Citation
Using inductive inference of past performance to build strategic cognitive adversary models

Material Information

Title:
Using inductive inference of past performance to build strategic cognitive adversary models
Creator:
Walczak, Steven Michael
Publication Date:
Language:
English
Physical Description:
xi, 144 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Board games ( jstor )
Chess ( jstor )
Decision trees ( jstor )
Educational games ( jstor )
Heuristics ( jstor )
Knowledge bases ( jstor )
Learning ( jstor )
Machine learning ( jstor )
Modeling ( jstor )
Numbers games ( jstor )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1990.
Bibliography:
Includes bibliographical references (leaves 135-143).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Steven Michael Walczak.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
026068619 ( ALEPH )
AHY9134 ( NOTIS )
24880978 ( OCLC )

Downloads

This item has the following downloads:


Full Text








USING INDUCTIVE INFERENCE OF
PAST PERFORMANCE TO BUILD STRATEGIC
COGNITIVE ADVERSARY MODELS












BY

STEVEN MICHAEL WALCZAK


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA


1990











ACKNOWLEDGMENTS


I would like to thank the faculty and staff of the Computer and Information

Sciences Department. Several staff members proved to be invaluable in helping me

cope with the university system. I specifically recognize the assistance of Ms. Jeffie

Woodham, Ms. Marlene Hughes, and Ms. Cheryl Coleman. Ms. Tammera Reedy

and Ms. Judy Longfellow were especially helpful in getting various figures and papers

printed and delivered to their final destinations. Professor Joseph Wilson provided

me with insight into the vision based problems encountered while performing my

research. Other faculty members including George Logothetis, Richard

Newman-Wolfe, Gerhard Ritter, and Ravi Varadarajan encouraged me to openly

discuss research issues in computer science and to enjoy being a student.

Two computer and information science graduate students helped significantly

during the early stages of my research. I would like to thank USA Captain

Christopher Wagner and USN Lieutenant Richard Fiola for there assistance in

uncovering the adversarial modeling techniques used by the armed services.

I would like to give special thanks to my committee members--Manuel

Bermudez, Walter Cunningham, Douglas Dankel, and Ira Fischler. Dr. Cunningham

has been a tremendous source of knowledge and help in establishing the techniques

used by the adversary modeling methodology to infer adversary playing styles. I owe






a great debt of gratitude to my committee chairman, Paul Fishwick. He spent many

tireless hours reviewing and analyzing my research and encouraging me.

My greatest human source of aid, comfort, and encouragement during my

graduate studies has been my wife. Karen gave three years of her life to work and

get us both through graduate school. She has my undying love.

I thank God for giving me the wisdom, intelligence, and motivation to pursue

my doctoral degree. The knowledge of His presence in my life has been the single

most important factor in my ability to complete graduate school. I hope and pray

that I will be faithful and wise enough to hear His calling in my life, that I may

choose to follow His path.












TABLE OF CONTENTS



ACKNOWLEDGEMENTS ...................


LIST OF TABLES ............

LIST OF FIGURES ...........

ABSTRACT ................

CHAPTERS


............................ .... vii


1 INTRODUCTION .......


Problem Statement .....................
Contribution of the Dissertation ...........
Outline of the Dissertation ...............

2 BACKGROUND AND RELATED WORK .....

Induction .............................
Tree Search ..........................
Athletic Adversarial Domains .............
Education Domain .....................
Adversarial Game Domains ...............
Law Enforcement Domains ...............
Military Domains ......................
Human Adversary Modeling Efforts
Domains ...................
Computer Planning in Military Domains
Political and Economic Domains ...........


Military

. .


Religion Domain ...................................
Chess Domain--Our Application Area ...................
Human Use of Adversary Modelin in Chess .........
Importance of Pawn Structures and Chess Openings ...
Philosophy ..................................
Psychology ..................................






Computer Chess and Artificial Intelligence .......... 38
Justification for Chess Domain Application .......... 43

3 ADVERSARY MODELING METHODOLOGY ............. 45

Knowledge Acquisition of Chunks ................... 46
Defining the Chunks to be Acquired ............... 46
Acquiring Geometric Chunks .................... 50
Acquiring Textual Chunks ....................... 58
Induction to Learn Repeated Chunk Patterns .............. 62
Acquiring an Adversary's Playing Style ................... 66
Application of the Adversary Model .................... 68

4 IA M ............................................... 75

Overview of IAM's Execution ......................... 75
Knowledge Acquisition and Inductive Learning Stage 76
Predicting Adversary Actions Stage ................ 82
Integrating IAM's Knowledge into a Chess Program ... 86
IAM's Performance Results ........................... 88
General Performance Results .................... 88
Performance with Respect to Time ................ 99
Performance with Respect to Textual Chunk Size .... 102
Performance with Respect to Geometric Chunk Size 104
The Effect of Domain Specific Knowledge .......... 107
Repeatability of the Demonstrated Results ......... 108

5 RESU LTS .......................................... 110

Reducing Opening Book Size ........................ 111
Revealing New Search Paths in the Game Tree ........... 113
Reducing Search Complexity Through Automatic Pruning ... 116
Summary of Research Contributions ................... 122

6 CONCLUSIONS AND FUTURE RESEARCH .............. 124

Conclusions ...................................... 124
Future Research Directions .......................... 125
Acquire More Geometric Chunks ................ 125
Increase the Knowledge Contained in Each Chunk .. 126
Place More Adversarial Knowledge in the Adversary
M odel ............................... 128

APPENDIX ............................. .......... ........ 132






REFERENCES ............................................. 135

BIOGRAPHICAL SKETCH ................................... 144












LIST OF TABLES


1. IAM Performance Measures for TEST1 ......................... 91

2. Effect of Likelihood on Prediction Ratios ........................ 96

3. IAM Performance Measures for TEST2 ......................... 97

4. Performance with Responsive Modification ....................... 98

5. Effect of Recency on IAM Performance for TEST1 ................ 100

6. Effect of Aging/Forgetting on IAM's Performance ................. 101

7. Performance (TEST1) with 10 Opening Moves .................... 103

8. Performance (TEST1) with 15 Opening Moves .................... 103

9. IAM Performance (TEST1) with Five-by-Five Chunks .............. 105

10. IAM Performance (TEST1) with Six-by-Six Chunks ................ 105

11. Effect of Domain Specific Knowledge .......................... 108

12. Ratio of Won and Lost Games for White and Black ............... 111

13. Statistical Analysis of Soviet Chess Weaknesses ................... 130











LIST OF FIGURES



Logic flow diagram. ..........................

Overview of the physical system .................

Sample Minimax tree. ........................

Sample cognitive chunk. .......................

Computer chess program ratings. .................

Chunks acquired for Botvinnik. .................

Internal representation of a board position. ........

Convolution templates .......................

Pseudo-code to perform convolution. .............

Step functions ..............................

Pseudo-code to perform step function ............

Pseudo-code to perform chunk coalescing. .........

Chunks on a chess board and their convolution values.

Chunk acquisition in Go.......................

Flowchart to collect textual move knowledge .......

The effect of noise on chunks. ..................

Four identical chunks. ........................

Fuzzy logic used by the heuristic rules ............

viii






19. An example of a game record--the 1948 Hague-Moscow Tournament. .. 77

20. Chunk occurrence in three different games ....................... 79

21. Pseudo-code to perform induction on geometric chunks ............. 81

22. Predictions from four chunks. ................................ 86

23. A game from the 1963 Botvinnik versus Petrosian match ............ 89

24. Another game from the 1963 Botvinnik versus Petrosian match ....... 90

25. Pseudo-code to suggest possible adversary moves. ................. 94

26. Chunk acquired with 5x5 size and corresponding smaller chunk ....... 106

27. Hypothetical game/search tree. ............................... 114

28. Prediction of Botvinnik's next move. .......................... 116

29. Standard search tree and the IAM improved search tree ............ 118

30. Search tree obtained from Correct Piece identification. .............. 121

31. Knight template and partial Rook template. ...................... 127











Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

USING INDUCTIVE INFERENCE OF
PAST PERFORMANCE TO BUILD STRATEGIC
COGNITIVE ADVERSARY MODELS

By

Steven Michael Walczak

December 1990


Chairman: Paul A. Fishwick
Major Department: Computer and Information Sciences

To perform optimally in adversarial domains, artificial intelligence programs

must be able to identically evaluate domain situations and develop plans as the

adversary. Knowledge about an adversary's evaluation criteria and objectives is

difficult to obtain since an adversary will guard this information in order to promote

a favorable outcome for himself. We present an adversary modeling methodology

which is capable of acquiring the evaluation criteria of an adversary through

observations of the adversary's past performances in the domain.

Our adversary modeling methodology uses induction to capture perceptual

chunks that are repeatedly displayed by a specific adversary. The acquired chunks

are then used to predict the adversary's actions during future encounters. Pattern

recognition is used to identify perceptual chunks in the current adversarial domain






situation which are analogous to chunks that have been previously learned from the

current adversary.

We have implemented our adversary modeling methodology in the program

IAM, Inductive Adversary Modeler, which has been applied to the domain of chess.

When high quality knowledge is available about an adversary's past performances,

the adversary modeling methodology accurately predicts ten percent of an adversary's

future action plans. The ability to accurately predict an adversary's move choices

reduces the complexity of the game tree search which enhances the playing ability

of current chess programs.











CHAPTER 1
INTRODUCTION

Adversarial situations surround us in varied forms on a daily basis. Examples

of adversarial situations include vying for a promotion at our place of employment,

Monday night's football game, and military conflicts throughout the world. These

adversarial situations are characterized by a minimum of two actors each of whom

is attempting to realize a goal that conflicts with the goal of the opposing actor. The

actors can be individuals or groups of individuals such as a football team players that

are cooperating to achieve the goal of the actor.

When humans are faced with the problem solving task of realizing a goal

against the desires of an adversary, they employ a variety of techniques that enable

them to outperform their adversary and accomplish their goal. The standard

cognitive problem solving methods of difference reduction and working backwards

are insufficient in adversarial domains. Human competitors use deductive reasoning

and analogy to previous domain situations to realize their goals in adversarial

domains. The use of analogy to gain a strategic advantage over an adversary

necessitates the studying of a specific adversary's previous performance, preferably

in situations analogous to the upcoming contest.

Human competitors performing in adversarial domains study the past

performances of their adversaries to identify strengths and weaknesses which can be






2

exploited respectively to gain a strategic or tactical advantage. This information is

preserved by the competitor as a model of the likely actions an adversary will

perform in particular situations. The model of the adversary's probable actions is

then exploited when an appropriate situation occurs during the course of the

competition. Such situations arise when the current game state--situation in the

adversarial domain--is analogous to a game state which occurred in the adversary's

past.


1.1 Problem Statement


Current artificial intelligence application efforts in adversarial domains are not

performed optimally. A factor which keeps adversary-based programs from

performing at expert levels is the lack of knowledge about the potential adversary.

Lenat and Feigenbaum (1987) claim that domain specific knowledge is required to

solve difficult problems and that the performance of intelligent programs improves

as additional domain knowledge is utilized.

Acquiring knowledge for programs in adversarial domains is hindered by the

fact that adversaries are unwilling to reveal their tactical and strategic goals, since

this would be counter-productive to the attainment of their goals. Candland (1980)

and Bolc (1987) have claimed that knowledge about any domain is available from a

wide variety of sources. Lenat (1982) proposes the use of the observable actions of

the adversary while performing in the adversarial domain. The proper selection of

adversarial domain situations to observe guarantees the acquisition of high quality

knowledge.






3
For the domain of chess, lack of knowledge about an adversary's tactics and

strategies is analogous to searching the game tree to a depth of only one ply. All

moves are considered as winning moves and the game proceeds in a random manner.

To acquire knowledge for use in adversarial domain application programs, we will

emulate the traits of human experts performing in adversarial domains, by analyzing

the previous performances of a particular adversary in relevant domain situations.


1.2 Contribution of the Dissertation


Adversarial domain programs require knowledge about an adversary's tactics

and strategies to continue to improve their performance. We need to consider two

separate factors that affect the strategic decisions an adversary will make. An

adversary's strategic decisions are affected by the plans or scripts that the adversary

feels will satisfy the current strategic goal. Before the adversary can select a plan of

action however, the domain situation must be evaluated to determine which plans are

appropriate to that particular situation.

Current programs performing tactical and strategic planning for adversarial

domains rely heavily on tree structures and search techniques such as iterative

deepening and alpha-beta pruning. The ability to accurately predict the choices that

will be made by an adversary in specific domain situations reduces computational

complexity and increases the depth of search. The increased depth of search that our

research enables is directly responsible for the augmented tactical ability of these

search based programs.






4

The purpose of our research is to develop a generalized method for modeling

the strategic decision making process of an adversary. The information flow diagram

for our adversary modeling methodology is shown in Figure 1. Our adversary

modeling methodology makes use of knowledge which is currently not used by

programs operating in adversarial domains. By increasing the knowledge available

to tactical and strategic planners, we will improve the overall performance of these

adversarial domain applications.


Repeated Patterns Adversar|y I|
*Adversary

Sct ion Predictor

Current Adversar ial Analogous Situation
Domain Situation Detector



Figure 1: Logic flow diagram.

Our adversary modeling methodology uses psychological principles of

organization and evaluation such as the Gestalt principles of organization (Anderson,


Record of
Previous Performances


ADVERSARY nODEL
by the Adversary -
S Knowledge Base
of Patterns
Knowledge Acquisition
to Acquire Patterns noun bg
-I Adversar ies

Inductive Inference
to Identify & Learn I


I


I






5
1980, and Pomerantz, 1986) to emulate the primary evaluation criteria used by an

adversary. Kodratoff and Michalski (1990) note that recent machine learning efforts

have focused on building cognitive learning architectures. We intend for the

adversary modeling methodology to be applicable across all adversarial domains and

hence, we include as little domain specific knowledge as possible. This means that

we are content to acquire the primitive elements that an adversary uses in evaluating

a domain situation without embellishing those elements with further domain specific

knowledge.

The psychological principles of organization are applied to a collection of

previous adversary performances in the domain (e.g., chess games played by the

adversary in prior tournaments and matches). We acquire the cognitive chunks used

by an adversary to evaluate the current domain situation. The acquired cognitive

chunks of the adversary help us to view the domain the same as our adversary and

to avoid influencing domain evaluations with personal biases.

We now have a large collection of chunks that have been displayed by the

adversary. The chunks are either textual patterns corresponding to the adversary's

verbal memory or geometric patterns corresponding to the adversary's visual memory.

The chunks we have acquired contain patterns that have been used only once and

may have occurred by chance. We use induction to eliminate these chance chunks

and create a valid and useful adversary model. The inductive phase identifies chunks

which are repeated in two or more games. The repetition of patterns is used to

indicate the probability that an adversary will use the acquired chunk in future






6

evaluations of domain situations. Because we are using induction as our learning

method, the adversary model created will be used as a heuristic guide instead of an

algorithmic solution.

Following the inductive learning, the collection of chunks that have been

repeatedly displayed by the adversary are stored in the adversary model. We

consider these chunks to be the primitive elements used by an adversary in evaluating

current domain situations. The chunks of the adversary model are then used to

predict adversary decisions when analogous situations occur in the adversarial

domain.

The knowledge obtained by the adversary modeling methodology is highly

reliable, because it comes directly from the adversary. A certain level of care must

be taken to guarantee that the adversary is performing normally in the domain and

not providing misinformation to the modeling methodology. We verify the adversary

modeling methodology by implementing the methodology for the adversarial domain

of chess. The implementation is called IAM, for Inductive Adversary Modeler. Our

justification for selecting the chess domain is given in section 2.10.6.

Berliner and Ebeling (1988) note that speed is a critical factor for evaluation

functions in chess as well as other adversarial domains. The knowledge that we are

learning about an adversary is available as historical records prior to the current

performance of the adversarial domain program. Therefore, the adversary modeling

methodology's knowledge acquisition can be performed prior to its use by the domain

program. This preprocessing effectively eliminates any time or complexity costs






7
generated by the adversary modeling methodology from being added to the current

program's evaluation function cost (Aho, 1974). The adversary model acts as a

heuristic coach to an existing domain program by predicting which moves an

adversary will make, as shown in Figure 2. The only part of the adversary model that

impacts the time order of the existing evaluation function is the analogous situation

detector that was shown in Figure 1.


Current
Game


Strategic


I
Advice J

Existing
Game Plaging Program

[__-


Figure 2: Overview of the physical system.


1.3 Outline of the Dissertation


We have presented a general overview of the problem of the lack of

knowledge about adversaries available to intelligent programs operating in






8

adversarial domains. Our solution to this problem has been briefly presented in this

chapter.

In Chapter 2, we present background material on induction and psychology

that affects our research. The research that has already been performed in several

of the adversarial domains including the domain of chess is discussed. Finally, we

present the reasons for selecting the domain of chess as our application domain.

In Chapter 3, we detail our research by first discussing the implementation

method and then the operation of the adversary modeling methodology. Each of the

acquisition and induction steps that have been briefly described in this chapter are

presented in detail.

Chapter 4 contains an example of the adversary modeling methodology in

action and provides quantitative performance results of our application program,

IAM. Our results and their impact on artificial intelligence search methods are

provided in Chapter 5. We discuss our conclusions and future research directions in

Chapter 6. The Appendix contains a glossary of chess and artificial intelligence

terms to aid the reader in understanding some of the domain descriptions that we

have presented.











CHAPTER 2
BACKGROUND AND RELATED WORK

In this chapter, we review the theoretical and applied work on induction. We

also describe the theoretical work on game tree search to emphasize our contribution

to this research area. Finally, we describe related work in the various adversarial

domains that have led to our solution with a detailed review of the domain of chess

which serves as our application domain. For each of the adversarial domains

presented, we describe the use of adversary modeling by humans and any applied

research that affects the implementation of our adversary modeling methodology.


2.1 Induction


Mankind has been using induction to perform problem solving and learning

tasks for many millennia. Rosch (1973) describes the way humans form concepts and

categories about their natural world. Polya (1973), in the early 1950s, formalized the

concept of induction. Induction is the process of discovering general laws or concepts

from observations and combinations of particular instances.

An example of induction for a natural world concept would be the

generalization that all ravens are black. This generalization is based on the

observation of a limited number of instances of a raven. Without seeing all the

possible examples or instances of raven, we are confident to induce that the color of






10
a raven is the same as the color of the instances we have seen--black. The strength

of our belief increases as more and more examples of black ravens are observed.

However, only one instance of a nonblack raven needs to be identified to invalidate

our induction.

From the above example we can see that inductive inference is learning in the

limit (Daley & Smith, 1986). We need to see sufficient observations of the right type

to form a correct generalized concept. Since we cannot guarantee that the right type

of observations are being provided to an inductive inference system, we must rely on

the use of a large number of examples of category instances to acquire the desired

concept. This means that the adversary modeling methodology we have developed

should perform more accurately as we provide greater numbers of examples of an

adversary's domain performance.

Since Polya's formalization of the concept of induction, theoretical research

on induction has primarily focused on the formation of concepts or classifications.

Angluin and Smith (1983) and Case and Smith (1983) present rigorous discussions

of the mathematical principles of induction. This theoretical background provides

a solid base for the use of inductive inference as a computational tool for learning

classes of objects or ideas.

The extraction of general principles from specific examples is a major research

goal in machine learning (Anderson, 1987). Induction has been applied to many

domains. Examples of applied inductive inference algorithms are Michalski's

INDUCE and Quinlan's ID3 (Michalski et al., 1983 and 1986). Each of these






11
example applications attempts to classify instances of a domain problem (e.g.,

soybean diseases for the INDUCE program) into maximally-specific groups that will

identify all known instances as well as accounting for unknown instances.

Prior to the theoretical discussions of inductive inference mentioned above,

Blum and Blum (1975) had already demonstrated that the ability to play games, or

at least to make technically correct moves, can be learned through inductive

inference. Induction has not been utilized in game domains due to the lack of high

quality play produced by Blums' algorithm. Our adversary modeling methodology

focuses on a narrower slice of the game domains, namely predicting adversary move

choices and playing style, and is able to learn high quality knowledge.

Adversarial domains present several specialized problems to the standard

inductive inference algorithms like ID3. Langley (1985) has noted that feedback is

an essential element of most learning methods. Inductive inference algorithms use

feedback to verify acquired concepts and to identify concepts that need to be

modified. In adversarial domains, feedback on the performance of an algorithm is

a not attainable until after the current domain situation has ended. Because of the

inherent time delay for feedback in adversarial domains, we substitute probabilistic

reasoning so that our confidence in the current adversary model increases as

additional examples of the same type are observed.

Another problem in adversarial domains is the dependence of the domain on

time. Actions occurring in the domain are temporally related. The inductive

inference applications cited above belong to the class of hierarchical or structure






12
induction methods. Structure induction methods are used to produce classification

systems. Muggleton (1990) proposes an inductive method called sequence induction

which has a time effect relationship to the domain. Sequence induction uses a series

of domain descriptions that are altered by actions operating on the domain to

produce control systems.

Mitchell et al. (1986) refer to a process of mechanized induction which

receives domain events as input and produces concepts for predicting future events.

For adversarial domains, sequence induction provides a method to produce

evaluation criteria for predicting an adversary's strategic decisions.


2.2 Tree Search


Programs in the adversarial domains of games invariably use trees as their

knowledge representation scheme. Finding solutions or move choices in game trees

involves searching the tree for the optimal finishing value. Von Neumann and

Morgenstern (1953) published the Minimax algorithm for backing the final game

values up through the game tree to the current node. Minimaxing attempts to

account for the effects of an adversary in the game domain by alternately raising the

minimum possible value or maximum possible value at each level of the tree.

If an exhaustive search can reach all of the leaf nodes of the tree, then the

true finishing value of the game can be raised to the current node and an optimal

game will ensue. However, games like chess, Go, and Othello have tree

representations that hold 1040 nodes or more. Game trees of this size prohibit an






13

exhaustive search. Instead, game programs will search down to a preselected depth

and apply an evaluation function that will determine a best guess for the true

finishing value obtainable from that node.

Evaluation functions use domain knowledge to predict the outcome of the

game from a specific position. All domain specific knowledge is contained in the

evaluation function. The quality of knowledge has a direct effect on the playing

ability of the game program (Berliner & Goetsch, 1984, and Lee & Mahajan, 1988).

Northwestern University's CHESS4.6 was able to defeat other chess programs which

searched several ply (levels) deeper into the game tree based on the strength of its

evaluation function (Elithorn & Banerji, 1984, and Wilkins, 1980).

The primary problem faced by search algorithms using tree structures is the

horizon effect. The horizon effect occurs when the solution to a search problem lies

at a depth in the tree that is beyond the current search depth. This is a realistic

problem for complex adversarial domains. In chess, the KRKN endgame, king and

rook versus king and knight, can require a twenty-seven move solution or a game tree

search depth of fifty-two ply (Bratko & Michie, 1980). Similarly, the KBBKN

endgame has solutions of sixty-six moves or one hundred thirty-two ply search depth

(Roycroft, 1988). Current chess programs search between eight and fourteen ply

before selecting a move.

Two different schools of thought exist concerning methods for continuing to

improve game program performance. The first school believes that the depth of the

search is the sole criterion affecting the playing ability of game programs.






14

Adelson-Velsky et al. (1988) state that increasing the depth of search of a game tree

increases the quality and reliability of game play. Berliner and Ebeling (1988)

specifically state that a deep tree search will outperform a shallower search which

uses more knowledge. Various techniques are used to increase the depth of search

for a program. Limiting the breadth of search and pruning of undesirable branches

through alpha-beta pruning and iterative deepening permit a deeper search for a

static number of evaluated tree nodes (Barr & Feigenbaum, 1981). Specialized

hardware can also be used to increase the speed and thereby the depth of search

(Ebeling, 1986).

An example of alpha-beta pruning is shown in Figure 3. The four leaf nodes

which have their branches cut by a diagonal line can each be pruned using alpha-beta

pruning. Once the three value is elevated to the left subtree level two node, then the

level one node is guaranteed a minimum value of three. Since each of the first leaf

nodes below the center and right subtrees have lower values than three, the

remaining leaf nodes do not need to be evaluated. This is called an alpha cutoff.

A beta cutoff is similar except that the cutoff is performed based on a guaranteed

maximum value for a Min node.

The other method for improving the performance of game programs is to

increase the knowledge that is available to them. We have already seen that better

quality knowledge enabled CHESS4.6 to outperform other programs that searched

several ply deeper into the game tree. The use of additional knowledge has the same

effect as searching deeper into the game tree. The projection ability of certain types




























Figure 3: Sample game tree with alpha-beta pruning.

of knowledge in the chess domain is nine ply for tactical knowledge, twenty-five ply

for positional knowledge, and forty-five ply for strategic knowledge (in Elithorn &

Banerji, 1984). Current game programs use mostly tactical knowledge along with

some positional knowledge.

For adversarial domains, the use of domain knowledge can be dangerous by

leading us into a false sense of security. Samuel (1959) warns that programs using

the Minimax algorithm must take into consideration the intent of the adversary.

Cecchini (in Crookal et al., 1987) claims that the actual intent of the adversary will

require programs to use different heuristics in order to win. Human tree search is

goal directed and the direction and method of search may change as new information


LEVEL

1





23


.(






16

becomes available (Frey, 1983). This last statement means that the preceding events

in a game and tournament affect the future strategic decisions of the adversary.

Our adversary modeling methodology follows the school of thought which

believes that more knowledge is required to continue improving the performance of

game programs. To avoid the complications of adversarial intent, our heuristics are

based on the primitive psychological elements that are used by an adversary to

evaluate the domain. Application of the adversary modeling heuristics occurs when

the current game board pattern is analogous to a previously induced situation. The

use of pattern recognition to increase search depths has already been implemented

by Samuel (1959 and 1967) in the form of signature tables and in the Russian chess

program KAISSA (Adelson-Velsky et al., 1975) to perform cutoffs or pruning.

The theoretical increase in search depth which we have achieved through the

adversary modeling methodology is detailed in Chapter 5. Several authors (Findler

& Meltzer, 1971, and Utgoff, 1986) have noted that the economy of a search

heuristic can be measured by the reduction in search effort and complexity which

produces a deeper search.


2.3 Athletic Adversarial Domains


The modeling of an adversary by humans is well documented in athletic

domains. Coaches of baseball (Aliston & Weiskopf, 1984), football (Riggins &

Winter, 1984), and other sports use scouting reports and films of their opponents to

study their opponent's strategies. This enables the coaches of athletic teams or






17

individuals to develop counter-strategies to foil the goals of the adversary and

promote their own goals.

In a personal communication, Owen J. Holyoak of the Exercise and Sport

Sciences Department at the University of Florida stated that the amount of

preparation performed by athletic team coaches corresponds directly to the perceived

threat of the adversary. When the technique of studying an adversary's past

performance is not fully used, surprising results can occur. At the 1982 homecoming

football game for the University of Florida (UF), the opponent was West Texas State

University (WTS). Although WTS was considered to be an underdog and was not

expected to score any points, they obtained a fourteen point lead early in the game

by using some plays that UF had not anticipated. The lack of preparation on UF's

part was responsible for the surprising start of the 1982 homecoming game.


2.4 Education Domain


The artificial intelligence domain of intelligent tutoring systems (ITSs) or

intelligent computer-aided instruction has already made successful use of models of

human thinking. These models are based on the availability of knowledge to the

human and that human's mental set at a particular point in time. The people being

modeled in ITS programs are the students who are attempting to learn. The student

models are used to predict the sources of error in student thinking which cause

incorrect responses to test items produced by the ITS.






18

Woolf (1984) has demonstrated an ITS which uses a model of the probable

knowledge currently held by the student for discourse planning. This model adapts

its understanding of the student's knowledge during the lesson to capture changes in

the student's cognitive state precipitated by the acquisition of new knowledge. Our

adversary modeling methodology operates similarly by constructing a model of the

probable evaluation mechanisms known to an adversary to predict future strategic

decisions of the adversary. Other examples of this modeling approach as used in

ITSs can be found in Sleeman and Brown (1982).


2.5 Adversarial Game Domains


Because of their finite size, games provide us with a convenient abstraction

of more difficult real world problems. The former Japanese use of the game of Go

to train their military leaders (Reiss Games, 1974) demonstrates the practical

extension of a game domain to a military domain. Pritchard (1973) elaborates

further by claiming that Go is guerilla warfare on a grand scale.

We will be using two person zero-sum perfect information games (Zagare,

1984) to demonstrate the adversary modeling methodology. Zero-sum games are

always adversarial in nature, since the wins of one side are equivalent to the losses

of the other side. Our selection of perfect information games was made to permit

an algorithmic evaluation of the results. The domain of chess which we will use for

our application of the adversary modeling methodology is reviewed in section 2.10.






19

Other types of games, such as games of chance, that do not fit our selection

of zero-sum perfect information games can also benefit from the adversary modeling

methodology. Our contribution which reduces the complexity of search is usable by

any game program that utilizes game tree search and heuristics to solve domain

problems. Berliner's BKG backgammon program (in Levy, 1988) is an example of

a game of chance which uses heuristic methods and game tree search to select

between moves.

Christensen and Korf (1986) claim that heuristic methods for one person

puzzles are equivalent to heuristic methods for two person games. The adversary

modeling methodology is capable of predicting moves in one person games that will

reduce the complexity of the current game position to one which has a known

solution. Other research using induction of previous games has been performed for

one person games. PREFER (Utgoff & Saxena, 1987) solves the 8-puzzle by

performing induction on previous 8-puzzle solutions.

Samuel (1959 and 1967) was the first researcher to use machine learning to

create an intelligent game playing program. Samuel's checker playing program uses

rote learning techniques to learn how to play checkers. This program is capable of

imitating the style of an adversary by mimicking the winning moves of the adversary

presented in the set of training instances. Programs which rely on rote learning tend

to become inflexible and require an equally large number of retraining instances to

adapt to a different style of play. The method of mimicking an adversary has been

shown to fail for the game of Go (Pritchard, 1973) and should not be considered as

a generalized method of strategic planning against adversaries.






20

Samuel developed signature tables to reduce search complexity. Signature

tables are used to remember the evaluation function value for specific positions.

When identical positions are identified through a pattern matching mechanism, the

signature tables are used to retrieve the corresponding node value. Wilcox (1985)

has also applied pattern recognition techniques to the game of Go to construct

NEMESIS, a strong Go playing program.

IAGO (Rosenbloom, 1982) uses the standard search techniques of alpha-beta

pruning and iterative deepening to play championship-level Othello. Rosenbloom

points out that the heuristics for IAGO need to change during different phases of the

game to capture shifts in the strategic importance of various game principles. For

example, in the middle game of Othello mobility is a critical factor, but in the end

game portion of Othello the quantity of stable pieces is the primary concern.

Berliner (1979) notes similar shifts in strategic reasoning for the domain of chess.

Our adversary modeling methodology compensates for the strategic shifts at various

points in the game by assigning time durations to the domain evaluation chunks that

are acquired for each adversary.

GINA (DeJong & Schultz, 1988) tries to learn to play Othello by using

experience gained from previous games. This approach is very similar to the one

used by Samuel for his checkers program. GINA attempts to exploit the weaknesses

of a specific opponent. Several of GINA's implementation features are applicable

across adversarial domains. These features are:






21

Serve as an external coach to an existing program in the domain.

Only use observable behavior to learn about an adversary. For GINA,

the observable behavior is the moves made by an adversary in prior

games.

Do not use feedback to modify learning, except for the results of the

games that have been studied.

The best example of strategic planning based on an adversary's modus

operandi is Waterman's (1970) poker player. Waterman has implemented a draw

poker playing program which analyzes an opponent's style of play to decide when a

bet is appropriate. This program learns over a period of several hands which

heuristics to use against a particular adversary. We expand upon Waterman's idea

by not limiting the program to a static number of predefined heuristics.


2.6 Law Enforcement Domains


Police departments and the Federal Bureau of Investigation each make use

of adversary modeling. As an example, a police detective will try to predict when,

where, and to whom a serial killer will perform his next crime. This is done

primarily by analyzing the criminal's previous performances of crimes. Similar

techniques are being considered for predicting the outcome of possible terrorist

activities in the RISK project sponsored by Lawrence Livermore National

Laboratories.






22

The key to successful drug enforcement is intelligence gathering and

information processing (Witkin, 1990). Raw data is voluminous and must be

converted to a useable form. Our adversary modeling methodology is capable of

finding patterns in raw data that correspond to resource movement problems (e.g.,

the actual drugs and the monetary compensations) and can be used by law

enforcement officials to increase their effectiveness.


2.7 Military Domains


Devastating results occur on the battlefield when sufficient preparation against

an adversary is not performed. Due to different sociological, geographical, and

educational backgrounds, potential adversaries will have unique behavioral rules

(Bond, 1986 and Boring, 1945). These behavioral rules govern varying beliefs and

goals concerning war and the use of nuclear weapons (McMillen, 1984 and

Sokolovskiy, 1968). Dunnigan (1982) and Ermarth (1978) state that if we assume

that an enemy will act strictly according to our own beliefs, then we will produce a

tactical blindness. Tactical blindness produces the inability to predict or fathom the

plans of our adversaries. Historic examples of the tactical and strategic blindness

produced by conscripting our own beliefs onto an adversary are the bombing of Pearl

Harbor by the Japanese in 1941, the defeat of the United States tactical forces at the

Chosin Reservoir during the Korean War (Wallace, 1990), and the Tet Offensive of

the North Vietnamese in 1968. Many soldier's lives might have been saved early in

World War II if the United States soldiers had understood the Japanese attitude

toward surrender (Boring, 1945).






23

2.7.1 Human Adversary Modeling Efforts in Military Domains

Military leaders (Gruner, 1988, and Ryan, 1988) claim that victory is often

contingent upon knowing your enemy. This means that military planners must be

able to evaluate the current situation using the same criteria as their adversary. A

vital component of intelligence information is briefings on an adversary's operational

characteristics or tactics (Gruner, 1988). Robert E. Lee, who is often cited as the

greatest military strategist (B. Davis, 1956), is a practical example of the power

accompanying detailed knowledge about an adversary. Lee was educated at West

Point, where all of the Union commanders were also trained, and served along side

many of the adversaries he would face in the Civil War during his tenure as an

officer in the United States Army and Cavalry (Snow, 1867). Lee's detailed

knowledge of the training and tactics of his adversaries enabled him to frequently

outperform better equipped and larger forces.

The United States military makes use of this concept in training our soldiers.

Each service branch maintains a cadre of personnel who are trained in Soviet-style

war tactics and strategy (Robbins, 1988). The training centers for the Army,

Air Force, and Navy are the National Training Center, Red Flag, and the Top Gun

school respectively. The purpose of these Soviet analogous fighting forces is to

provide a realistic training environment to better prepare the military for the

possibility of an actual war against Soviet forces or forces that use Soviet-style tactics.

A historic example of adversary modeling for military purposes occurred

during World War II. A detailed psychological profile of Hitler was constructed by






24

Langer (Langer, 1972) for the Allied forces. This model of Hitler was constructed

solely from second-hand information. The psychological model of Hitler accurately

predicted the method of his ultimate demise via suicide.

2.7.2 Computer Planning in Military Domains

Military games and strategic and tactical planners make use of varying levels

of adversary modeling, ranging from none in typical multi-player games and training

simulators to moderately advanced models in certain strategic planning programs.

Failing to account for the actions of an adversary severely inhibits strategic planning.

Military planning tools must account for the probable actions of an adversary based

on the adversary's intentions and strategic style, otherwise tactical blindness will

result.

The tactical troop movement planner TACPLAN (Andriole et al., 1986) is

typical of most military planning tools. TACPLAN only accounts for static elements

of the domain such as the effect of terrain on equipment movement. Knowledge

about adversaries is nonexistent. The next step towards an adversary model is

programs similar to ARES (Young & Lehner, 1986) which acknowledges the

presence of an adversary. ARES is also a tactical planner; however, the plans

generated by ARES attempt to account for the possible blocking actions of an

adversary. Although this approach claims to account for the actions of an adversary,

the actual outcome is not that different from the plans generated by programs like

TACPLAN. The lack of difference between the two approaches is caused because

the actions of an adversary are prescribed in advance which is similar to only






25

permitting a chess adversary to make forward moves. Infinitely many strategic

choices can be made by an adversary that will not be covered by the prescribed

action choices that are permitted by the tactical planning algorithm.

The POLITICS system (Carbonell, 1981) proceeds in the development plan

by introducing goals that an adversary is trying to accomplish, such as blocking the

current goal of the planning program. The adversarial goals used in POLITICS are

statically defined and are based on preconceived notions of an adversary's intent.

Basing the actions of an adversary on static predefined goals runs the risk of creating

tactical blindness mentioned in the introduction to section 2.7. The actual goals of

an adversary, much less how the adversary intends to accomplish those goals, may not

be known during the contest or may be contrary to the expected goals (Crookall et

al., 1987 and Narendra & Thathachar, 1989). POLITICS does provide a good

example of a program in which a thinking adversary attempts to achieve specific

goals.

TM (Erickson & Zytkow, 1988) is another goal based system for making

tactical decisions. The program BOGEY is an external learning program which

generates large numbers of simulations of tactical engagements with specific goals.

BOGEY, like GINA in section 2.5, is another example of the utility of coaching

programs that support existing domain specific programs through an external learning

function.

The MARK I and MARK II systems (Davis, 1988a and 1988b and Davis et

al., 1986) embody the mind set of an adversary in DEWT, Deterrence Escalation






26

control and War Termination, models. Currently the MARK systems have two

decision models, East and West. While MARK I and MARK II attempt to capture

differences in the mind sets between adversaries, the models are statically defined

like the goals of the POLITICS program. These static mind sets are based on

predefined conceptions of an adversary's intentions which may not be accurate and

will not account for the modification over time to decision heuristics used by an

adversary. One of the contributions that artificial intelligence can provide to military

planning is reasoning that accounts for the temporal relevance of actions (Bonasso,

1988).

The use of multiple models to generate goals for different adversaries

demonstrated by the MARK systems is a reasonable approach to separate the varying

evaluation criteria used by various adversaries. An improvement to the MARK

systems is accomplished by acquiring the evaluation criteria and traits of particular

adversaries in specific situations along with the temporal relevance for each of the

evaluation criteria of an adversary.

We mentioned in section 1.2 that geometric or visual relationships between

domain resources are used by the inductive learning mechanism of the adversary

modeling methodology. This means that military domains are required to represent

the physical resources of the domain in a geometric manner. The services are

trained to use standardized formations such as the column, echelon, line, and wedge

formations (Department of the Army, 1986). Typical board wargames make use of

a hexagonal map for plotting the movement of resources. By using such map






27

representations, the use of specific formations by a particular commander can be

acquired by our adversary modeling methodology.


2.8 Political and Economic Domains


Top level policy decisions made by government officials are also affected by

prior knowledge about an adversary's goals and strategies (Cimbala, 1987).

Government negotiators attempt to discover the limitations faced by an adversary to

accomplish the optimal outcome of an agreement or treaty between the two

adversaries. The use of knowledge of the political mind set for a geographic region

permits negotiators to interpret information from the proper perspective (Cimbala,

1987).

The political use of adversary modeling can also be extended to corporate

politics. We can acquire patterns from adversarial actions such as corporate

takeovers or mergers. This knowledge enables corporate entities to be better

defended or prepared for such adversarial actions.

Other domains that are concerned with the geometric relationships of

elements in the domain can benefit directly from the geometric pattern acquisition

phase of the adversary modeling methodology. Examples of such domains include

city planning and architecture. For the city planning domain, patterns of streets from

other cities which have desirable and equivalent traffic patterns can be acquired for

use in the current city. The primary benefits to these domains will be economic

savings and time savings for the reduced complexity of the design phase for

individual projects.







28

2.9 Religion Domain


The importance of knowing the intentions and strategic style of an adversary

has been recorded throughout history. An early example of these writings comes

from the first century A.D. when the apostle Paul (1 Corinthians 9:19-22) writes:

that I might win the more. And to the Jews I became as
a Jew, that I might win Jews; to those who are under the
Law, as under the Law to those without law, as
without law ... .To the weak I became weak, that I
might win the weak.

Paul is stating that he will model the lifestyle of the people he is attempting to win.

The training centers for the military at the National Training Center, Red Flag, and

the Top Gun schools are modern examples of the approach used by Paul.


2.10 Chess Domain--Our Application Area


We mentioned in section 2.5 that games like chess can be thought of as

abstractions of actual military contests. Although the outcome of these games may

be viewed as trivial compared to the outcomes of military conflicts, personal

reputation, and financial prosperity are the stakes of such games for expert game

players.

In the remainder of this section, we demonstrate the use of adversary

modeling by humans in the chess domain and the philosophical and psychological

foundations that enable the adversary modeling methodology to operate in the chess

domain. Finally, we present a detailed background of past and current computer

chess programs.






29
2.10.1 Human Use of Adversary Modeling in Chess

Before playing for the World Chess Championship against Capablanca who

was considered to be invincible and was appropriately called "The Chess Machine",

Alekhine studied and analyzed Capablanca's prior games (Schonberg, 1973). From

his study, Alekhine was able to determine a weakness in Capablanca's play that he

was able to exploit to become the new World Chess Champion. Other chess masters

including the World Chess Champions Botvinnik, Tal, and Kasparov prepare for

chess matches by rigorously studying the previous games of their opponents

(Horowitz, 1973).

Nunn (Nunn & Griffiths, 1987) describes a game against Ost-Hansen in which

both players based their game strategy on expectations of their adversary's playing

style. These expectations were formed from prior experience against the adversary.

Although no one was able to capitalize upon the information, it has been noted that

prior to his World Championship match against Spassky, Fischer varied his standard

opening for white from e4, pawn to King four, only three times during tournament

level play (Evans, 1970).

Several of the chess grandmasters, including Lasker and Tal (Horowitz, 1973

and Schonberg, 1973), have been described as psychological players. These players

would intentionally make the move that would be most disturbing to their opponents,

even though such moves were frequently not the best tactical choice. Levy (1984)

has described his style of play against computer chess programs similarly, stating that

his move choices are often made to confuse the computer.






30

The advent of intelligent programs in various game domains, including chess,

has caused humans expert game players to consider the computer as an adversary.

Subsequently, human competitors who will face an intelligent computer program as

an adversary have created models of adversary strategy and playing style that

represent the computer game programs. Certain human competitors are considered

to be specialists against computer styles of play.

A famous example of this specialization is the series of bets which David Levy

(1984) has had with the computer chess community. In these bets, Levy claims that

no computer chess program will be able to beat him. The first bet which had a ten

year duration was won by Levy, however, DEEP THOUGHT has defeated Levy in

the latest of the bets. A recent discussion on the rec.games.chess news network by

Feng-Hsiung Hsu indicates an increase in the number of human opponents of

computer chess programs using "anti-computer" playing styles against programs

entered in major chess tournaments.

When Mike Valvo defeated DEEP THOUGHT, the reigning computer chess

program champion for the past three years, in a two game postal chess match via

electronic mail, it was noted (Newborn & Kopec, 1989) that Valvo had observed

DEEP THOUGHT play a number of games and had acquired the playing style of

DEEP THOUGHT. The chess program did not have a similar opportunity. Berliner

(1988) claims that various chess players have asked for the previous game records of

HITECH, the 1989 computer chess program co-champion, so that they might be

prepared for a possible match.






31
2.10.2 Importance of Pawn Structures and Chess Openings

We pay special attention to pawn formations during our chunk learning

process. Philidor has stated that the win or loss of a game depends on the quality

of the pawn positions during the game (Pachman, 1975). The strength and weakness

of pawn formations is of paramount importance (Lasker, 1973). Pawn formations can

be used to reveal information concerning positional advantages (Reinfeld, 1987).

Furthermore, three of the six criteria for determining the character of a position

given by Capablanca relate to the effectiveness of the pawns in the game (Pachman,

1975).

Pawn formations contain information about the positional nature of the chess

game (Soltis, 1976). The geometric patterns of pawns displayed by a specific

adversary can be used to identify his playing style. By definition we can recognize

open and closed board positions solely from the absence or presence of long pawn

structures. The general playing style of the adversary is indicated by the predominant

style displayed by the adversary in the studied games. Adversaries will tend to

remain with the particular style that they feel most comfortable in playing. Chigorin

played only open style games (Pachman, 1975). Our own analysis of the games of

Botvinnik and Spassky has indicated that each of these chess grandmasters prefers

a particular style, closed and open respectively.

Openings also play an important role in determining the positional advantage

of the game. Current computer chess programs utilize a book or database of

opening moves which have a duration from two moves to more than ten moves. The

evaluation algorithms which enable chess programs to estimate the strength of the






32

current game position are not used until the book is exited. Typically, exits from

opening books occur only when an adversary makes a move that is not contained in

the book or the current opening line of play is exhausted. This means that current

computer chess programs can find themselves from ten to twenty percent of the way

through a game, for a game lasting fifty moves, before they start to calculate the

value of the current position.

The opening sequences of play that are used as the opening book for current

chess programs are stored verbatim. Mednis (1990) has demonstrated that particular

opening formations can be achieved through a wide variety of the actual move order.

A statistical analysis of the opening move sequences displayed by an adversary

accounts for opening move order variations by finding the upper bound, lower bound,

and mean time of specific moves.

Selecting an opening line of play is usually performed at random or in

response to the adversary's choice of opening move. By studying the playing style

and previous games of an adversary, specific opening lines that are well known to an

adversary are identified. This knowledge can be used by a chess program to emulate

human chess experts by choosing openings that are not as well known to the

adversary. Nunn (Nunn & Griffiths, 1987) avoids the Pelikan variation of the Sicilian

Defense in one of his games against Anthony because Nunn knows from experience

that Anthony is a Pelikan specialist. Other examples of selecting opening lines of

play based on the prior performance of an adversary are given by Nunn.

Acquisition of the opening sequence knowledge held by the adversary will

enable current chess computer programs to select lines of play which are strategically






33
advantageous. Holding (1985) supports our perspective by stating that we should use

information about players playing style to disrupt their normal play.

2.10.3 Philosophy

Dennett (1987) states that current chess programs concentrate their time on

branches of a game tree which represent the adversary's best response as viewed by

the program. While this would appear to be a reasonable approach, Dennett

elaborates by stating that a rational adversary might have a higher ulterior motive

than avoiding defeat in chess. For example, a young Soviet chess player with a rating

of master is performing well in a European chess tournament. The young Soviet has

acquired a score of 8.5 points and is about to face the tournament champion, a fellow

Soviet grandmaster who has scored 11 points thus far. The grandmaster and young

master both believe that the grandmaster will win the final game. However, no other

player in the tournament has a score greater than 9 points. The Soviet grandmaster

realizes that an outcome of a draw with himself will greatly raise the young Soviet

master's chess rating without damaging his own rating and might improve the young

master's confidence and future playing ability. Therefore, the Soviet grandmaster

who has already won the tournament by virtue of his previous victories offers the

younger master-level player a draw midway through the game.

Another example of the problems faced by computer programs in dealing with

the intentionality of human adversaries is provided by Levy (1984) who claims that

he will intentionally choose inferior moves when playing against a computer. The

inferior move choices are selected by Levy because he believes that computer chess






34

programs have greater difficulty in playing a winning game against such an irrational

style of play.

The Levy example above demonstrates one of the traits of human adversaries.

Humans, especially when competing against a machine, attempt to mislead their

adversary about their intentions and capabilities. We must be able to distinguish

between relevant information and misinformation about an adversary's playing style.

Incorporating the detection of misinformation from an adversary into the inductive

learning mechanism adds an additional heuristic element. The addition of another

heuristic element reduces the probability of acquiring a valid adversary model.

The effect of an adversary's misinformation attempts is negated by selecting

previous domain performances that limit the possibility of misinformation. By only

using the previous games of an adversary from tournament play in which the

adversary has something at stake, misinformation is eliminated. For our application

in the chess domain we use the World Championship Match games of adversaries

taken from the compendium by Gelo (1988).

2.10.4 Psychology

Chase and Simon (1973) have extended the research performed by de Groot

which claims that the depth of game tree search and the total number of moves

considered is approximately the same for both novices and chess masters. However,

the ability to recall chess positions after a five second presentation of the positions

is markedly different. Chess masters are able to reconstruct chess positions involving

twenty or more pieces with complete accuracy. Chess novices are only able to place






35

four or five of the pieces correctly. The difference between the masters and the

novices disappears when the subjects are asked to reproduce random patterns of

chess pieces. Each group can only place three or four of the pieces from the random

board configurations. The apparent decrease in ability by the master level players

is attributed to the fact that the random configurations of pieces contained no

inherent domain knowledge.

Based on Miller's (1956) hypothesis, the chess masters are forming higher

level chunks composed of groups of pieces from the game positions. These chunks

contain domain specific knowledge from the prior experience of the chess masters.

Chase and Simon (1988) report that human chess masters store from 10,000 to

100,000 chess patterns.

The skill of chess masters increases as more patterns are acquired. Novice

chess players move away from slow deductive reasoning towards the fast perceptual

processing used by masters as more and more patterns are learned (Chase & Simon,

1973). The chunking theory of learning (Laird et al., 1986) supports this viewpoint

in stating that performance improves via the acquisition of knowledge about patterns

and chunks. Richard Seltzer, the father of fourteen year old chess master Bobby

Seltzer, in a personal communication relates that Bobby's increase in skill is reflected

by an improved perception of the board position and subsequent position evaluation.

Bratko et al. (1986) state that natural positional moves are closely related to

the chunk structures of a game position. Specific chunks cause chess players to

generate tactical and strategic plans. The chunk displayed in Figure 4 will cause an






36
adversary to immediately consider plans for a back-row mate with either the queen

or one of the rooks. Back-row mate plans result from the knowledge contained in

the chunk concerning the opponent's king's limited mobility.





Ili .





Figure 4: Sample cognitive chunk.

Humans recall chess chunks through familiarity with common board

configurations. Simon and Gilmartin (1973) developed MAPP which reconstructs

chess positions by simulating human eye movements recorded in their study. Simon

and Gilmartin's research shows that humans tend to group or chunk together pieces

which share geometric proximity. We will use geometric proximity as the primary

filter for selecting chess pieces to be formed into chunks. Proximity is one of the

Gestalt principles of organization (Anderson, 1980) along with similarity, good

continuation, and symmetry. We will also utilize the Gestalt principles of similarity

and good continuation to construct chunks. Chase and Simon (1973) found that

proximity and piece color, or similarity, were the two strongest factors affecting chunk

pattern recall by their chess master subjects.

The significant use of patterns of game pieces by chess masters has led us to

make the following three part hypothesis:






37
Chess masters acquire several thousand patterns of pieces that are

used to evaluate a game position.

Chess masters will continue to use a specific strategy while winning

games and tournaments.

Chess masters will tend to reduce the complexity of a game position

by moving board positions that do not contain familiar patterns into

positions that do contain familiar patterns.

We present the general hypothesis for all adversarial domains in Chapter 3. This

version of the hypothesis is the chess domain translation from the general hypothesis

and is presented here to examine the underlying foundations. The first part of our

hypothesis is a direct result from the research of Chase and Simon examined above.

The second part of our hypothesis is supported by the Reinforcement Theory

of psychology and is an extension of the results from various psychological

experiments (e.g., the water jug problem (Anderson, 1980) and the 1023-choice task

(Laird et al., 1986)) which indicate that human subjects will continue to use a learned

strategy for solving a class of problems even when a more efficient strategy would

also solve a specific problem.

While the third part of our hypothesis is intuitively appealing, direct support

comes from the research of Horgan et al. (1989) and Saariluoma (1984) which

demonstrates the use of chunking and analogy to similar positions by chess masters

in solving chess problems. The chunking of several pieces into a known pattern

produces a cognitive economy while evaluating complex board positions.







38

The search complexity of game trees is reduced as a result of the application

of our hypothesis. The strategic and tactical decisions of an adversary can be

accurately predicted because the adversary will tend to recreate familiar patterns to

reduce the cognitive complexity of evaluating the current game position. A practical

example of the use of familiar patterns is demonstrated in the 1978 Budapest

tournament game between Pinter and Bronstein (Roycroft, 1988). This game

resulted in a KNKBB endgame. Chess literature had previously claimed that a

position known as the Kling and Horowitz position (Kb6Nb7) will result in a draw

for the KNKBB endgame. Pinter's knowledge of the Kling and Horowitz position

led him to consistently maneuver to create this position. Pinter succeeded in forming

the Kling and Horowitz position three different times in three different corners of

the board. However, Pinter's haste to create the chunk caused him to make inferior

moves nearly thirty percent of the time according to the 250,000,000 position BBN

database (Roycroft, 1988) which exhaustively solves the KNKBB endgame in an

optimal number of moves.

Computer chess programs need to simulate the cognitive economy of chess

masters to continue to improve in performance. Lenat et al. (1979) state that

intelligent systems must perform expectation filtering, using predictions to filter

unsurprising data. Such a filtering process is provided in adversarial domains by the

ability to predict adversary moves in specific situations.

2.10.5 Computer Chess and Artificial Intelligence

The first chess "computer" was the automaton constructed by Torres y

Quevedo circa 1890 (Bell, 1978). Torres y Quevedo's chess machine was capable of






39

accurately playing certain variations of the KRK, King and Rook versus King,

endgame.

Since Claude Shannon's (1950) paper on the requirements for a chess playing

program, many artificial intelligence researchers have devoted themselves to the

creation of master level game playing programs. Shannon defines two main

strategies for developing chess playing programs. These strategies are the type A

strategy which generates all possible moves to a certain depth of the game tree and

then uses an evaluation function to choose the best move and the type B strategy

which uses heuristics for suggesting moves without an exhaustive search.

The type A strategy chess programs dominate the type B strategy competitors.

Northwestern University's type A chess program CHESS4.X (Frey, 1983), were X is

a version number, reigned during the 1970s as the North American computer chess

program champion and was the World Computer Chess program champion from

1977 until 1980 (Kopec & Newborn, 1987). The three type A programs BELLE,

HITECH, and DEEP THOUGHT have each set the high water mark for chess

program playing strength at USCF ratings of 2200, 2400, and 2500 respectively.

These three programs have also been the North American computer chess program

champions from 1985 to the present (Newborn & Kopec, 1989).

The chart in Figure 5 shows the progress in playing strength of chess programs

with respect to search speeds. The bottom axis is logarithmic which means that the

linear progression in strength of chess programs during the 1970s was accomplished

by an exponential increase in search speeds. Recent progress depicted by the dashed







40


I





\ N

























'4;
I N
B -
















6I IC
L **

iS
'4 0






\ N C"
v w n

\ i 's &
\~ 8 ^
\5r .T 1S
^u I Il
A ^ P *
\ ~~ 5N m






41

curve reflects the hyper-exponential increase in search speeds required to maintain

the linear growth in playing ability. BELLE, HITECH, and DEEP THOUGHT each

use specialized hybrid architectures to achieve the dramatic increases in search speed

depicted in Figure 5 (Ebeling, 1986). HITECH uses a VLSI architecture with fine

grained parallelism to perform pattern matching which is cognitively similar to the

pattern recognition performed by human chess masters. DEEP THOUGHT

(Newborn & Kopec, 1989) uses special VLSI circuitry to achieve a search rate of

720,000 position nodes per second.

Although type A strategy chess programs are the best performing chess

programs, their performance against grandmaster rated human players has been

abysmal. In an exhibition round robin tournament played in October 1989 at Boston,

four human grandmasters crushed four of the top computer chess programs, including

HITECH and DEEP THOUGHT (Hamilton, 1990), with a score of 14.5 to 1.5.

World Chess Champion Gary Kasparov defeated DEEP THOUGHT in a two game

match played in New York prior to the Boston tournament. Kasparov commented

that DEEP THOUGHT lacked experience and was unable to analyze the reasons for

some of the moves he made (Geitner, 1989).

Berliner and Goetsch (1984) state that the performance of chess programs is

proportional to the amount of knowledge used by the program's evaluation function.

Bratko and Michie (1980) demonstrated the need to use domain specific knowledge

to solve certain domain tasks by having CHESS4.5 and the AL1 program try to solve

the KRKN endgame. CHESS4.5 was unable to solve two of the three chess






42

problems used in Bratko and Michie's research. However, AL1 which uses a type

B heuristic method with domain knowledge was able to solve all of the endgame

problems.

Type B strategies vary in their approaches to playing chess. The first heuristic

method uses plans or goals that have been constructed from domain specific

expertise to solve chess problems. AL1 is a goal based chess program which solves

KRKN endgames. LEBL (Tadepalli, 1989) uses goals during King-Pawn endgames

to reduce search complexity. Wilkins (1980 and 1982) also uses domain specific

knowledge to produce plans which his program PARADISE tries to achieve. Goal

based planning allows PARADISE to accurately play quiescent middle game

positions. The use of goals to achieve higher playing performance in chess programs

simulates the cognitive method of means-ends analysis used by humans in certain

problem solving tasks (Anderson, 1980).

Another heuristic approach which is similar to the cognitive methods analyzed

by Chase and Simon, described in section 2.10.4, simulates the human cognitive

process of chunking. This heuristic approach tries to identify common configurations

or chunks and then uses these chunks to identify analogous solutions to chess

problems. Quinlan's ID3 program (in Michalski et al., 1983) tries to inductively

classify equivalent board configurations in the KRKN endgame. The ID3 induced

patterns are used to determine if the King-Rook side has a winning position within

a specified number of moves. Currently the program has solved the KRKN endgame

for all positions which can be won in three ply. Campbell's (1988) CHUNKER also






43

takes advantage of board configurations to produce chunks which are used to solve

King-Pawn endgames.

Each of the heuristic methods described above have been implemented on

endgame or middle game portions of chess problems and deal primarily with tactical

solutions to chess problems. The knowledge available to chess programs can be

significantly augmented by extending the research of Quinlan and Campbell in the

following ways:

Collect chunks from complete chess games instead of just the endgame

segment.

Use the chunks of pieces to predict an adversary's tactical and strategic

movement decisions.

2.10.6 Justification for Chess Domain Application

,' Shapiro (1987) presents several reasons for choosing chess as an experimental

test-bench.

1. The game constitutes a fully defined and well-formalized domain.

2. The game is sufficiently complex to challenge the highest levels of

human intellectual capacity.

3. A large body of knowledge has been accumulated and is available in

literature form.

4. A generally accepted numerical scale of performance is available so

that increases in performance can be rated.

Our application domain needs to be sufficiently complex so that exhaustive search






44
is not a feasible solution. Typical chess game trees consist of 1043 nodes (Berliner

& Goetsch, 1984) which prohibits exhaustive search solutions given current

technological capabilities. Because chess is a well understood domain and there is

an accepted numerical rating scale, performance improvements contributed by the

adversary modeling methodology can be evaluated.












CHAPTER 3
ADVERSARY MODELING METHODOLOGY

In this chapter, we examine in detail each of the functions required to create

an adversary model from the adversary modeling methodology. The foundation for

our adversary modeling methodology is the three part hypothesis:

Relevant patterns and plans are learned by the adversary from

practical experience in the domain.

The strategic and tactical actions of an adversary are repeated as long

as the actions produce a positive result.

An adversary tends to reduce the complexity of a situation to enable

a more precise strategic evaluation of the situation by using the plans

and patterns already acquired in the domain.

Evidence supporting this hypothesis and the specific chess domain translation are

presented in section 2.10.4.

To accurately predict the strategic and tactical plans of an adversary, we must

first be able to perceive and evaluate the domain the same as our adversary. This

means that we need to acquire the evaluation criteria used by our adversary. The

first and third parts of our foundational hypothesis indicate that all competitors in

adversarial domains make use of patterns that have been previously acquired in the

domain for their evaluation process.







46

3.1 Knowledge Acquisition of Chunks


For chess, the patterns involve both the geometric patterns of pieces and the

textual patterns related to the actual moves that are executed. The geometric

patterns correspond to chunks that are stored in visual memory, while the textual

patterns correspond to chunks that are stored in the adversary's verbal memory.

3.1.1 Defining the Chunks to be Acquired

We rely on the Gestalt principles of organization and the research of Chase

and Simon described in section 2.10.4 to construct our definition of the elements that

are to be contained in a geometric chunk. Because we are trying to find chunks of

pieces that are used to reduce the complexity of a situation, our chunks necessarily

have a minimum of two pieces. Although chunks need only two pieces to satisfy our

definition, chunks are generally required to contain a minimum of three pieces to be

used in our predictions of probable adversary moves. Chase and Simon (1973)

indicate that the chunks used by chess masters contain three to four pieces. Two

piece chunks are strategically significant in the opening game segment of chess due

to the minimal number of pieces that have moved from their starting positions and

in the end game segment due to the previous elimination of pieces. However, two

piece chunks in middle game positions often occur without effect from the strategy

or cognitive simplification efforts of the adversary. The chunking mechanism

emphasizes chunks which contain a minimum of three pieces.

Based on the principal of proximity, chunks contain only pieces which are

adjacent to another piece. Pieces contained in a chunk are all of the same color to






47
satisfy the principle of similarity. Requiring chunks to be of a single color, namely

the pieces played by the adversary, restricts the contents of the chunk to be under

direct control of the adversary. An adversary cannot control the move choices of an

opponent except for inter-locked pawn chains and therefore cannot use a chunk with

mixed colors of pieces for long range strategic planning.

Research performed by Church and Church (Holding, 1985) demonstrates that

human chess players have difficulty in processing information from pieces on

diagonals. From this research we only chunk pieces by proximity which are adjacent

along either the horizontal axis or vertical axis of the playing board. Other

adversarial game domains like checkers can rotate the major axes to include the

diagonals and exclude the horizontal and vertical axes.

The Gestalt principle of good continuation or meaningfulness is used to chunk

pieces which are capable of affecting the strategic meaning of a chunk from the

diagonals. The bishops, queens, and king are all included in a chunk if they are

diagonally adjacent to another piece in the chunk. Meaningfulness is also used to

include pawns which are diagonally behind, to reflect the movement capabilities of

the pawn piece, a piece in a chunk.

Furthermore, we need to define the maximum size of each chunk to build an

efficient knowledge representation structure for storing the chunks. Humans

compose complex chunks hierarchically from smaller, less complex chunks. We use

a maximum chunk size of a four-by-four, or sixteen square, board area. All of the

chunks collected by IAM from a twelve game analysis of M. Botvinnik are shown in







48

Figure 6. From these chunks we can see that a sixteen square area is sufficient to

hold all of the chunks found. We increased the maximum size of defined chunks to

five-by-five and six-by-six to measure the affect of a larger chunk size. Our research

indicated that only one additional chunk was captured during an analysis of twenty

games played by Botvinnik (see Section 4.2.4, Figure 22). This marginal increase in

acquired chunks does not merit the increased cost in space requirements for larger

chunk definitions. The sixteen square size chunk is the minimal implementation that

also permits the use of a chunk to acquire additional knowledge we deemed useful,

including the pieces remaining at the end of each game.

Because of the significance of pawn formations in the analysis of chess

positions, we acquire a second type of geometric chunk which is aimed at acquiring

pawn formations. As noted above, pawn formations are stable and an adversary can

rely on the presence of an opponent's pawns in a chunk. Our similarity principle is

modified to collect all similar pieces, i.e. pawns, instead of similar colors. The

proximity and meaningfulness principles remain the same for pawn chunks. Because

pawn structures can extend over a large area of the game board, we increase the size

of the maximum defined chunk from a sixteen square area to a thirty-six square area.

Support for increasing the maximum chunk size for the pawn formation chunks was

found when several pawn structures were acquired during our research which would

have exceeded the sixteen square area used in the standard geometric chunk

definition.








a I

aII

I aaaI
al
onl I

a11






aa8


a L


a


Figure 6: Chunks acquired for Botvinnik.


w


a

a_






_a _
aa



Ag ac

















Figure 6--continued.

We have emulated the cognitive perceptual processes of human chess players

to define the composition of geometric chunks. Our definition of the elements which

are viable chunk members acquires chunks which are easily perceived and which

contain the maximum amount of knowledge concerning the current game situation.

Our definition of textual chunks is less complicated. Human chess players

remember verbatim dozens of opening lines of play that they have experienced. We

emulate this rote learning by chunking the first five moves made by the adversary

verbatim from the game records of the studied games. Our research indicates that

chess masters tend to leave an opening line of play between the fifth and tenth move

of the game. We are interested in acquiring knowledge that is high in quality and

reliability and therefore we rote learn only the first five moves. One of our

experiments considered the effect of increasing the textual chunk size to ten and

fifteen moves and the results of this experiment are detailed in Chapter 4.

3.1.2 Acquiring Geometric Chunks

From our background study presented in Chapter 2, we note the significance

of perceptual processing in the development of chess expertise. We take advantage

of this aspect of expert chess performance by viewing the game board and pieces as






51
an image to be processed into chunks. By viewing the game board as an image, we

consider via analogy previously defined computer vision techniques for processing

images and collecting information.

The convolution operator is used in the image algebra (Ritter et al., 1988) of

computer vision to collect information from the area surrounding a specific pixel,

which is analogous the proximity principle we use for defining chunks. For games,

a homomorphism can be drawn between pixels in an image and pieces on a game

board. The meaningfulness principle that we use in our chunk definition requires us

to capture knowledge from different areas surrounding a piece, depending on the

current piece being analyzed. The technique of varying templates enables us to alter

the summing template used by the convolution operator to correspond to an

individual piece. This permits us to collect information about the proximal pieces

that is dependent upon the central piece.

We maintain an internal representation of the game board in an eight-by-eight

array with subscripts 'i' and 'j'. The first subscript denotes the column and the

second subscript denotes the row. Our internal representation is similar to the one

described by Shannon (1950) with negative values assigned to non-adversary pieces

and the numbers one through six representing the pawns, knights, bishops, rooks,

queens, and kings respectively. Unoccupied squares are denoted by the value zero.

A board position and its corresponding internal representation are demonstrated in

Figure 7. The piece value at each board location is denoted by the algebraic symbol

Xij.


























i= 1 2 3 4 5 6 7 8

Figure 7: Internal representation of a board position.

The three different templates used in our implementation of the convolution

algorithm are shown in Figure 8. Each template can be thought of as a three-by-

three array with the center location denoted by the subscript (0,0). Other template

subscripts range from negative one to one. The general template in Figure 8 is used

for bishops, queens, and the king with the other pieces being convolved by the

appropriately named template. The pseudo-code in Figure 9 demonstrates how the

convolution operation is performed. When an adversary's piece is detected, the

appropriate template is loaded into a copy of the current template--TP. The values

depicted in the templates were a proximal piece is located are summed and stored

in the return array Y using an inclusive or operator--G. We have defined the

predicate function, ON_BOARD, to verify that the array subscripts 'i + M' and 'j + N'

correspond to a valid board location. The return array Y is an eight-by-eight copy

of the internal board representation with all values initially set to zero. This


j=8

7




4
3

B 2


-4 -2 B 8 8 -4 -6 8

8 -3 -5 -1 -3 8 -1 -1
-1 -1 8 a -1 8 8 8

8a B B 8 -i a a
8 1 1 1 8 a -z2
1 8 2 8 2 1 8
8 8 8 8 8 1 3 1

48 5 4 8 6 B










4 8 16 4 8 16 D 8 0
2 32 2 32 2 32
1 128 64 0 128 0 0 128 0

General Pawn Rook, Knight
Figure 8: Convolution templates.


Case Xj of
3, 5, 6 : TP the general template ;; (See Figure 8)
1 : TP the pawn template
2, 4 : TP +- the rook-knight template
otherwise Return ;; Non-adversary piece, no need to convolve
End Case
For N = -1 to 1 do
For M = -1 to 1 do
If ONBOARD(i+N, j+M) and X(i+N), +m) <> 0
Then Begin
Yij +- Y,j TPN,M
Y(i+N),(j+M) Y(i+N),(j+M) TP-N,-M
End Begin

Figure 9: Pseudo-code to perform convolution.

effectively captures all pieces satisfying our chunk definition qualities and returns

their relational position to the center piece location in Y.

The internal representation of the game board is updated after each move

that is read from the game record. For example, if the move 'al-cl' is read to move

the rook, in the lower-left corner of the board in Figure 7, two squares to the right,






54

then the value at X3,1 would be set to four and the value of X1,1 would be changed

from four to zero. Currently we use a modified algebraic notation for recognizing

the moves played during the game. Other notations can be easily translated to

algebraic notation for use by the IAM program. The chunks we desire to capture are

the ones that have just been formed by the adversary, which improves our capability

for predicting definite adversary actions. Therefore, we collect chunks following the

adversary's move.

The appropriate step function shown in Figure 10 is applied to the board

representation X and is stored in an eight-by-eight array Y. Pseudo-code which

reduces the complexity of the ensuing board search by removing all pieces that do

not satisfy the appropriate similarity constraint is shown in Figure 11. We use the

step functions to increase the speed of the search algorithm which locates pieces that

can be chunked together. The convolution operation shown in Figure 9 is then

performed at the site of the current board search. If any neighbors are identified as

belonging to the chunk, then the convolution operator is applied iteratively to all

related neighbors. This acquires the chunk as soon as the left-most and bottom-most

piece, with respect to the al square, is found.

Chunks of pieces which have never been moved by the adversary do not

represent any significant new geometric relational or strategic knowledge. To reduce

the total number of chunks being stored, we discard chunks which have greater than

half of their pieces still in their original starting positions. To further reduce the

storage requirements of the adversary modeling methodology we take advantage of













1- -




a


to


Standard

Geometric Chunks


Piece
Values
8
-1 1

P4

*$
wM


Pawn Formation Chunks


Figure 10: Step functions.


For I = 1 to 8 do

For J = 1 to 8 do

If gathering standard geometric chunks AND XI, > 0

Then Yj, 1
Else If gathering pawn formation chunks AND ABS(X,) = 1

Then YI, 1

Else Y,4 4- 0
For I = 1 to 8 do
For J = 1 to 8 do

If Y, = 1 Then perform convolution

Figure 11: Pseudo-code to perform step function.

chunk durations. If a chunk is created on the tenth move of a game and persists

until the twentieth move, then only one chunk is created instead of ten chunks. The


Piece Values


I






56

times of a chunk's inception and subsequent dissolution are stored with each chunk.

We also use the time tags associated with each chunk to eliminate storing a chunk

which is dissolved by the adversary and then recreated at a later time during the

same game. Pseudo-code to perform the chunk comparisons, which produce the

storage savings just mentioned, is shown in Figure 12. The variable 'C' represents

the double linked list of chunk records with the subscript 'I' denoting the position in

the list and the variable 'D' represents the most recent chunk that has been identified

for the current game.


For I = 1 to the total number of chunks created for this game do
If C, = D
Then Begin
If D.start_time < CI.start_time
Then CI.starttime ,- D.start_time
If D.end_time > C,.end_time
Then C,.end_time D.endtime
D -0
Exit the loop
End Begin
Figure 12: Pseudo-code to perform chunk coalescing.

An example of the chunks identified by the convolution operator and the

convolution values returned is shown in Figure 13 where the chunks have been

acquired for the white player. The position shown in Figure 13 is taken from the

fifth game of the 1951 World Championship match between Botvinnik and Bronstein

after white's eighteenth move. The three chunks acquired by the convolution



































8 33 162 2

16 8 8 8 224 195 8

120 174 7
8 32 2 28
B 32 2 28
a-I- E 1 L


Blanks = 8


Figure 13: Chunks on a chess board and their convolution values.






58

algorithm are outlined in the figure. The pawn structure in the middle-left of the

board could become a strong factor in the endgame and possibly create a passed

pawn. This knowledge is available from the geometric relationship of the pieces in

the chunk. The other two chunks are a standard offensive configuration (i.e., the

queen and rook) and a standard defensive configuration (i.e., the fianchettoed bishop

and castled king) that are found in many chess games.

The generality of the geometric chunk acquisition technique is demonstrated

by applying the convolution operator to the adversarial domain of the Go game.

Because only one type of piece is used, the general template from Figure 8 is the

only template used in performing the convolution. The results of the convolution

algorithm applied to a section of the Go game board are shown in Figure 14. This

chunk represents the Go formation called eyes. Eyes are an important defensive

strategic feature in Go. Other Go formations such as ladders, knots, false eyes, and

tiger's mouths are also acquired by the adversary modeling methodology's

convolution algorithm.

3.1.3 Acquiring Textual Chunks

The background research presented in Chapter 2 indicates that chess masters

make use of their verbal memory to recall specific opening sequences of play. Our

adversary modeling methodology acquires this facet of expert chess performance in

addition to the geometric piece chunks.

Chess masters can recall verbatim particular opening sequences of play, which

indicates the use of rote learning. We emulate the cognitive process of rote learning






















The heavU black line Values in squares correspond
represents the edge to pieces located at game
of the game board. board intersections

Figure 14: Chunk acquisition in Go.

to acquire textual chunks. Each move from a previously played game is read in

textually, in modified algebraic notation, from the game record used as input. The

algebraic notation is then translated to update the current board representation

maintained by the adversary modeling methodology application. Prior to translating

the textual notation used to record the moves of an adversary's game, specific types

of moves are saved in the "move" section of the adversary knowledge base verbatim.

Due to the availability of transpositions which lead to the same board position

through several different move sequences, we store each of the first five moves

performed by an adversary individually. A statistical representation of the frequency

that a move occurs and the average game turn is generated and stored with each

move. The statistical representation includes: the number of times a specific move


168 35 34 98 138

24 12






60

has been observed, the mean move number at which the move has occurred, and the

color of the pieces being played.

From the statistical knowledge of each of the moves displayed by the

adversary, the adversary model predicts opening sequences of play based on the

average over-all performance of an adversary. The knowledge of piece color is

crucial because chess masters use different modes of play for each of the two colors.

As the white player, a chess master can choose which opening sequence to play and

has greater control over the position of the pieces following the opening. However,

as the black player, a chess master plays a responsive mode instead of the controlling

mode of white. The opening moves of black are made as responses to the white

player's move choices. Therefore, if an identical move is made by an adversary as

both a white piece and black piece player, then the two moves are considered as

different moves and are stored separately.

Other types of moves are useful for predicting the playing style of an

adversary. We save the number of pawn moves that occur during the first ten moves

of each game and the number of attacking moves made during the first ten and

twenty moves of each game to aid in inferring the playing style of an adversary. Any

other move which can be identified textually is also a candidate for the textual move

knowledge of the adversary model. A flowchart for the process of collecting textual-

based knowledge is shown in Figure 15. The frequency and direction of castling

moves are also acquired to assist in prediction of possible castle moves during an

adversary's future games.


































To demonstrate the flowchart, assume that the move c4--pawn to queen's

bishop four--is made by the adversary on his fourth game turn. Since the move

occurs during the first five moves of the game, the move will be saved into the

textual chunk knowledge base. If the move already exists in the textual chunk

knowledge base, then the number of times this move has been observed is

incremented and the statistics regarding the time of the move are updated to reflect

the new knowledge. Because a pawn--the queen's bishop pawn--is being moved, we

increment the count of pawn moves made during the first ten moves of the game.

We save the total number of games analyzed, the number of games the

adversary played as white, the number of games won by the adversary, the number






62

of games won as white, and the average length of all analyzed games. This

information is used by the adversary model to determine the frequency of a specific

move and the likelihood that the move will be repeated by the adversary.

Other adversarial domains can benefit from the textual knowledge acquisition

performed by the adversary modeling methodology. For example, the acquisition of

textual patterns will capture standard operating procedures and protocols, such as

how to accept an enemy's surrender, that have been learned by an adversary in a

military domain.


3.2 Induction to Learn Repeated Chunk Patterns


We have acquired the geometric chunks, both standard and pawn structure

chunks, through the computer vision technique of applying convolutions to the board

image. Our goal with the adversary modeling methodology is to learn the chunks

which are used by the adversary to evaluate domain situations. We propose that the

chunks which are commonly used by an adversary can be identified through their

repeated use across several games.

Induction is used to identify groups of chunks from the collection of chunks

acquired from each of the analyzed games. Each group which contains a minimum

of two chunks, indicating the observation of the chunk in two separate games, is

identified as a cognitive chunk used by the adversary in domain evaluations. Chunks

that are found only in single games may occur by chance and are therefore not used

to predict the tactical and strategic decisions of an adversary. Because of our






63 .
inductive premise, we require that a minimum of two adversary games be studied

prior to performing the induction.

The induction algorithm requires chunks to be identical before being grouped

together. Our induction algorithm is noise intolerant since even small variations in

a chunk can have a large effect on the strategic evaluation value of a chunk. An

example of three chunks which only differ by the location or value of a single piece

and which have significantly different strategic meanings is shown in Figure 16. The

first chunk shows a strong pawn formation, while the second chunk shows two pawns

being attacked by an opponent's pawn, and the third chunk shows a pawn formation

with a double pawn which may become a liability towards the end of the game. The

noise intolerance of the induction algorithm is not a real problem since most

adversarial domains keep very accurate records of the past performances of domain

entities.




i sij I iil SIui





Figure 16: The effect of noise on chunks.

Chunks are considered identical when each has the same number and type of

pieces arranged in the same geometric configurations. The strategic significance of

a chunk is usually invariant with respect to left and right alignment, so chunks are

reflected through a vertical axis of the game board while trying to identify identical






64

chunks. When the adversary is playing the black pieces, we reflect the pieces through

the mid-board horizontal axis to group chunks that occur while the adversary is

playing either color of pieces. Chunks considered to be identical, as shown in Figure

17, are grouped together and stored in the knowledge base that forms the adversary

model.


Figure 17: Four identical chunks.

We use a frame based system to store the chunks found by the induction

algorithm in the adversary model's knowledge base. The slots in each frame are the

actual chunk pieces and relative positions of the induced chunks, the number of times

this chunk has been observed, the earliest move number when the chunk was created,

the latest move number when the chunk was dissolved, the outcome (win or loss) of






65
the games in which the chunk occurred, and the color of the adversary's pieces.

Although we have not implemented this slot, chunks which are observed frequently

can have a "best move" slot added which would model the scripts used by chess

masters for certain configurations.

The adversary model's knowledge base of patterns known to an adversary can

be constructed in parts. At least two games need to be processed by the convolution

operator to acquire potential chunks. After the potential chunks have been acquired,

the induction learning mechanism processes the potential chunks of each game by

first checking to see if the chunk already exists in the adversary model's knowledge

base. If the chunk already exists in the knowledge base then the appropriate slots

are updated to reflect the new information associated with the new potential chunk.

Should the chunk not exist in the current knowledge base, then the chunk is

compared against the chunks of every other game that has just been processed by the

convolution operation. Any identical chunks are grouped and stored in the

knowledge base as a new pattern known to the adversary.

The induction learning mechanism is applied separately to general chunks

containing different pieces of the same color and pawn structure chunks containing

only pawns of both colors. Following the induction procedure, the adversary model

has a knowledge base which identifies all the recurring textual and geometric

patterns from the previous games of a particular adversary. Multiple adversaries can

be modelled by the adversary modeling methodology with each adversary having a

separate knowledge base. The adversary model's knowledge bases of specific






66

adversaries are used to infer the playing style of an adversary and to predict the

tactical and strategic movement choices of an adversary.


3.3 Acquiring an Adversary's Playing Style


Knowing the general playing style of an adversary enables chess programs to

select moves that will place the adversary in a disadvantageous strategic position.

The general playing style of an adversary is inferred from the collection of patterns

contained in the adversary model knowledge base. We use a rule-based inference

engine to heuristically determine the playing style of an adversary.

Currently, we infer two different styles of play. The inference engine indicates

that an adversary prefers open positions, closed positions, or has no obvious

preference. The closed position style of play is further used to imply strong

positional play by the adversary. Likewise, an open position playing style implies a

strong tactical player.

The rule base heuristics address the following domain specific factors which

indicate an adversary's playing style preference:

The presence and relative size of pawn structures.

The number of pawn moves made during the opening sequence.

The presence of pawn chains, or inter-locked pawn structures.

The number of opening moves that are attacking--move a piece across

the mid-board horizontal axis.

Each rule is evaluated to determine if a specific playing style is demonstrated. The







67

rules are not competitive, but instead they lend support to the adversary model's

belief that a particular playing style is preferred by the adversary by adjusting the

likelihood value attached to each playing style. The playing style with the greatest

likelihood is then inferred as the general playing style of the adversary.


Pawn Moves








8 4 18









0 1.25 1.75 3.8
Pawn Chains


Xe3

Opei
StgJi

Clos'
StJ I


Pawn Structures of 5 Pawns
1.0i -


ed
.e


8.8


0 1.5 2.0 5.0+









8 1.5 3.8 18.8+
Attack ing Moves


Figure 18: Fuzzy logic used by the heuristic rules.

The heuristic rules use fuzzy logic during the evaluation process. Graphs of

the fuzzy values for each of the heuristic rules is shown in Figure 18, with the fuzzy

regions shaded. Heuristic rules that have a value in the fuzzy areas which lie close

to the division between the two styles of play currently inferred do not support either

hypothesis. Our decision to use fuzzy logic prevents us from assigning a playing style

to an adversary that is based on statistically insignificant data with respect to the

fuzzy set values.


1.8


1.0






8.8






68

The heuristic rules used to infer an adversary's playing style are not

competitive, new heuristic rules can be added at any location of the rule base. As

new domain specific knowledge becomes available that will infer additional playing

styles or support the current playing style inferences, the current rule base can be

easily augmented.

The determination of an adversary's playing style preference has been added

to the generalized adversary modeling methodology to increase the adversary model's

performance in the chess domain. The playing style inference engine makes use of

domain specific knowledge in addition to the textual and geometric patterns and for

this reason is not considered as a main part of the adversary modeling methodology.

The pattern acquisition, induction learning mechanism, and adversary action

prediction tool each use only minimal amounts of domain specific knowledge so the

adversary model implementation can be used in other adversarial domains with little

or no modifications.


3.4 Application of the Adversary Model


The adversary model consists of a knowledge base of textual and geometric,

or visual, patterns known to the adversary. The third part of our hypothesis

presented at the beginning of this chapter indicates how we are going to use this

knowledge. An adversary will attempt to reduce the complexity of evaluating the

current domain situation by maneuvering the board position to contain one or more

of the pattern chunks with which he is familiar.






69
The adversary model also contains two processes which utilize the knowledge

about patterns known by an adversary. The first process detects analogous situations

in the domain and the second process attempts to predict the most probable

adversary action. We define an analogous domain situation to exist when a chunk

in the knowledge base is almost complete in the domain. The second process then

predicts the necessary move to complete the chunk.

Currently a move window size of one is used when detecting analogous

domain situations. For chess, this means that the chunk must be able to have all of

the pieces of the chunk in their correct relational positions in exactly one move.

Board positions that are considered to be analogous to chunks in the knowledge base

may only have at most two pieces out of position, but are normally required to only

have one piece out of position. An extension to the move window size has been

implemented to permit the adversary model to eliminate interfering pieces when a

castle is the required move to complete a chunk.

We detect the analogous situations we have defined for the chess domain by

performing pattern recognition on the game board. The board representation of the

adversary's game currently in progress is searched prior to each move to be made by

the adversary until a piece belonging to the adversary is located. Next all of the

chunks in the adversary model's knowledge base which can fit into the current board

location are compared against the current board pattern. Inconsistencies between

the chunks in the knowledge base and the current board configuration are counted.

An inconsistency exists if a board location corresponding to the location of a piece






70

in a knowledge base chunk is empty or contains a different piece. If only one

inconsistency exists, or two if the missing pieces are the king and rook, then the

analogous situation detection process checks to see if any viable move can complete

the chunk.

Chunks in the knowledge base occupy a square area of the game board

defined by the pieces farthest to the left, top, right, and bottom of the chunk. The

square areas of chunks frequently contain blank or non-occupied squares. These

squares serve as wild cards so that the actual square on the game board may be

blank or contain any other piece. This is done to simulate the effect of hierarchically

composing larger chunks from smaller chunks.

Each chunk in the knowledge base which can be completed suggests the

appropriate move to complete the chunk as the adversary's next move choice.

Several chunks can each have analogous situations for any particular turn of the

current game. The collection of possible moves that enable an adversary to complete

a chunk are evaluated by the adversary move prediction process to select the most

likely adversary move.

The adversary move prediction process uses a heuristic rule-based inference

engine to evaluate the collection of suggested moves that will complete chunks. The

inference engine is similar to the one described above for analyzing the playing style

of an adversary. An initial probability value is generated for each move by dividing

the number of times the chunk suggesting the move has been observed by the total

number of prior adversary games analyzed. The heuristic rules then add or subtract






71
from the base probability of each suggested move. The move with the highest

probability is then used to predict the adversary's next move. If two or more moves

have nearly equal probabilities, then the inference engine makes multiple move

predictions.

Computer chess programs can use the probability value associated with each

move as a measure of belief. When the adversary model infers a probability for a

move that is less than some specified value, the chess program can choose to ignore

the adversary model's prediction.

The heuristics used in the prediction inference engine attempt to follow

several psychological principles of cognitive organization and economy. The general

description of the purpose of our heuristics follows.

Large chunks are favored over small chunks. Chunk size is dependent on the

number of pieces contained in a chunk.

Chunks containing major pieces are favored over chunks containing only

pawns.

Favor moves which have been suggested by more than one chunk.

Reduce the probability of chunks that have only been observed in lost games.

Increase the probability of chunks whose move suggestion causes a gain in

material advantage. (This heuristic is only used in quiescent positions due to

the volatile nature of material advantage in non-quiescent positions.)

Chunks which have occurred while using the same color pieces as the current

game are favored over chunks that have only been observed for the opposite

color.






72

Adjust the probability to account for temporal discrepancies. If the current

game move number is outside of the move range defined by the two slots for

the chunk's time of inception and dissolution, then reduce the probability

proportional to the distance between the current move and the chunk's

temporal range.

Eliminate smaller chunks that attempt to borrow pieces from an existing

larger chunk.

Reduce the probability of chunks containing only two pieces during the

middle game segment.

If the adversary has just dissolved a chunk, then do not recreate the chunk

immediately.

Reduce the probability of chunks that have a maximum dissolution game turn

that is within the time range covered by the textual opening move knowledge

base.

These heuristics attempt to find the move suggestion which has the greatest number

of situational factors in common with the current game condition, such as the color

of the pieces and the time or game turn within the game. Additionally, the heuristics

promote the moves that will create the largest possible chunk and therefore afford

the greatest cognitive economy to the chess player. The heuristic which supports

moves suggested by multiple chunks is simulating the construction of hierarchically

complex chunks from smaller chunks. Our general description of the heuristic rules

indicates a minimal usage of domain specific knowledge which augments our

inter-domain application capabilities.






73
The role of the adversary model as a coach needs to be remembered.

Predictions of probable adversary actions are only made when the current game

situation resembles domain situations which have already been learned inductively.

After analyzing nine and then twelve games of Botvinnik the resultant adversary

models were used to predict the actions of Botvinnik in a new game which had not

been previously analyzed. For each of models' predictions, over forty percent of the

new game's board configurations had no similarities to the chunks in the knowledge

base.

Geometric chunk patterns are uncommon during the opening game segment

of a chess game. The textual knowledge base of patterns known to the adversary is

used to supplement the lack of knowledge in the geometric chunk pattern knowledge

base. When a computer chess program equipped with the adversary model is playing

the white pieces, the statistical analysis of opening moves displayed by the adversary

is used to select an opening sequence of play which is unfamiliar to the adversary.

This grants a strategic advantage to the computer chess program with the first move

of the game. Predictions for the opening moves of the adversary are made by

selecting the move with the highest probability value. Probability values are assigned

by using the statistical values stored with each opening move displayed by the

adversary to attain a base probability corresponding to the frequency each move has

been observed. This base probability is then modified to account for any difference

between the current time of the game and the mean game turn of the knowledge

base textual move pattern.






74

Moves which have already been made in the current game cause the identical

move in the knowledge base to be rejected. Additionally, when the adversary is

playing the black pieces, we implemented "response code" in the prediction process

which heuristically favors textual move patterns that have previously followed the

current move just made by the white player. The effect of the response code is

examined in Chapter 4.











CHAPTER 4
IAM

In this chapter we present a detailed examination of our implementation of

our adversary modeling methodology for the domain of chess. Our program is

named IAM, an acronym for Inductive Adversary Modeler. We first present several

detailed examples that demonstrate IAM's functionality. Next we review the

effectiveness of IAM by examining the predictive capabilities of our adversary model

while performing against an actual opponent.


4.1 Overview of IAM's Execution


IAM is defined in two stages. The first stage performs the knowledge

acquisition of chunks and the learning by induction phases of the adversary modeling

methodology which establish the knowledge base of geometric and textual chunks

contained in the adversary model. The second stage applies the knowledge in the

knowledge base to a game in progress to predict probable adversary moves and to

identify the adversary's playing style.

IAM has several global data structures which facilitate the algorithmic design

of the program. Chunks, both textual and geometric, and the suggested moves of the

adversary model are stored in dynamic data structures. The form of these structures

and the pointers to them are among the global data structures. Additionally,






76

application constants such as the size of geometric chunks and pawn structure chunks,

the number of opening moves to acquire, the internal representation of the chess

board, and various constants used by the heuristics of the two inference engines are

stored as global variables. The use of these global structures allows us to easily

change the amount, size, or type of knowledge that is to be acquired and provides

a means for fine tuning the heuristics used in predicting an adversary's move choices.

4.1.1 Knowledge Acquisition and Inductive Learning Stage

Input to IAM, for the knowledge acquisition and learning stage, consists of

two or three game records of the adversary's previous performances. The records are

written in algebraic notation as shown for the sample game in Figure 19. The first

line of the game record indicates the color of pieces played by the adversary, the

length of the game, and the game result, either win or loss, for the adversary. The

remaining lines are the actual moves of a particular game.

We use only the previous games of an adversary which have resulted in a win

or loss. Drawn games are often influenced by factors, like the state of health of the

adversary, that do not impart any knowledge about the strategic tendencies or

evaluation criteria of an adversary. A good example of this comes from the 1957

World Chess Championship match between Botvinnik and Smyslov. Three of the last

four games in the match ended in a draw with an average game length of thirteen

turns. Botvinnik was down by three points in this match and the first draw was

probably offered by Botvinnik so that he could rest and prepare for the next game.

The final two draws were offered and accepted because the result of the match was









Botvinnik vs. Keres (Moscow, May 4/5, 1948)

1601
1. d4 d5 21. Nb4 Rd8 41. Kd3 Ne6
2. Nf3 Bf5 22. Qf5 Rd6 42. Nd5 Kc6
3. c4 e6 23. Rfcl Rxc6 43. h4 Nd8
4. cxd5 exd5 24. Rxc6 Rd8 44. Nf4 Kd6
5. Qb3 Nc6 25. Rxb6 cxb6 45. Nh5 Ne6
6. Bg5 Be7 26. Nc6 Qc7 46. Ke3 Ke7
7. Bxe7 Ngxe7 27. Nxd8 Qxd8 47. d5 Nc5
8. e3 Qd6 28. Qc2 Qc7 48. Nxg7 Kd6
9. Nbd2 0-0 29. Qxc7 Nxc7 49. Ne6 Nd7
10. Rcl a5 30. Nbl Kf8 50. Kd4 Ne5
11. a3 Rfc8 31. Kfl Ke7 51. Ng7 Nc4
12. Bd3 a4 32. Ke2 Kd6 52. Nf5+ Kc7
13. Qc2 Bxd3 33. Kd3 Kc6 53. Kc3 Kd7
14. Qxd3 Nd8 34. Nc3 Ne8 54. g4 Ne5
15. 0-0 Ne6 35. Na2 f6 55. g5 fxg5
16. Rc3 b5 36. f3 Nc7 56. hxg5 Nf3
17. Qc2 Rcb8 37. Nb4+ Kd6 57. Kb4 Nxg5
18. Nel Nc8 38. e4 dxe4+ 58. e5 h5
19. Rc6 Qe7 39. fxe4 Ne6 59. e6+ Kd8
20. Nd3 Nb6 40. Ke3 Nc7 60. Kxb5
1:0
Figure 19: An example of a game record--the 1948 Hague-Moscow Tournament.


a forgone conclusion, similar to our example in Chapter 2 of the young Soviet chess

master and the older Soviet grandmaster.

Additional knowledge can be inferred from the win/loss result of each game

played by an adversary. When the induction procedure detects a series of lost games

by the adversary, then the possibility that the adversary will be altering his

established strategy is inferred. The adversary move prediction inference engine

decreases the probability of chunks observed only in lost games from predicting the

next adversary move. Likewise, an overall winning performance by an adversary

reinforces the likelihood of chunks being repeated in future games.






78

Each move of every game is read in as input separately. The textual chunking

mechanism then identifies if the current move has any of the qualifications for being

saved in the textual portion of the knowledge base. If so, a dynamic data structure

for holding textual moves is created and the move is stored.

Next, the textual move is translated and the internal board representation is

updated. After each move made by the adversary, potential chunks for each game

are identified by the convolution operator we discussed in Chapter 3. Chunks which

are identical to another chunk in the same game are coalesced into a single chunk

(see the pseudo-code in Figure 12). This step, for the twelve games played by

Botvinnik in the 1948 Hague-Moscow Championship tournament, reduced the 699

chunks identified by the convolution operator to 272 potential chunks that the

inductive learning algorithm must process. Chunks from different games are

considered unique at this point in processing.

Pawn structure chunks are collected by the convolution operator separately

from the standard geometric piece chunks. Because the occurrence of identical pawn

structures during different segments of a game yields knowledge about an adversary's

playing style, pawn structure chunks are not coalesced.

After the individual geometric chunks have been collected for each game, the

induction algorithm processes the chunks to find patterns which are duplicated across

game boundaries. The induction algorithm found twenty-eight unique chunks, shown

in Figure 6, from the 272 potential chunks identified for Botvinnik in the

Hague-Moscow tournament. Each of the twenty-eight chunks induced was found in






79

at least two games. For example, the partial board representations for the three

games which produced one of the twenty-eight chunks are shown in Figure 20.


Botv inn ik

vs.

Reshevsky


Xeres

VS.

Botvinnik


Botv innik

vs.

Euue


occurrence in three different games.






80

The induction algorithm, with pseudo-code shown in Figure 21, verifies if a

potential or induced chunk already exists in the adversary pattern knowledge base

and updates the knowledge base appropriately. When a chunk is found to be

repeated in multiple games, we use the function REMOVE to eliminate multiple

instances of the chunk from the list of potential chunks produced from our

convolution algorithm. The induced chunks which do not exist in the current

knowledge base are then written into new frames, created by the function NEW, in

the knowledge base by the function WRITE.

The potential chunks collected by the convolution operator which are not

grouped into a valid chunk to be saved in the knowledge base are written to a

temporary file. This temporary file serves as a database of non-repeating chunks

which is used to identify chunk repetitions that occur outside of the current games

being induced. All chunks in this temporary database are used by the induction

algorithm to group chunks that have not been previously grouped. The database is

deleted after all games from a specific time period or tournament have been

analyzed. The deletion of the database forces the induced chunks to have a temporal

relevance, or recency, to the potential chunks being analyzed.

The textual move chunks are then analyzed to gather the statistics to be used

by the prediction stage of IAM. The statistics that are gathered include the mean

game turn that a move was executed, the number of times a particular move was

observed, the results of the games in which the move was observed, and the color of

the pieces being played. This statistical analysis requires that only one entry be made









For I = 1 to number of chunks found for the current games do
Begin
Flag_1 FALSE ;; indicates when a chunk has been repeated
Flag_2 ,- FALSE ;; indicates when a chunk is duplicated in the KB
For J = I+1 to number of chunks found for the current games do
;; First check this chunk against the chunks that were just found
If C, = Cj
Then Begin
Increment number of times C, has been observed
If Ci.start_time > Cj.starttime
Then Ci.starttime ,- Cj.start_time
If Ci.end time < Cj.endtime
Then Ci.endtime Cj.endtime
REMOVE(Cj)
Flag_l TRUE
End Begin
For J = 1 to number of chunks currently in the chunk Knowledge Base do
;; Next check this chunk against all chunks already in the Knowledge Base
If CI = Cj
Then Begin ;; Update KB knowledge
Increment number of times Cj has been observed
If C1.starttime < Cj.starttime
Then Cj.starttime 4- Ci.starttime
If Ci.endtime > Cj.end_time
Then Cj.endtime C- Ci.endtime
Flag_2 TRUE
End Begin
If Flag_l = TRUE AND Flag_2 = FALSE
Then Begin
NEW(geometric_frame, F)
F C,
WRITE(F, Geometric_Chunk_Knowledge_Base)
End Begin
End Begin
Figure 21: Pseudo-code to perform induction on geometric chunks.






82

in the knowledge base for each specific move. Additional information concerning the

total number of games analyzed, number of games played as white, number of games

won, and number of games won as white is stored with the textual move patterns.

The pawn structure chunks are also analyzed statistically at the same time as

the textual move patterns. The statistical analysis gathers knowledge that is used by

the adversary playing style inference engine. This knowledge is about the presence,

quantity, and quality of pawn structures and inter-locked pawn chains. The quality

of a pawn structure is determined by the number of pawns involved in the structure.

The knowledge from this analysis is stored with the textual moves in the adversary

patterns knowledge base.

Following the induction on geometric chunks and the statistical analysis of

pawn structures and rote learned textual chunks, the knowledge base of patterns

known by an adversary is available for use by the prediction stage of the adversary

model. The adversary pattern knowledge base is constructed incrementally. As new

games of an adversary become available the chunk knowledge from those games can

be quickly incorporated into the existing knowledge base.

4.1.2 Predicting Adversary Actions Stage

After an adversary patterns knowledge base has been constructed, the

adversary model then attempts to predict adversary actions in relevant domain

situations. Simulations of a game played by an adversary are used to verify the

prediction capabilities of the adversary modeling methodology. Each of the

simulated games used to represent a future contest against an adversary has never

been analyzed by IAM.






83
Prior to each of these simulated games, the preferred playing style of the

adversary is inferred from the chunks in the knowledge base. We had IAM study

twelve games of two different chess grandmasters who display preferences for the two

different playing styles currently inferred by IAM. The two players were Botvinnik,

who prefers closed positions, and Spassky, who prefers open positions. By analyzing

the presence and frequency of pawn structures and the frequency of pawn and

attacking moves, IAM correctly identified the playing style preference of each

adversary. Knowledge about the playing style of an adversary is then used by the

current domain program which the adversary model is coaching to select specific

moves that manipulate the game board away from the adversary's preferred style.

During the opening portion of the simulated game, IAM relies exclusively on

the textual move knowledge base which contains the statistical analysis of opening

patterns displayed by the adversary. Prior to the adversary's move, IAM predicts the

adversary's move from the moves currently in the knowledge base. Following the

actual move made by the adversary, the corresponding move in the knowledge base

is eliminated from further consideration. For an adversary playing the white pieces,

IAM has had accurate predictions ranging from forty percent to one hundred percent,

with a mean prediction accuracy of seventy-five percent.

IAM's opening sequence predictions are less accurate for an adversary playing

the black pieces. This is because the adversary is playing responsively to his

opponent's moves. Accuracy for opening move predictions against a black piece

playing adversary ranged from zero to twenty percent.






84

By modifying the textual chunk learning algorithm to remember the white

player's move which preceded an adversary's black move, we were able to

significantly increase IAM's opening sequence prediction accuracy to range between

forty and one hundred percent, with a mean of sixty percent. This modification

produces results which are similar to IAM's performance against the adversary while

playing the white pieces.

The number of textual chunks stored in the knowledge base increases

proportionately, since identical moves by the adversary which follow a different white

move are now considered as unique moves. The average increase in the knowledge

base size is forty-five percent. For example, after analyzing twenty-two games of

Botvinnik with the original textual chunking algorithm, thirty-one opening move

textual chunks were produced. With the response code modification forty-seven

opening move textual chunks were saved to the knowledge base. The actual textual

chunks are also larger to contain the adversary's opponent's move. Because of the

relatively small number of chunks in the knowledge base for a specific opponent, the

increase in prediction accuracy warrants the small increase in size to the knowledge

base.

All move predictions following the opening sequence are made primarily from

the contents of the geometric piece chunk knowledge base. Before each move to be

made by the adversary, the current board representation is searched to find pieces

belonging to the adversary. This is done because each of the chunks contained in the

knowledge base consists of pieces belonging to the adversary.






85
Once an adversary's piece has been located on the board representation, each

chunk in the knowledge base is compared against the board representation to find

identical matches and analogous chunks, which have one or two pieces out of place.

Because chunks occupy a specific square area of the game board, chunks which are

to large to fit into the current board location are ignored to reduce processing time.

The detection of analogous board positions is performed via pattern matching. When

an analogous board position has been identified, IAM then checks to see if a legal

chess move would complete the analogous chunk. An example of this process is

shown in Figure 22. Four of the chunks shown in Figure 6 were used to identify

analogous positions on the current game board which is prior to Botvinnik's

thirteenth move from the game displayed in Figure 19. The corresponding moves

that would produce the chunks are also displayed in the figure.

The collection of moves suggested by the analogous chunks is then processed

by an inference engine to heuristically determine the most probable move. The

inference engine predicted the castling move for the Figure 22 board position. This

prediction represents the result of applying the various psychologically based

heuristics concerning chunk size and relevance to the current domain. The queen

being en prise is domain specific information and is therefore not considered by the

move prediction inference engine. However, the actual move made by Botvinnik,

Qc2 or b3-c2, is one of the moves suggested by an analogous chunk and reviewed by

the inference engine. An experiment to augment the adversary pattern knowledge

base with a minimal amount of domain specific knowledge is presented in section 4.2.































Figure 22: Predictions from four chunks.


The process of searching the current board configuration to identify analogous

positions is repeated before each move to be made by the adversary. Existing chess

programs can use IAM's predictive capabilities by having IAM search the possible

board configuration at a specific node of the game tree and finding analogous chunk

situations in the proposed board configuration. IAM will then predict the most

probable adversary move that corresponds to our domain simplification hypothesis

of Chapter 3.

4.1.3 Integrating IAM's Knowledge into a Chess Program

Our induction based adversary modeling methodology produces knowledge

which can be incorporated into current chess programs by two methods. Knowledge






87
about the specific opening sequences preferred by an adversary can be used to

manually optimize a chess program's opening book. Optimization of the opening

book entails augmenting the opening book with opening sequences that are unknown

to the adversary and adjusting the preferences of the chess program to select these

new opening sequences.

Current chess programs use a move generator to limit the breadth of their

game trees. Move generators suggest a number of moves, typically five, which are

considered as the only moves available from a specific position. We describe the

reduction of tree complexity by the adversary modeling methodology in Chapter 5.

IAM predicts the move that an adversary will execute next from a particular board

position. These predictions can be used by chess programs to further limit the search

of their game trees.

A chess program would have the adversary model make predictions of the

next move in the game tree while simultaneously generating the five best moves with

its move generator. When predictions from the adversary model are available, the

chess program would use the predictions to expand its current game tree. Otherwise,

the chess program would continue to use the moves supplied by its move generator.

Combinations of adversary model predicted moves and move generator suggested

moves can be used by heuristically ordering the moves suggested by both methods

and expanding the game tree with the first five moves from the ordered move list.






88

4.2 IAM's Performance Results


In this section we describe several experiments that have been performed with

IAM to test the overall capabilities of the adversary modeling methodology. Each

experiment is described preceding the analysis of the results.

The games of the adversary Botvinnik are all taken from his World

Championship matches from the 1948 Hague-Moscow tournament through his 1961

match against Tal. The two games which are used to simulate Botvinnik's

performance in a future match, come from Botvinnik's 1963 match against Petrosian,

and are shown in Figure 23 and Figure 24. The game TEST1 is the first game from

the Petrosian match resulting in a win or loss in which Botvinnik played the white

pieces. TEST2 is the next game from the match, following TEST1, in which

Botvinnik played the black pieces.

4.2.1 General Performance Results

The purpose of our first experiment is to determine the base performance

results of IAM so that we can compare these results against other experimental

results. We anticipate that the predictive capabilities of IAM will increase as more

chunks are learned. The records of every game played by Botvinnik from his World

Championship Matches, prior to 1963, are used as input.

4.2.1.1 Performance versus the adversary as the white player

TEST1 is simulated for IAM following the construction of the adversary

model's knowledge base for each of the tournaments of Botvinnik's career. The






89

length in game turns of TEST1 is fifty-seven moves. The performance of IAM in this

experiment is shown in Table 1.


TEST1--Botvinnik playing White

1. d4 d5 21. Re2 Nb6 41. Nc5 Bf5
2. c4 e6 22. Rhel Nc4 42. Kg3 a4
3. Nc3 Be7 23. Bxc4 Rxc4 43. Kf4 a3
4. cxd5 exd5 24. Rd2 Re8 44. Ke5 Rb4
5. Bf4 c6 25. Re3 a6 45. Nd3 Rb5
6. e3 Bf5 26. b3 Rc6 46. Kd6 Kf7
7. g4 Be6 27. Na4 b6 47. Kc6 Bxd3
8. h3 Nf6 28. Nb2 a5 48. Rxd3 Rb2
9. Bd3 c5 29. Nd3 f6 49. Rxa3 Rg2
10. Nf3 Nc6 30. h4 Bf7 50. Kxd5 Rxg5+
11. Kfl 0-0 31. Rxe8+ Be6 51. Kc6 h5
12. Kg2 cxd4 32. Qe3 Bf7 52. d5 Rg2
13. Nxd4 Nxd4 33. g5 Be6 53. d6 Rc2+
14. exd4 Nd7 34. Nf4 Bf7 54. Kd7 h4
15. Qc2 Nf6 35. Nd3 Be6 55. f4 Rf2
16. f3 Rc8 36. gxf6 Qxf6 56. Kc8 Rxf4
17. Be5 Bd6 37. Qg5 Qxg5+ 57. Ra7+
18. Rael Bxe5 38. hxg5 a4
19. Rxe5 g6 39. bxa4 Rc4 1:0
20. Qf2 Nd7 40. a5 bxa5


Figure 23: A game from


the 1963 Botvinnik versus Petrosian match.


The first column, #G, is the number of games currently analyzed to produce

the adversary pattern knowledge base. The #Ch and #M columns are the number

of geometric chunks and textual opening move chunks respectively. The first number

in the #M column represents the total number of chunks and the second number is

the chunks pertaining to white moves, for TEST1, or black moves, for TEST2. The

next three columns are the number of predictions made by IAM (#P), the number

of predictions which exactly matched the adversary's ensuing move (#C), and the




Full Text

USING INDUCTIVE INFERENCE OF
PAST PERFORMANCE TO BUILD STRATEGIC
COGNITIVE ADVERSARY MODELS
BY
STEVEN MICHAEL WALCZAK
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1990

ACKNOWLEDGMENTS
I would like to thank the faculty and staff of the Computer and Information
Sciences Department. Several staff members proved to be invaluable in helping me
cope with the university system. I specifically recognize the assistance of Ms. Jeffie
Woodham, Ms. Marlene Hughes, and Ms. Cheryl Coleman. Ms. Tammera Reedy
and Ms. Judy Longfellow were especially helpful in getting various figures and papers
printed and delivered to their final destinations. Professor Joseph Wilson provided
me with insight into the vision based problems encountered while performing my
research. Other faculty members including George Logothetis, Richard
Newman-Wolfe, Gerhard Ritter, and Ravi Varadarajan encouraged me to openly
discuss research issues in computer science and to enjoy being a student.
Two computer and information science graduate students helped significantly
during the early stages of my research. I would like to thank USA Captain
Christopher Wagner and USN Lieutenant Richard Fióla for there assistance in
uncovering the adversarial modeling techniques used by the armed services.
I would like to give special thanks to my committee members-Manuel
Bermudez, Walter Cunningham, Douglas Dankel, and Ira Fischler. Dr. Cunningham
has been a tremendous source of knowledge and help in establishing the techniques
used by the adversary modeling methodology to infer adversary playing styles. I owe
11

a great debt of gratitude to my committee chairman, Paul Fishwick. He spent many
tireless hours reviewing and analyzing my research and encouraging me.
My greatest human source of aid, comfort, and encouragement during my
graduate studies has been my wife. Karen gave three years of her life to work and
get us both through graduate school. She has my undying love.
I thank God for giving me the wisdom, intelligence, and motivation to pursue
my doctoral degree. The knowledge of His presence in my life has been the single
most important factor in my ability to complete graduate school. I hope and pray
that I will be faithful and wise enough to hear His calling in my life, that I may
choose to follow His path.

TABLE OF CONTENTS
ACKNOWLEDGEMENTS ii
LIST OF TABLES vii
LIST OF FIGURES viii
ABSTRACT x
CHAPTERS
1 INTRODUCTION 1
Problem Statement 2
Contribution of the Dissertation 3
Outline of the Dissertation 7
2 BACKGROUND AND RELATED WORK 9
Induction 9
Tree Search 12
Athletic Adversarial Domains 16
Education Domain 17
Adversarial Game Domains 18
Law Enforcement Domains 21
Military Domains 22
Human Adversary Modeling Efforts in Military
Domains 23
Computer Planning in Military Domains 24
Political and Economic Domains 27
Religion Domain 28
Chess Domain--Our Application Area 28
Human Use of Adversary Modelin in Chess 29
Importance of Pawn Structures and Chess Openings ... 31
Philosophy 33
Psychology 34
iv

Computer Chess and Artificial Intelligence 38
Justification for Chess Domain Application 43
3 ADVERSARY MODELING METHODOLOGY 45
Knowledge Acquisition of Chunks 46
Defining the Chunks to be Acquired 46
Acquiring Geometric Chunks 50
Acquiring Textual Chunks 58
Induction to Learn Repeated Chunk Patterns 62
Acquiring an Adversary’s Playing Style 66
Application of the Adversary Model 68
4 IAM 75
Overview of IAM’s Execution 75
Knowledge Acquisition and Inductive Learning Stage . . 76
Predicting Adversary Actions Stage 82
Integrating IAM’s Knowledge into a Chess Program ... 86
IAM’s Performance Results 88
General Performance Results 88
Performance with Respect to Time 99
Performance with Respect to Textual Chunk Size .... 102
Performance with Respect to Geometric Chunk Size . . 104
The Effect of Domain Specific Knowledge 107
Repeatability of the Demonstrated Results 108
5 RESULTS 110
Reducing Opening Book Size Ill
Revealing New Search Paths in the Game Tree 113
Reducing Search Complexity Through Automatic Pruning ... 116
Summary of Research Contributions 122
6 CONCLUSIONS AND FUTURE RESEARCH 124
Conclusions 124
Future Research Directions 125
Acquire More Geometric Chunks 125
Increase the Knowledge Contained in Each Chunk . . . 126
Place More Adversarial Knowledge in the Adversary
Model 128
APPENDIX 132
v

REFERENCES 135
BIOGRAPHICAL SKETCH 144
vi

LIST OF TABLES
1. IAM Performance Measures for TEST1 91
2. Effect of Likelihood on Prediction Ratios 96
3. IAM Performance Measures for TEST2 97
4. Performance with Responsive Modification 98
5. Effect of Recency on IAM Performance for TEST1 100
6. Effect of Aging/Forgetting on IAM’s Performance 101
7. Performance (TEST1) with 10 Opening Moves 103
8. Performance (TEST1) with 15 Opening Moves 103
9. IAM Performance (TEST1) with Five-by-Five Chunks 105
10. IAM Performance (TEST1) with Six-by-Six Chunks 105
11. Effect of Domain Specific Knowledge 108
12. Ratio of Won and Lost Games for White and Black Ill
13. Statistical Analysis of Soviet Chess Weaknesses 130
Vll

LIST OF FIGURES
1. Logic flow diagram 4
2. Overview of the physical system 7
3. Sample Minimax tree 15
4. Sample cognitive chunk 36
5. Computer chess program ratings 40
6. Chunks acquired for Botvinnik 49
7. Internal representation of a board position 52
8. Convolution templates 53
9. Pseudo-code to perform convolution 53
10. Step functions 55
11. Pseudo-code to perform step function 55
12. Pseudo-code to perform chunk coalescing 56
13. Chunks on a chess board and their convolution values 57
14. Chunk acquisition in Go 59
15. Flowchart to collect textual move knowledge 61
16. The effect of noise on chunks 63
17. Four identical chunks 64
18. Fuzzy logic used by the heuristic rules 67
viii

19. An example of a game record--the 1948 Hague-Moscow Tournament. . . 77
20. Chunk occurrence in three different games 79
21. Pseudo-code to perform induction on geometric chunks 81
22. Predictions from four chunks 86
23. A game from the 1963 Botvinnik versus Petrosian match 89
24. Another game from the 1963 Botvinnik versus Petrosian match 90
25. Pseudo-code to suggest possible adversary moves 94
26. Chunk acquired with 5x5 size and corresponding smaller chunk 106
27. Hypothetical game/search tree 114
28. Prediction of Botvinnik’s next move 116
29. Standard search tree and the IAM improved search tree 118
30. Search tree obtained from Correct Piece identification 121
31. Knight template and partial Rook template 127
IX

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
USING INDUCTIVE INFERENCE OF
PAST PERFORMANCE TO BUILD STRATEGIC
COGNITIVE ADVERSARY MODELS
By
Steven Michael Walczak
December 1990
Chairman: Paul A. Fishwick
Major Department: Computer and Information Sciences
To perform optimally in adversarial domains, artificial intelligence programs
must be able to identically evaluate domain situations and develop plans as the
adversary. Knowledge about an adversary’s evaluation criteria and objectives is
difficult to obtain since an adversary will guard this information in order to promote
a favorable outcome for himself. We present an adversary modeling methodology
which is capable of acquiring the evaluation criteria of an adversary through
observations of the adversary’s past performances in the domain.
Our adversary modeling methodology uses induction to capture perceptual
chunks that are repeatedly displayed by a specific adversary. The acquired chunks
are then used to predict the adversary’s actions during future encounters. Pattern
recognition is used to identify perceptual chunks in the current adversarial domain
x

situation which are analogous to chunks that have been previously learned from the
current adversary.
We have implemented our adversary modeling methodology in the program
IAM, Inductive Adversary Modeler, which has been applied to the domain of chess.
When high quality knowledge is available about an adversary’s past performances,
the adversary modeling methodology accurately predicts ten percent of an adversary’s
future action plans. The ability to accurately predict an adversary’s move choices
reduces the complexity of the game tree search which enhances the playing ability
of current chess programs.
xi

CHAPTER 1
INTRODUCTION
Adversarial situations surround us in varied forms on a daily basis. Examples
of adversarial situations include vying for a promotion at our place of employment,
Monday night’s football game, and military conflicts throughout the world. These
adversarial situations are characterized by a minimum of two actors each of whom
is attempting to realize a goal that conflicts with the goal of the opposing actor. The
actors can be individuals or groups of individuals such as a football team players that
are cooperating to achieve the goal of the actor.
When humans are faced with the problem solving task of realizing a goal
against the desires of an adversary, they employ a variety of techniques that enable
them to outperform their adversary and accomplish their goal. The standard
cognitive problem solving methods of difference reduction and working backwards
are insufficient in adversarial domains. Human competitors use deductive reasoning
and analogy to previous domain situations to realize their goals in adversarial
domains. The use of analogy to gain a strategic advantage over an adversary
necessitates the studying of a specific adversary’s previous performance, preferably
in situations analogous to the upcoming contest.
Human competitors performing in adversarial domains study the past
performances of their adversaries to identify strengths and weaknesses which can be
1

2
exploited respectively to gain a strategic or tactical advantage. This information is
preserved by the competitor as a model of the likely actions an adversary will
perform in particular situations. The model of the adversary’s probable actions is
then exploited when an appropriate situation occurs during the course of the
competition. Such situations arise when the current game state-situation in the
adversarial domain-is analogous to a game state which occurred in the adversary’s
past.
1.1 Problem Statement
Current artificial intelligence application efforts in adversarial domains are not
performed optimally. A factor which keeps adversary-based programs from
performing at expert levels is the lack of knowledge about the potential adversary.
Lenat and Feigenbaum (1987) claim that domain specific knowledge is required to
solve difficult problems and that the performance of intelligent programs improves
as additional domain knowledge is utilized.
Acquiring knowledge for programs in adversarial domains is hindered by the
fact that adversaries are unwilling to reveal their tactical and strategic goals, since
this would be counter-productive to the attainment of their goals. Candland (1980)
and Bole (1987) have claimed that knowledge about any domain is available from a
wide variety of sources. Lenat (1982) proposes the use of the observable actions of
the adversary while performing in the adversarial domain. The proper selection of
adversarial domain situations to observe guarantees the acquisition of high quality
knowledge.

3
For the domain of chess, lack of knowledge about an adversary’s tactics and
strategies is analogous to searching the game tree to a depth of only one ply. All
moves are considered as winning moves and the game proceeds in a random manner.
To acquire knowledge for use in adversarial domain application programs, we will
emulate the traits of human experts performing in adversarial domains, by analyzing
the previous performances of a particular adversary in relevant domain situations.
1.2 Contribution of the Dissertation
Adversarial domain programs require knowledge about an adversary’s tactics
and strategies to continue to improve their performance. We need to consider two
separate factors that affect the strategic decisions an adversary will make. An
adversary’s strategic decisions are affected by the plans or scripts that the adversary
feels will satisfy the current strategic goal. Before the adversary can select a plan of
action however, the domain situation must be evaluated to determine which plans are
appropriate to that particular situation.
Current programs performing tactical and strategic planning for adversarial
domains rely heavily on tree structures and search techniques such as iterative
deepening and alpha-beta pruning. The ability to accurately predict the choices that
will be made by an adversary in specific domain situations reduces computational
complexity and increases the depth of search. The increased depth of search that our
research enables is directly responsible for the augmented tactical ability of these
search based programs.

4
The purpose of our research is to develop a generalized method for modeling
the strategic decision making process of an adversary. The information flow diagram
for our adversary modeling methodology is shown in Figure 1. Our adversary
modeling methodology makes use of knowledge which is currently not used by
programs operating in adversarial domains. By increasing the knowledge available
to tactical and strategic planners, we will improve the overall performance of these
adversarial domain applications.
figure 1: Logic flow diagram.
Our adversary modeling methodology uses psychological principles of
organization and evaluation such as the Gestalt principles of organization (Anderson,

5
1980, and Pomerantz, 1986) to emulate the primary evaluation criteria used by an
adversary. Kodratoff and Michalski (1990) note that recent machine learning efforts
have focused on building cognitive learning architectures. We intend for the
adversary modeling methodology to be applicable across all adversarial domains and
hence, we include as little domain specific knowledge as possible. This means that
we are content to acquire the primitive elements that an adversary uses in evaluating
a domain situation without embellishing those elements with further domain specific
knowledge.
The psychological principles of organization are applied to a collection of
previous adversary performances in the domain (e.g., chess games played by the
adversary in prior tournaments and matches). We acquire the cognitive chunks used
by an adversary to evaluate the current domain situation. The acquired cognitive
chunks of the adversary help us to view the domain the same as our adversary and
to avoid influencing domain evaluations with personal biases.
We now have a large collection of chunks that have been displayed by the
adversary. The chunks are either textual patterns corresponding to the adversary’s
verbal memory or geometric patterns corresponding to the adversary’s visual memory.
The chunks we have acquired contain patterns that have been used only once and
may have occurred by chance. We use induction to eliminate these chance chunks
and create a valid and useful adversary model. The inductive phase identifies chunks
which are repeated in two or more games. The repetition of patterns is used to
indicate the probability that an adversary will use the acquired chunk in future

6
evaluations of domain situations. Because we are using induction as our learning
method, the adversary model created will be used as a heuristic guide instead of an
algorithmic solution.
Following the inductive learning, the collection of chunks that have been
repeatedly displayed by the adversary are stored in the adversary model. We
consider these chunks to be the primitive elements used by an adversary in evaluating
current domain situations. The chunks of the adversary model are then used to
predict adversary decisions when analogous situations occur in the adversarial
domain.
The knowledge obtained by the adversary modeling methodology is highly
reliable, because it comes directly from the adversary. A certain level of care must
be taken to guarantee that the adversary is performing normally in the domain and
not providing misinformation to the modeling methodology. We verify the adversary
modeling methodology by implementing the methodology for the adversarial domain
of chess. The implementation is called IAM, for Inductive Adversary Modeler. Our
justification for selecting the chess domain is given in section 2.10.6.
Berliner and Ebeling (1988) note that speed is a critical factor for evaluation
functions in chess as well as other adversarial domains. The knowledge that we are
learning about an adversary is available as historical records prior to the current
performance of the adversarial domain program. Therefore, the adversary modeling
methodology’s knowledge acquisition can be performed prior to its use by the domain
program. This preprocessing effectively eliminates any time or complexity costs

7
generated by the adversary modeling methodology from being added to the current
program’s evaluation function cost (Aho, 1974). The adversary model acts as a
heuristic coach to an existing domain program by predicting which moves an
adversary will make, as shown in Figure 2. The only part of the adversary model that
impacts the time order of the existing evaluation function is the analogous situation
detector that was shown in Figure 1.
Game
Figure 2: Overview of the physical system.
1.3 Outline of the Dissertation
We have presented a general overview of the problem of the lack of
knowledge about adversaries available to intelligent programs operating in

8
adversarial domains. Our solution to this problem has been briefly presented in this
chapter.
In Chapter 2, we present background material on induction and psychology
that affects our research. The research that has already been performed in several
of the adversarial domains including the domain of chess is discussed. Finally, we
present the reasons for selecting the domain of chess as our application domain.
In Chapter 3, we detail our research by first discussing the implementation
method and then the operation of the adversary modeling methodology. Each of the
acquisition and induction steps that have been briefly described in this chapter are
presented in detail.
Chapter 4 contains an example of the adversary modeling methodology in
action and provides quantitative performance results of our application program,
IAM. Our results and their impact on artificial intelligence search methods are
provided in Chapter 5. We discuss our conclusions and future research directions in
Chapter 6. The Appendix contains a glossary of chess and artificial intelligence
terms to aid the reader in understanding some of the domain descriptions that we
have presented.

CHAPTER 2
BACKGROUND AND RELATED WORK
In this chapter, we review the theoretical and applied work on induction. We
also describe the theoretical work on game tree search to emphasize our contribution
to this research area. Finally, we describe related work in the various adversarial
domains that have led to our solution with a detailed review of the domain of chess
which serves as our application domain. For each of the adversarial domains
presented, we describe the use of adversary modeling by humans and any applied
research that affects the implementation of our adversary modeling methodology.
2.1 Induction
Mankind has been using induction to perform problem solving and learning
tasks for many millennia. Rosch (1973) describes the way humans form concepts and
categories about their natural world. Polya (1973), in the early 1950s, formalized the
concept of induction. Induction is the process of discovering general laws or concepts
from observations and combinations of particular instances.
An example of induction for a natural world concept would be the
generalization that all ravens are black. This generalization is based on the
observation of a limited number of instances of a raven. Without seeing all the
possible examples or instances of raven, we are confident to induce that the color of
9

10
a raven is the same as the color of the instances we have seen--black. The strength
of our belief increases as more and more examples of black ravens are observed.
However, only one instance of a nonblack raven needs to be identified to invalidate
our induction.
From the above example we can see that inductive inference is learning in the
limit (Daley & Smith, 1986). We need to see sufficient observations of the right type
to form a correct generalized concept. Since we cannot guarantee that the right type
of observations are being provided to an inductive inference system, we must rely on
the use of a large number of examples of category instances to acquire the desired
concept. This means that the adversary modeling methodology we have developed
should perform more accurately as we provide greater numbers of examples of an
adversary’s domain performance.
Since Polya’s formalization of the concept of induction, theoretical research
on induction has primarily focused on the formation of concepts or classifications.
Angluin and Smith (1983) and Case and Smith (1983) present rigorous discussions
of the mathematical principles of induction. This theoretical background provides
a solid base for the use of inductive inference as a computational tool for learning
classes of objects or ideas.
The extraction of general principles from specific examples is a major research
goal in machine learning (Anderson, 1987). Induction has been applied to many
domains. Examples of applied inductive inference algorithms are Michalski’s
INDUCE and Quinlan’s ID3 (Michalski et al., 1983 and 1986). Each of these

11
example applications attempts to classify instances of a domain problem (e.g.,
soybean diseases for the INDUCE program) into maximally-specific groups that will
identify all known instances as well as accounting for unknown instances.
Prior to the theoretical discussions of inductive inference mentioned above,
Blum and Blum (1975) had already demonstrated that the ability to play games, or
at least to make technically correct moves, can be learned through inductive
inference. Induction has not been utilized in game domains due to the lack of high
quality play produced by Blums’ algorithm. Our adversary modeling methodology
focuses on a narrower slice of the game domains, namely predicting adversary move
choices and playing style, and is able to learn high quality knowledge.
Adversarial domains present several specialized problems to the standard
inductive inference algorithms like ID3. Langley (1985) has noted that feedback is
an essential element of most learning methods. Inductive inference algorithms use
feedback to verify acquired concepts and to identify concepts that need to be
modified. In adversarial domains, feedback on the performance of an algorithm is
a not attainable until after the current domain situation has ended. Because of the
inherent time delay for feedback in adversarial domains, we substitute probabilistic
reasoning so that our confidence in the current adversary model increases as
additional examples of the same type are observed.
Another problem in adversarial domains is the dependence of the domain on
time. Actions occurring in the domain are temporally related. The inductive
inference applications cited above belong to the class of hierarchical or structure

12
induction methods. Structure induction methods are used to produce classification
systems. Muggleton (1990) proposes an inductive method called sequence induction
which has a time effect relationship to the domain. Sequence induction uses a series
of domain descriptions that are altered by actions operating on the domain to
produce control systems.
Mitchell et al. (1986) refer to a process of mechanized induction which
receives domain events as input and produces concepts for predicting future events.
For adversarial domains, sequence induction provides a method to produce
evaluation criteria for predicting an adversary’s strategic decisions.
2.2 Tree Search
Programs in the adversarial domains of games invariably use trees as their
knowledge representation scheme. Finding solutions or move choices in game trees
involves searching the tree for the optimal finishing value. Von Neumann and
Morgenstern (1953) published the Minimax algorithm for backing the final game
values up through the game tree to the current node. Minimaxing attempts to
account for the effects of an adversary in the game domain by alternately raising the
minimum possible value or maximum possible value at each level of the tree.
If an exhaustive search can reach all of the leaf nodes of the tree, then the
true finishing value of the game can be raised to the current node and an optimal
game will ensue. However, games like chess, Go, and Othello have tree
representations that hold 1040 nodes or more. Game trees of this size prohibit an

13
exhaustive search. Instead, game programs will search down to a preselected depth
and apply an evaluation function that will determine a best guess for the true
finishing value obtainable from that node.
Evaluation functions use domain knowledge to predict the outcome of the
game from a specific position. All domain specific knowledge is contained in the
evaluation function. The quality of knowledge has a direct effect on the playing
ability of the game program (Berliner & Goetsch, 1984, and Lee & Mahajan, 1988).
Northwestern University’s CHESS4.6 was able to defeat other chess programs which
searched several ply (levels) deeper into the game tree based on the strength of its
evaluation function (Elithorn & Banerji, 1984, and Wilkins, 1980).
The primary problem faced by search algorithms using tree structures is the
horizon effect. The horizon effect occurs when the solution to a search problem lies
at a depth in the tree that is beyond the current search depth. This is a realistic
problem for complex adversarial domains. In chess, the KRKN endgame, king and
rook versus king and knight, can require a twenty-seven move solution or a game tree
search depth of fifty-two ply (Bratko & Michie, 1980). Similarly, the KBBKN
endgame has solutions of sixty-six moves or one hundred thirty-two ply search depth
(Roycroft, 1988). Current chess programs search between eight and fourteen ply
before selecting a move.
Two different schools of thought exist concerning methods for continuing to
improve game program performance. The first school believes that the depth of the
search is the sole criterion affecting the playing ability of game programs.

14
Adelson-Velsky et al. (1988) state that increasing the depth of search of a game tree
increases the quality and reliability of game play. Berliner and Ebeling (1988)
specifically state that a deep tree search will outperform a shallower search which
uses more knowledge. Various techniques are used to increase the depth of search
for a program. Limiting the breadth of search and pruning of undesirable branches
through alpha-beta pruning and iterative deepening permit a deeper search for a
static number of evaluated tree nodes (Barr & Feigenbaum, 1981). Specialized
hardware can also be used to increase the speed and thereby the depth of search
(Ebeling, 1986).
An example of alpha-beta pruning is shown in Figure 3. The four leaf nodes
which have their branches cut by a diagonal line can each be pruned using alpha-beta
pruning. Once the three value is elevated to the left subtree level two node, then the
level one node is guaranteed a minimum value of three. Since each of the first leaf
nodes below the center and right subtrees have lower values than three, the
remaining leaf nodes do not need to be evaluated. This is called an alpha cutoff.
A beta cutoff is similar except that the cutoff is performed based on a guaranteed
maximum value for a Min node.
The other method for improving the performance of game programs is to
increase the knowledge that is available to them. We have already seen that better
quality knowledge enabled CF1ESS4.6 to outperform other programs that searched
several ply deeper into the game tree. The use of additional knowledge has the same
effect as searching deeper into the game tree. The projection ability of certain types

15
LEMEL
Figure 3: Sample game tree with alpha-beta pruning.
of knowledge in the chess domain is nine ply for tactical knowledge, twenty-five ply
for positional knowledge, and forty-five ply for strategic knowledge (in Elithorn &
Banerji, 1984). Current game programs use mostly tactical knowledge along with
some positional knowledge.
For adversarial domains, the use of domain knowledge can be dangerous by
leading us into a false sense of security. Samuel (1959) warns that programs using
the Minimax algorithm must take into consideration the intent of the adversary.
Cecchini (in Crookal et al., 1987) claims that the actual intent of the adversary will
require programs to use different heuristics in order to win. Human tree search is
goal directed and the direction and method of search may change as new information

16
becomes available (Frey, 1983). This last statement means that the preceding events
in a game and tournament affect the future strategic decisions of the adversary.
Our adversary modeling methodology follows the school of thought which
believes that more knowledge is required to continue improving the performance of
game programs. To avoid the complications of adversarial intent, our heuristics are
based on the primitive psychological elements that are used by an adversary to
evaluate the domain. Application of the adversary modeling heuristics occurs when
the current game board pattern is analogous to a previously induced situation. The
use of pattern recognition to increase search depths has already been implemented
by Samuel (1959 and 1967) in the form of signature tables and in the Russian chess
program KAISSA (Adelson-Velsky et al., 1975) to perform cutoffs or pruning.
The theoretical increase in search depth which we have achieved through the
adversary modeling methodology is detailed in Chapter 5. Several authors (Findler
& Meltzer, 1971, and Utgoff, 1986) have noted that the economy of a search
heuristic can be measured by the reduction in search effort and complexity which
produces a deeper search.
2.3 Athletic Adversarial Domains
The modeling of an adversary by humans is well documented in athletic
domains. Coaches of baseball (Aliston & Weiskopf, 1984), football (Riggins &
Winter, 1984), and other sports use scouting reports and films of their opponents to
study their opponent’s strategies. This enables the coaches of athletic teams or

17
individuals to develop counter-strategies to foil the goals of the adversary and
promote their own goals.
In a personal communication, Owen J. Holyoak of the Exercise and Sport
Sciences Department at the University of Florida stated that the amount of
preparation performed by athletic team coaches corresponds directly to the perceived
threat of the adversary. When the technique of studying an adversary’s past
performance is not fully used, surprising results can occur. At the 1982 homecoming
football game for the University of Florida (UF), the opponent was West Texas State
University (WTS). Although WTS was considered to be an underdog and was not
expected to score any points, they obtained a fourteen point lead early in the game
by using some plays that UF had not anticipated. The lack of preparation on UF’s
part was responsible for the surprising start of the 1982 homecoming game.
2.4 Education Domain
The artificial intelligence domain of intelligent tutoring systems (ITSs) or
intelligent computer-aided instruction has already made successful use of models of
human thinking. These models are based on the availability of knowledge to the
human and that human’s mental set at a particular point in time. The people being
modeled in ITS programs are the students who are attempting to learn. The student
models are used to predict the sources of error in student thinking which cause
incorrect responses to test items produced by the ITS.

18
Woolf (1984) has demonstrated an ITS which uses a model of the probable
knowledge currently held by the student for discourse planning. This model adapts
its understanding of the student’s knowledge during the lesson to capture changes in
the student’s cognitive state precipitated by the acquisition of new knowledge. Our
adversary modeling methodology operates similarly by constructing a model of the
probable evaluation mechanisms known to an adversary to predict future strategic
decisions of the adversary. Other examples of this modeling approach as used in
ITSs can be found in Sleeman and Brown (1982).
2.5 Adversarial Game Domains
Because of their finite size, games provide us with a convenient abstraction
of more difficult real world problems. The former Japanese use of the game of Go
to train their military leaders (Reiss Games, 1974) demonstrates the practical
extension of a game domain to a military domain. Pritchard (1973) elaborates
further by claiming that Go is guerilla warfare on a grand scale.
We will be using two person zero-sum perfect information games (Zagare,
1984) to demonstrate the adversary modeling methodology. Zero-sum games are
always adversarial in nature, since the wins of one side are equivalent to the losses
of the other side. Our selection of perfect information games was made to permit
an algorithmic evaluation of the results. The domain of chess which we will use for
our application of the adversary modeling methodology is reviewed in section 2.10.

19
Other types of games, such as games of chance, that do not fit our selection
of zero-sum perfect information games can also benefit from the adversary modeling
methodology. Our contribution which reduces the complexity of search is usable by
any game program that utilizes game tree search and heuristics to solve domain
problems. Berliner’s BKG backgammon program (in Levy, 1988) is an example of
a game of chance which uses heuristic methods and game tree search to select
between moves.
Christensen and Korf (1986) claim that heuristic methods for one person
puzzles are equivalent to heuristic methods for two person games. The adversary
modeling methodology is capable of predicting moves in one person games that will
reduce the complexity of the current game position to one which has a known
solution. Other research using induction of previous games has been performed for
one person games. PREFER (Utgoff & Saxena, 1987) solves the 8-puzzle by
performing induction on previous 8-puzzle solutions.
Samuel (1959 and 1967) was the first researcher to use machine learning to
create an intelligent game playing program. Samuel’s checker playing program uses
rote learning techniques to learn how to play checkers. This program is capable of
imitating the style of an adversary by mimicking the winning moves of the adversary
presented in the set of training instances. Programs which rely on rote learning tend
to become inflexible and require an equally large number of retraining instances to
adapt to a different style of play. The method of mimicking an adversary has been
shown to fail for the game of Go (Pritchard, 1973) and should not be considered as
a generalized method of strategic planning against adversaries.

20
Samuel developed signature tables to reduce search complexity. Signature
tables are used to remember the evaluation function value for specific positions.
When identical positions are identified through a pattern matching mechanism, the
signature tables are used to retrieve the corresponding node value. Wilcox (1985)
has also applied pattern recognition techniques to the game of Go to construct
NEMESIS, a strong Go playing program.
IAGO (Rosenbloom, 1982) uses the standard search techniques of alpha-beta
pruning and iterative deepening to play championship-level Othello. Rosenbloom
points out that the heuristics for IAGO need to change during different phases of the
game to capture shifts in the strategic importance of various game principles. For
example, in the middle game of Othello mobility is a critical factor, but in the end
game portion of Othello the quantity of stable pieces is the primary concern.
Berliner (1979) notes similar shifts in strategic reasoning for the domain of chess.
Our adversary modeling methodology compensates for the strategic shifts at various
points in the game by assigning time durations to the domain evaluation chunks that
are acquired for each adversary.
GINA (DeJong & Schultz, 1988) tries to learn to play Othello by using
experience gained from previous games. This approach is very similar to the one
used by Samuel for his checkers program. GINA attempts to exploit the weaknesses
of a specific opponent. Several of GINA’s implementation features are applicable
across adversarial domains. These features are:

21
• Serve as an external coach to an existing program in the domain.
• Only use observable behavior to learn about an adversary. For GINA,
the observable behavior is the moves made by an adversary in prior
games.
• Do not use feedback to modify learning, except for the results of the
games that have been studied.
The best example of strategic planning based on an adversary’s modus
operandi is Waterman’s (1970) poker player. Waterman has implemented a draw
poker playing program which analyzes an opponent’s style of play to decide when a
bet is appropriate. This program learns over a period of several hands which
heuristics to use against a particular adversary. We expand upon Waterman’s idea
by not limiting the program to a static number of predefined heuristics.
2,6 Law Enforcement Domains
Police departments and the Federal Bureau of Investigation each make use
of adversary modeling. As an example, a police detective will try to predict when,
where, and to whom a serial killer will perform his next crime. This is done
primarily by analyzing the criminal’s previous performances of crimes. Similar
techniques are being considered for predicting the outcome of possible terrorist
activities in the RISK project sponsored by Lawrence Livermore National
Laboratories.

22
The key to successful drug enforcement is intelligence gathering and
information processing (Witkin, 1990). Raw data is voluminous and must be
converted to a useable form. Our adversary modeling methodology is capable of
finding patterns in raw data that correspond to resource movement problems (e.g.,
the actual drugs and the monetary compensations) and can be used by law
enforcement officials to increase their effectiveness.
2.7 Military Domains
Devastating results occur on the battlefield when sufficient preparation against
an adversary is not performed. Due to different sociological, geographical, and
educational backgrounds, potential adversaries will have unique behavioral rules
(Bond, 1986 and Boring, 1945). These behavioral rules govern varying beliefs and
goals concerning war and the use of nuclear weapons (McMillen, 1984 and
Sokolovskiy, 1968). Dunnigan (1982) and Ermarth (1978) state that if we assume
that an enemy will act strictly according to our own beliefs, then we will produce a
tactical blindness. Tactical blindness produces the inability to predict or fathom the
plans of our adversaries. Historic examples of the tactical and strategic blindness
produced by conscripting our own beliefs onto an adversary are the bombing of Pearl
Harbor by the Japanese in 1941, the defeat of the United States tactical forces at the
Chosin Reservoir during the Korean War (Wallace, 1990), and the Tet Offensive of
the North Vietnamese in 1968. Many soldier’s lives might have been saved early in
World War II if the United States soldiers had understood the Japanese attitude
toward surrender (Boring, 1945).

23
2.7,1 Human Adversary Modeling Efforts in Military Domains
Military leaders (Gruner, 1988, and Ryan, 1988) claim that victory is often
contingent upon knowing your enemy. This means that military planners must be
able to evaluate the current situation using the same criteria as their adversary. A
vital component of intelligence information is briefings on an adversary’s operational
characteristics or tactics (Gruner, 1988). Robert E. Lee, who is often cited as the
greatest military strategist (B. Davis, 1956), is a practical example of the power
accompanying detailed knowledge about an adversary. Lee was educated at West
Point, where all of the Union commanders were also trained, and served along side
many of the adversaries he would face in the Civil War during his tenure as an
officer in the United States Army and Cavalry (Snow, 1867). Lee’s detailed
knowledge of the training and tactics of his adversaries enabled him to frequently
outperform better equipped and larger forces.
The United States military makes use of this concept in training our soldiers.
Each service branch maintains a cadre of personnel who are trained in Soviet-style
war tactics and strategy (Robbins, 1988). The training centers for the Army,
Air Force, and Navy are the National Training Center, Red Flag, and the Top Gun
school respectively. The purpose of these Soviet analogous fighting forces is to
provide a realistic training environment to better prepare the military for the
possibility of an actual war against Soviet forces or forces that use Soviet-style tactics.
A historic example of adversary modeling for military purposes occurred
during World War II. A detailed psychological profile of Hitler was constructed by

24
Langer (Langer, 1972) for the Allied forces. This model of Hitler was constructed
solely from second-hand information. The psychological model of Hitler accurately
predicted the method of his ultimate demise via suicide.
2.7.2 Computer Planning in Military Domains
Military games and strategic and tactical planners make use of varying levels
of adversary modeling, ranging from none in typical multi-player games and training
simulators to moderately advanced models in certain strategic planning programs.
Failing to account for the actions of an adversary severely inhibits strategic planning.
Military planning tools must account for the probable actions of an adversary based
on the adversary’s intentions and strategic style, otherwise tactical blindness will
result.
The tactical troop movement planner TACPLAN (Andriole et al., 1986) is
typical of most military planning tools. TACPLAN only accounts for static elements
of the domain such as the effect of terrain on equipment movement. Knowledge
about adversaries is nonexistent. The next step towards an adversary model is
programs similar to ARES (Young & Lehner, 1986) which acknowledges the
presence of an adversary. ARES is also a tactical planner; however, the plans
generated by ARES attempt to account for the possible blocking actions of an
adversary. Although this approach claims to account for the actions of an adversary,
the actual outcome is not that different from the plans generated by programs like
TACPLAN. The lack of difference between the two approaches is caused because
the actions of an adversary are prescribed in advance which is similar to only

25
permitting a chess adversary to make forward moves. Infinitely many strategic
choices can be made by an adversary that will not be covered by the prescribed
action choices that are permitted by the tactical planning algorithm.
The POLITICS system (Carbonell, 1981) proceeds in the development plan
by introducing goals that an adversary is trying to accomplish, such as blocking the
current goal of the planning program. The adversarial goals used in POLITICS are
statically defined and are based on preconceived notions of an adversary’s intent.
Basing the actions of an adversary on static predefined goals runs the risk of creating
tactical blindness mentioned in the introduction to section 2.7. The actual goals of
an adversary, much less how the adversary intends to accomplish those goals, may not
be known during the contest or may be contrary to the expected goals (Crookall et
al., 1987 and Narendra & Thathachar, 1989). POLITICS does provide a good
example of a program in which a thinking adversary attempts to achieve specific
goals.
TM (Erickson & Zytkow, 1988) is another goal based system for making
tactical decisions. The program BOGEY is an external learning program which
generates large numbers of simulations of tactical engagements with specific goals.
BOGEY, like GINA in section 2.5, is another example of the utility of coaching
programs that support existing domain specific programs through an external learning
function.
The MARK I and MARK II systems (Davis, 1988a and 1988b and Davis et
al., 1986) embody the mind set of an adversary in DEWT, Deterrence Escalation

26
control and War Termination, models. Currently the MARK systems have two
decision models, East and West. While MARK 1 and MARK II attempt to capture
differences in the mind sets between adversaries, the models are statically defined
like the goals of the POLITICS program. These static mind sets are based on
predefined conceptions of an adversary’s intentions which may not be accurate and
will not account for the modification over time to decision heuristics used by an
adversary. One of the contributions that artificial intelligence can provide to military
planning is reasoning that accounts for the temporal relevance of actions (Bonasso,
1988).
The use of multiple models to generate goals for different adversaries
demonstrated by the MARK systems is a reasonable approach to separate the varying
evaluation criteria used by various adversaries. An improvement to the MARK
systems is accomplished by acquiring the evaluation criteria and traits of particular
adversaries in specific situations along with the temporal relevance for each of the
evaluation criteria of an adversary.
We mentioned in section 1.2 that geometric or visual relationships between
domain resources are used by the inductive learning mechanism of the adversary
modeling methodology. This means that military domains are required to represent
the physical resources of the domain in a geometric manner. The services are
trained to use standardized formations such as the column, echelon, line, and wedge
formations (Department of the Army, 1986). Typical board wargames make use of
a hexagonal map for plotting the movement of resources. By using such map

27
representations, the use of specific formations by a particular commander can be
acquired by our adversary modeling methodology.
2.8 Political and Economic Domains
Top level policy decisions made by government officials are also affected by
prior knowledge about an adversary’s goals and strategies (Cimbala, 1987).
Government negotiators attempt to discover the limitations faced by an adversary to
accomplish the optimal outcome of an agreement or treaty between the two
adversaries. The use of knowledge of the political mind set for a geographic region
permits negotiators to interpret information from the proper perspective (Cimbala,
1987).
The political use of adversary modeling can also be extended to corporate
politics. We can acquire patterns from adversarial actions such as corporate
takeovers or mergers. This knowledge enables corporate entities to be better
defended or prepared for such adversarial actions.
Other domains that are concerned with the geometric relationships of
elements in the domain can benefit directly from the geometric pattern acquisition
phase of the adversary modeling methodology. Examples of such domains include
city planning and architecture. For the city planning domain, patterns of streets from
other cities which have desirable and equivalent traffic patterns can be acquired for
use in the current city. The primary benefits to these domains will be economic
savings and time savings for the reduced complexity of the design phase for
individual projects.

28
2.9 Religion Domain
The importance of knowing the intentions and strategic style of an adversary
has been recorded throughout history. An early example of these writings comes
from the first century A.D. when the apostle Paul (1 Corinthians 9:19-22) writes:
that I might win the more. And to the Jews I became as
a Jew, that I might win Jews; to those who are under the
Law, as under the Law ... to those without law, as
without law ... To the weak I became weak, that I
might win the weak.
Paul is stating that he will model the lifestyle of the people he is attempting to win.
The training centers for the military at the National Training Center, Red Flag, and
the Top Gun schools are modern examples of the approach used by Paul.
2,10 Chess Domain-Our Application Area
We mentioned in section 2.5 that games like chess can be thought of as
abstractions of actual military contests. Although the outcome of these games may
be viewed as trivial compared to the outcomes of military conflicts, personal
reputation, and financial prosperity are the stakes of such games for expert game
players.
In the remainder of this section, we demonstrate the use of adversary
modeling by humans in the chess domain and the philosophical and psychological
foundations that enable the adversary modeling methodology to operate in the chess
domain. Finally, we present a detailed background of past and current computer
chess programs.

29
2.10.1 Human Use of Adversary Modeling in Chess
Before playing for the World Chess Championship against Capablanca who
was considered to be invincible and was appropriately called "The Chess Machine",
Alekhine studied and analyzed Capablanca’s prior games (Schonberg, 1973). From
his study, Alekhine was able to determine a weakness in Capablanca’s play that he
was able to exploit to become the new World Chess Champion. Other chess masters
including the World Chess Champions Botvinnik, Tal, and Kasparov prepare for
chess matches by rigorously studying the previous games of their opponents
(Horowitz, 1973).
Nunn (Nunn & Griffiths, 1987) describes a game against 0st-Hansen in which
both players based their game strategy on expectations of their adversary’s playing
style. These expectations were formed from prior experience against the adversary.
Although no one was able to capitalize upon the information, it has been noted that
prior to his World Championship match against Spassky, Fischer varied his standard
opening for white from e4, pawn to King four, only three times during tournament
level play (Evans, 1970).
Several of the chess grandmasters, including Lasker and Tal (Horowitz, 1973
and Schonberg, 1973), have been described as psychological players. These players
would intentionally make the move that would be most disturbing to their opponents,
even though such moves were frequently not the best tactical choice. Levy (1984)
has described his style of play against computer chess programs similarly, stating that
his move choices are often made to confuse the computer.

30
The advent of intelligent programs in various game domains, including chess,
has caused humans expert game players to consider the computer as an adversary.
Subsequently, human competitors who will face an intelligent computer program as
an adversary have created models of adversary strategy and playing style that
represent the computer game programs. Certain human competitors are considered
to be specialists against computer styles of play.
A famous example of this specialization is the series of bets which David Levy
(1984) has had with the computer chess community. In these bets, Levy claims that
no computer chess program will be able to beat him. The first bet which had a ten
year duration was won by Levy, however, DEEP THOUGHT has defeated Levy in
the latest of the bets. A recent discussion on the rec.games.chess news network by
Feng-Hsiung Hsu indicates an increase in the number of human opponents of
computer chess programs using "anti-computer" playing styles against programs
entered in major chess tournaments.
When Mike Valvo defeated DEEP THOUGHT, the reigning computer chess
program champion for the past three years, in a two game postal chess match via
electronic mail, it was noted (Newborn & Kopec, 1989) that Valvo had observed
DEEP THOUGHT play a number of games and had acquired the playing style of
DEEP THOUGHT. The chess program did not have a similar opportunity. Berliner
(1988) claims that various chess players have asked for the previous game records of
HITECH, the 1989 computer chess program co-champion, so that they might be
prepared for a possible match.

31
2.10.2 Importance of Pawn Structures and Chess Openings
We pay special attention to pawn formations during our chunk learning
process. Philidor has stated that the win or loss of a game depends on the quality
of the pawn positions during the game (Pachman, 1975). The strength and weakness
of pawn formations is of paramount importance (Lasker, 1973). Pawn formations can
be used to reveal information concerning positional advantages (Reinfeld, 1987).
Furthermore, three of the six criteria for determining the character of a position
given by Capablanca relate to the effectiveness of the pawns in the game (Pachman,
1975).
Pawn formations contain information about the positional nature of the chess
game (Soltis, 1976). The geometric patterns of pawns displayed by a specific
adversary can be used to identify his playing style. By definition we can recognize
open and closed board positions solely from the absence or presence of long pawn
structures. The general playing style of the adversary is indicated by the predominant
style displayed by the adversary in the studied games. Adversaries will tend to
remain with the particular style that they feel most comfortable in playing. Chigorin
played only open style games (Pachman, 1975). Our own analysis of the games of
Botvinnik and Spassky has indicated that each of these chess grandmasters prefers
a particular style, closed and open respectively.
Openings also play an important role in determining the positional advantage
of the game. Current computer chess programs utilize a book or database of
opening moves which have a duration from two moves to more than ten moves. The
evaluation algorithms which enable chess programs to estimate the strength of the

32
current game position are not used until the book is exited. Typically, exits from
opening books occur only when an adversary makes a move that is not contained in
the book or the current opening line of play is exhausted. This means that current
computer chess programs can find themselves from ten to twenty percent of the way
through a game, for a game lasting fifty moves, before they start to calculate the
value of the current position.
The opening sequences of play that are used as the opening book for current
chess programs are stored verbatim. Mednis (1990) has demonstrated that particular
opening formations can be achieved through a wide variety of the actual move order.
A statistical analysis of the opening move sequences displayed by an adversary
accounts for opening move order variations by finding the upper bound, lower bound,
and mean time of specific moves.
Selecting an opening line of play is usually performed at random or in
response to the adversary’s choice of opening move. By studying the playing style
and previous games of an adversary, specific opening lines that are well known to an
adversary are identified. This knowledge can be used by a chess program to emulate
human chess experts by choosing openings that are not as well known to the
adversary. Nunn (Nunn & Griffiths, 1987) avoids the Pelikan variation of the Sicilian
Defense in one of his games against Anthony because Nunn knows from experience
that Anthony is a Pelikan specialist. Other examples of selecting opening lines of
play based on the prior performance of an adversary are given by Nunn.
Acquisition of the opening sequence knowledge held by the adversary will
enable current chess computer programs to select lines of play which are strategically

33
advantageous. Holding (1985) supports our perspective by stating that we should use
information about players playing style to disrupt their normal play.
2.10.3 Philosophy
Dennett (1987) states that current chess programs concentrate their time on
branches of a game tree which represent the adversary’s best response as viewed by
the program. While this would appear to be a reasonable approach, Dennett
elaborates by stating that a rational adversary might have a higher ulterior motive
than avoiding defeat in chess. For example, a young Soviet chess player with a rating
of master is performing well in a European chess tournament. The young Soviet has
acquired a score of 8.5 points and is about to face the tournament champion, a fellow
Soviet grandmaster who has scored 11 points thus far. The grandmaster and young
master both believe that the grandmaster will win the final game. However, no other
player in the tournament has a score greater than 9 points. The Soviet grandmaster
realizes that an outcome of a draw with himself will greatly raise the young Soviet
master’s chess rating without damaging his own rating and might improve the young
master’s confidence and future playing ability. Therefore, the Soviet grandmaster
who has already won the tournament by virtue of his previous victories offers the
younger master-level player a draw midway through the game.
Another example of the problems faced by computer programs in dealing with
the intentionality of human adversaries is provided by Levy (1984) who claims that
he will intentionally choose inferior moves when playing against a computer. The
inferior move choices are selected by Levy because he believes that computer chess

34
programs have greater difficulty in playing a winning game against such an irrational
style of play.
The Levy example above demonstrates one of the traits of human adversaries.
Humans, especially when competing against a machine, attempt to mislead their
adversary about their intentions and capabilities. We must be able to distinguish
between relevant information and misinformation about an adversary’s playing style.
Incorporating the detection of misinformation from an adversary into the inductive
learning mechanism adds an additional heuristic element. The addition of another
heuristic element reduces the probability of acquiring a valid adversary model.
The effect of an adversary’s misinformation attempts is negated by selecting
previous domain performances that limit the possibility of misinformation. By only
using the previous games of an adversary from tournament play in which the
adversary has something at stake, misinformation is eliminated. For our application
in the chess domain we use the World Championship Match games of adversaries
taken from the compendium by Gelo (1988).
2.10.4 Psychology
Chase and Simon (1973) have extended the research performed by de Groot
which claims that the depth of game tree search and the total number of moves
considered is approximately the same for both novices and chess masters. However,
the ability to recall chess positions after a five second presentation of the positions
is markedly different. Chess masters are able to reconstruct chess positions involving
twenty or more pieces with complete accuracy. Chess novices are only able to place

35
four or five of the pieces correctly. The difference between the masters and the
novices disappears when the subjects are asked to reproduce random patterns of
chess pieces. Each group can only place three or four of the pieces from the random
board configurations. The apparent decrease in ability by the master level players
is attributed to the fact that the random configurations of pieces contained no
inherent domain knowledge.
Based on Miller’s (1956) hypothesis, the chess masters are forming higher
level chunks composed of groups of pieces from the game positions. These chunks
contain domain specific knowledge from the prior experience of the chess masters.
Chase and Simon (1988) report that human chess masters store from 10,000 to
100,000 chess patterns.
The skill of chess masters increases as more patterns are acquired. Novice
chess players move away from slow deductive reasoning towards the fast perceptual
processing used by masters as more and more patterns are learned (Chase & Simon,
1973). The chunking theory of learning (Laird et al., 1986) supports this viewpoint
in stating that performance improves via the acquisition of knowledge about patterns
and chunks. Richard Seltzer, the father of fourteen year old chess master Bobby
Seltzer, in a personal communication relates that Bobby’s increase in skill is reflected
by an improved perception of the board position and subsequent position evaluation.
Bratko et al. (1986) state that natural positional moves are closely related to
the chunk structures of a game position. Specific chunks cause chess players to
generate tactical and strategic plans. The chunk displayed in Figure 4 will cause an

36
adversary to immediately consider plans for a back-row mate with either the queen
or one of the rooks. Back-row mate plans result from the knowledge contained in
the chunk concerning the opponent’s king’s limited mobility.
configurations. Simon and Gilmartin (1973) developed MAPP which reconstructs
chess positions by simulating human eye movements recorded in their study. Simon
and Gilmartin’s research shows that humans tend to group or chunk together pieces
which share geometric proximity. We will use geometric proximity as the primary
filter for selecting chess pieces to be formed into chunks. Proximity is one of the
Gestalt principles of organization (Anderson, 1980) along with similarity, good
continuation, and symmetry. We will also utilize the Gestalt principles of similarity
and good continuation to construct chunks. Chase and Simon (1973) found that
proximity and piece color, or similarity, were the two strongest factors affecting chunk
pattern recall by their chess master subjects.
The significant use of patterns of game pieces by chess masters has led us to
make the following three part hypothesis:

37
• Chess masters acquire several thousand patterns of pieces that are
used to evaluate a game position.
• Chess masters will continue to use a specific strategy while winning
games and tournaments.
• Chess masters will tend to reduce the complexity of a game position
by moving board positions that do not contain familiar patterns into
positions that do contain familiar patterns.
We present the general hypothesis for all adversarial domains in Chapter 3. This
version of the hypothesis is the chess domain translation from the general hypothesis
and is presented here to examine the underlying foundations. The first part of our
hypothesis is a direct result from the research of Chase and Simon examined above.
The second part of our hypothesis is supported by the Reinforcement Theory
of psychology and is an extension of the results from various psychological
experiments (e.g., the water jug problem (Anderson, 1980) and the 1023-choice task
(Laird et al., 1986)) which indicate that human subjects will continue to use a learned
strategy for solving a class of problems even when a more efficient strategy would
also solve a specific problem.
While the third part of our hypothesis is intuitively appealing, direct support
comes from the research of Horgan et al. (1989) and Saariluoma (1984) which
demonstrates the use of chunking and analogy to similar positions by chess masters
in solving chess problems. The chunking of several pieces into a known pattern
produces a cognitive economy while evaluating complex board positions.

38
The search complexity of game trees is reduced as a result of the application
of our hypothesis. The strategic and tactical decisions of an adversary can be
accurately predicted because the adversary will tend to recreate familiar patterns to
reduce the cognitive complexity of evaluating the current game position. A practical
example of the use of familiar patterns is demonstrated in the 1978 Budapest
tournament game between Pinter and Bronstein (Roycroft, 1988). This game
resulted in a KNKBB endgame. Chess literature had previously claimed that a
position known as the Kling and Horowitz position (Kb6Nb7) will result in a draw
for the KNKBB endgame. Pinter’s knowledge of the Kling and Horowitz position
led him to consistently maneuver to create this position. Pinter succeeded in forming
the Kling and Horowitz position three different times in three different corners of
the board. However, Pinter’s haste to create the chunk caused him to make inferior
moves nearly thirty percent of the time according to the 250,000,000 position BBN
database (Roycroft, 1988) which exhaustively solves the KNKBB endgame in an
optimal number of moves.
Computer chess programs need to simulate the cognitive economy of chess
masters to continue to improve in performance. Lenat et al. (1979) state that
intelligent systems must perform expectation filtering, using predictions to filter
unsurprising data. Such a filtering process is provided in adversarial domains by the
ability to predict adversary moves in specific situations.
2.10.5 Computer Chess and Artificial Intelligence
The first chess "computer" was the automaton constructed by Torres y
Quevedo circa 1890 (Bell, 1978). Torres y Quevedo’s chess machine was capable of

39
accurately playing certain variations of the KRK, King and Rook versus King,
endgame.
Since Claude Shannon’s (1950) paper on the requirements for a chess playing
program, many artificial intelligence researchers have devoted themselves to the
creation of master level game playing programs. Shannon defines two main
strategies for developing chess playing programs. These strategies are the type A
strategy which generates all possible moves to a certain depth of the game tree and
then uses an evaluation function to choose the best move and the type B strategy
which uses heuristics for suggesting moves without an exhaustive search.
The type A strategy chess programs dominate the type B strategy competitors.
Northwestern University’s type A chess program CHESS4.X (Frey, 1983), were X is
a version number, reigned during the 1970s as the North American computer chess
program champion and was the World Computer Chess program champion from
1977 until 1980 (Kopec & Newborn, 1987). The three type A programs BELLE,
HITECH, and DEEP THOUGHT have each set the high water mark for chess
program playing strength at USCF ratings of 2200, 2400, and 2500 respectively.
These three programs have also been the North American computer chess program
champions from 1985 to the present (Newborn & Kopec, 1989).
The chart in Figure 5 shows the progress in playing strength of chess programs
with respect to search speeds. The bottom axis is logarithmic which means that the
linear progression in strength of chess programs during the 1970s was accomplished
by an exponential increase in search speeds. Recent progress depicted by the dashed


41
curve reflects the hyper-exponential increase in search speeds required to maintain
the linear growth in playing ability. BELLE, HITECH, and DEEP THOUGHT each
use specialized hybrid architectures to achieve the dramatic increases in search speed
depicted in Figure 5 (Ebeling, 1986). HITECH uses a VLSI architecture with fine
grained parallelism to perform pattern matching which is cognitively similar to the
pattern recognition performed by human chess masters. DEEP THOUGHT
(Newborn & Kopec, 1989) uses special VLSI circuitry to achieve a search rate of
720,000 position nodes per second.
Although type A strategy chess programs are the best performing chess
programs, their performance against grandmaster rated human players has been
abysmal. In an exhibition round robin tournament played in October 1989 at Boston,
four human grandmasters crushed four of the top computer chess programs, including
HITECH and DEEP THOUGHT (Hamilton, 1990), with a score of 14.5 to 1.5.
World Chess Champion Gary Kasparov defeated DEEP THOUGHT in a two game
match played in New York prior to the Boston tournament. Kasparov commented
that DEEP THOUGHT lacked experience and was unable to analyze the reasons for
some of the moves he made (Geitner, 1989).
Berliner and Goetsch (1984) state that the performance of chess programs is
proportional to the amount of knowledge used by the program’s evaluation function.
Bratko and Michie (1980) demonstrated the need to use domain specific knowledge
to solve certain domain tasks by having CHESS4.5 and the AL1 program try to solve
the KRKN endgame. CHESS4.5 was unable to solve two of the three chess

42
problems used in Bratko and Michie’s research. However, AL1 which uses a type
B heuristic method with domain knowledge was able to solve all of the endgame
problems.
Type B strategies vary in their approaches to playing chess. The first heuristic
method uses plans or goals that have been constructed from domain specific
expertise to solve chess problems. AL1 is a goal based chess program which solves
KRKN endgames. LEBL (Tadepalli, 1989) uses goals during King-Pawn endgames
to reduce search complexity. Wilkins (1980 and 1982) also uses domain specific
knowledge to produce plans which his program PARADISE tries to achieve. Goal
based planning allows PARADISE to accurately play quiescent middle game
positions. The use of goals to achieve higher playing performance in chess programs
simulates the cognitive method of means-ends analysis used by humans in certain
problem solving tasks (Anderson, 1980).
Another heuristic approach which is similar to the cognitive methods analyzed
by Chase and Simon, described in section 2.10.4, simulates the human cognitive
process of chunking. This heuristic approach tries to identify common configurations
or chunks and then uses these chunks to identify analogous solutions to chess
problems. Quinlan’s ID3 program (in Michalski et al., 1983) tries to inductively
classify equivalent board configurations in the KRKN endgame. The ID3 induced
patterns are used to determine if the King-Rook side has a winning position within
a specified number of moves. Currently the program has solved the KRKN endgame
for all positions which can be won in three ply. Campbell’s (1988) CHUNKER also

43
takes advantage of board configurations to produce chunks which are used to solve
King-Pawn endgames.
Each of the heuristic methods described above have been implemented on
endgame or middle game portions of chess problems and deal primarily with tactical
solutions to chess problems. The knowledge available to chess programs can be
significantly augmented by extending the research of Quinlan and Campbell in the
following ways:
• Collect chunks from complete chess games instead of just the endgame
segment.
• Use the chunks of pieces to predict an adversary’s tactical and strategic
movement decisions.
2.10.6 Justification for Chess Domain Application
Shapiro (1987) presents several reasons for choosing chess as an experimental
test-bench.
1. The game constitutes a fully defined and well-formalized domain.
2. The game is sufficiently complex to challenge the highest levels of
human intellectual capacity.
3. A large body of knowledge has been accumulated and is available in
literature form.
4. A generally accepted numerical scale of performance is available so
that increases in performance can be rated.
Our application domain needs to be sufficiently complex so that exhaustive search

44
is not a feasible solution. Typical chess game trees consist of 1043 nodes (Berliner
& Goetsch, 1984) which prohibits exhaustive search solutions given current
technological capabilities. Because chess is a well understood domain and there is
an accepted numerical rating scale, performance improvements contributed by the
adversary modeling methodology can be evaluated.

CHAPTER 3
ADVERSARY MODELING METHODOLOGY
In this chapter, we examine in detail each of the functions required to create
an adversary model from the adversary modeling methodology. The foundation for
our adversary modeling methodology is the three part hypothesis:
• Relevant patterns and plans are learned by the adversary from
practical experience in the domain.
• The strategic and tactical actions of an adversary are repeated as long
as the actions produce a positive result.
• An adversary tends to reduce the complexity of a situation to enable
a more precise strategic evaluation of the situation by using the plans
and patterns already acquired in the domain.
Evidence supporting this hypothesis and the specific chess domain translation are
presented in section 2.10.4.
To accurately predict the strategic and tactical plans of an adversary, we must
first be able to perceive and evaluate the domain the same as our adversary. This
means that we need to acquire the evaluation criteria used by our adversary. The
first and third parts of our foundational hypothesis indicate that all competitors in
adversarial domains make use of patterns that have been previously acquired in the
domain for their evaluation process.
45

46
3.1 Knowledge Acquisition of Chunks
For chess, the patterns involve both the geometric patterns of pieces and the
textual patterns related to the actual moves that are executed. The geometric
patterns correspond to chunks that are stored in visual memory, while the textual
patterns correspond to chunks that are stored in the adversary’s verbal memory.
3.1.1 Defining the Chunks to be Acquired
We rely on the Gestalt principles of organization and the research of Chase
and Simon described in section 2.10.4 to construct our definition of the elements that
are to be contained in a geometric chunk. Because we are trying to find chunks of
pieces that are used to reduce the complexity of a situation, our chunks necessarily
have a minimum of two pieces. Although chunks need only two pieces to satisfy our
definition, chunks are generally required to contain a minimum of three pieces to be
used in our predictions of probable adversary moves. Chase and Simon (1973)
indicate that the chunks used by chess masters contain three to four pieces. Two
piece chunks are strategically significant in the opening game segment of chess due
to the minimal number of pieces that have moved from their starting positions and
in the end game segment due to the previous elimination of pieces. However, two
piece chunks in middle game positions often occur without effect from the strategy
or cognitive simplification efforts of the adversary. The chunking mechanism
emphasizes chunks which contain a minimum of three pieces.
Based on the principal of proximity, chunks contain only pieces which are
adjacent to another piece. Pieces contained in a chunk are all of the same color to

47
satisfy the principle of similarity. Requiring chunks to be of a single color, namely
the pieces played by the adversary, restricts the contents of the chunk to be under
direct control of the adversary. An adversary cannot control the move choices of an
opponent except for inter-locked pawn chains and therefore cannot use a chunk with
mixed colors of pieces for long range strategic planning.
VResearch performed by Church and Church (Holding, 1985) demonstrates that
human chess players have difficulty in processing information from pieces on
diagonals. From this research we only chunk pieces by proximity which are adjacent
along either the horizontal axis or vertical axis of the playing board. Other
adversarial game domains like checkers can rotate the major axes to include the
diagonals and exclude the horizontal and vertical axes.
The Gestalt principle of good continuation or meaningfulness is used to chunk
pieces which are capable of affecting the strategic meaning of a chunk from the
diagonals. The bishops, queens, and king are all included in a chunk if they are
diagonally adjacent to another piece in the chunk. Meaningfulness is also used to
include pawns which are diagonally behind, to reflect the movement capabilities of
the pawn piece, a piece in a chunk.
Furthermore, we need to define the maximum size of each chunk to build an
efficient knowledge representation structure for storing the chunks. Humans
compose complex chunks hierarchically from smaller, less complex chunks. We use
a maximum chunk size of a four-by-four, or sixteen square, board area. All of the
chunks collected by IAM from a twelve game analysis of M. Botvinnik are shown in

48
Figure 6. From these chunks we can see that a sixteen square area is sufficient to
hold all of the chunks found. We increased the maximum size of defined chunks to
five-by-five and six-by-six to measure the affect of a larger chunk size. Our research
indicated that only one additional chunk was captured during an analysis of twenty
games played by Botvinnik (see Section 4.2.4, Figure 22). This marginal increase in
acquired chunks does not merit the increased cost in space requirements for larger
chunk definitions. The sixteen square size chunk is the minimal implementation that
also permits the use of a chunk to acquire additional knowledge we deemed useful,
including the pieces remaining at the end of each game.
Because of the significance of pawn formations in the analysis of chess
positions, we acquire a second type of geometric chunk which is aimed at acquiring
pawn formations. As noted above, pawn formations are stable and an adversary can
rely on the presence of an opponent’s pawns in a chunk. Our similarity principle is
modified to collect all similar pieces, i.e. pawns, instead of similar colors. The
proximity and meaningfulness principles remain the same for pawn chunks. Because
pawn structures can extend over a large area of the game board, we increase the size
of the maximum defined chunk from a sixteen square area to a thirty-six square area.
Support for increasing the maximum chunk size for the pawn formation chunks was
found when several pawn structures were acquired during our research which would
have exceeded the sixteen square area used in the standard geometric chunk
definition.

49
a
B
a
a
a
a
a
A
B.
c
a
R
R
a
a
a
a
a
R
B
A
&
a
R
a
£
a
a
a
£
É
a
A
a
a
R
a
a
B
A
a
A
a
B
a
a
a
A
£2
R
6
a
a
A
B
a
£
c
R
R
a
a
a
Figure 6: Chunks acquired for Botvinnik.

50
We have emulated the cognitive perceptual processes of human chess players
to define the composition of geometric chunks. Our definition of the elements which
are viable chunk members acquires chunks which are easily perceived and which
contain the maximum amount of knowledge concerning the current game situation.
Our definition of textual chunks is less complicated. Human chess players
remember verbatim dozens of opening lines of play that they have experienced. We
emulate this rote learning by chunking the first five moves made by the adversary
verbatim from the game records of the studied games. Our research indicates that
chess masters tend to leave an opening line of play between the fifth and tenth move
of the game. We are interested in acquiring knowledge that is high in quality and
reliability and therefore we rote learn only the first five moves. One of our
experiments considered the effect of increasing the textual chunk size to ten and
fifteen moves and the results of this experiment are detailed in Chapter 4.
3.1.2 Acquiring Geometric Chunks
From our background study presented in Chapter 2, we note the significance
of perceptual processing in the development of chess expertise. We take advantage
of this aspect of expert chess performance by viewing the game board and pieces as

51
an image to be processed into chunks. By viewing the game board as an image, we
consider via analogy previously defined computer vision techniques for processing
images and collecting information.
The convolution operator is used in the image algebra (Ritter et al., 1988) of
computer vision to collect information from the area surrounding a specific pixel,
which is analogous the proximity principle we use for defining chunks. For games,
a homomorphism can be drawn between pixels in an image and pieces on a game
board. The meaningfulness principle that we use in our chunk definition requires us
to capture knowledge from different areas surrounding a piece, depending on the
current piece being analyzed. The technique of varying templates enables us to alter
the summing template used by the convolution operator to correspond to an
individual piece. This permits us to collect information about the proximal pieces
that is dependent upon the central piece.
We maintain an internal representation of the game board in an eight-by-eight
array with subscripts T and ‘j\ The first subscript denotes the column and the
second subscript denotes the row. Our internal representation is similar to the one
described by Shannon (1950) with negative values assigned to non-adversary pieces
and the numbers one through six representing the pawns, knights, bishops, rooks,
queens, and kings respectively. Unoccupied squares are denoted by the value zero.
A board position and its corresponding internal representation are demonstrated in
Figure 7. The piece value at each board location is denoted by the algebraic symbol

52
j=8
7
6
5
4
3
2
1
i :
Figure 7: Internal representation of a board position.
The three different templates used in our implementation of the convolution
algorithm are shown in Figure 8. Each template can be thought of as a three-by-
three array with the center location denoted by the subscript (0,0). Other template
subscripts range from negative one to one. The general template in Figure 8 is used
for bishops, queens, and the king with the other pieces being convolved by the
appropriately named template. The pseudo-code in Figure 9 demonstrates how the
convolution operation is performed. When an adversary’s piece is detected, the
appropriate template is loaded into a copy of the current template-TP. The values
depicted in the templates were a proximal piece is located are summed and stored
in the return array Y using an inclusive or operator-®. We have defined the
predicate function, ON BOARD, to verify that the array subscripts ‘i + M’ and ‘j + N’
correspond to a valid board location. The return array Y is an eight-by-eight copy
of the internal board representation with all values initially set to zero. This
-4
-2
0
0
0
-4
-6
0
0
-3
-5
-1
-3
0
-1
-1
-1
-1
0
0
-1
0
0
0
0
0
0
0
0
-1
0
0
0
1
1
1
0
0
-2
0
1
0
2
0
0
2
1
0
0
0
0
0
0
1
3
1
4
0
0
5
4
0
6
0
12345678

53
4
8
16
2
/
/
32
1
128
64
4
8
16
2
7~
/
32
0
128
0
General Pawn
Figure 8: Convolution templates.
0
a
0
2
/
/
32
0
128
0
Rook, Knight
Case Xj j of
3, 5, 6 : TP <- the general template ;; (See Figure 8)
1 : TP <- the pawn template
2, 4 : TP <- the rook-knight template
otherwise Return ;; Non-adversary piece, no need to convolve
End Case
For N = -1 to 1 do
For M = -1 to 1 do
If ON_BOARD(i + N, j + M) and X(i+N)(j+M) < > 0
Then Begin
Yy - Yy © TPn m
Y(¡+n),g+m) Y(i+N)i(j+M) © TP.NrM
End Begin
:igure 9: Pseudo-code to perform convolution.
effectively captures all pieces satisfying our chunk definition qualities and returns
their relational position to the center piece location in Y.
The internal representation of the game board is updated after each move
that is read from the game record. For example, if the move ‘al-cl’ is read to move
the rook, in the lower-left corner of the board in Figure 7, two squares to the right,

54
then the value at X31 would be set to four and the value of Xu would be changed
from four to zero. Currently we use a modified algebraic notation for recognizing
the moves played during the game. Other notations can be easily translated to
algebraic notation for use by the IAM program. The chunks we desire to capture are
the ones that have just been formed by the adversary, which improves our capability
for predicting definite adversary actions. Therefore, we collect chunks following the
adversary’s move.
The appropriate step function shown in Figure 10 is applied to the board
representation X and is stored in an eight-by-eight array Y. Pseudo-code which
reduces the complexity of the ensuing board search by removing all pieces that do
not satisfy the appropriate similarity constraint is shown in Figure 11. We use the
step functions to increase the speed of the search algorithm which locates pieces that
can be chunked together. The convolution operation shown in Figure 9 is then
performed at the site of the current board search. If any neighbors are identified as
belonging to the chunk, then the convolution operator is applied iteratively to all
related neighbors. This acquires the chunk as soon as the left-most and bottom-most
piece, with respect to the al square, is found.
Chunks of pieces which have never been moved by the adversary do not
represent any significant new geometric relational or strategic knowledge. To reduce
the total number of chunks being stored, we discard chunks which have greater than
half of their pieces still in their original starting positions. To further reduce the
storage requirements of the adversary modeling methodology we take advantage of

55
For I = 1 to 8 do
For J = 1 to 8 do
If gathering standard geometric chunks AND X, j > 0
Then Yn «- 1
Else If gathering pawn formation chunks AND ABS(X, j) = 1
Then Yu - 1
Else Y, j «- 0
For I = 1 to 8 do
For J = 1 to 8 do
If Y, j = 1 Then perform convolution
Figure 11: Pseudo-code to perform step function.
chunk durations. If a chunk is created on the tenth move of a game and persists
until the twentieth move, then only one chunk is created instead of ten chunks. The

56
times of a chunk’s inception and subsequent dissolution are stored with each chunk.
We also use the time tags associated with each chunk to eliminate storing a chunk
which is dissolved by the adversary and then recreated at a later time during the
same game. Pseudo-code to perform the chunk comparisons, which produce the
storage savings just mentioned, is shown in Figure 12. The variable ‘C represents
the double linked list of chunk records with the subscript T denoting the position in
the list and the variable ‘D’ represents the most recent chunk that has been identified
for the current game.
For I = 1 to the total number of chunks created for this game do
If C, = D
Then Begin
If D.start time < C,.start_time
Then Q.start_time <- D.start_time
If D.end_time > C[.end_time
Then Cj.end_time «- D.end_time
D - 0
Exit the loop
End Begin
Figure 12: Pseudo-code to perform chunk coalescing.
An example of the chunks identified by the convolution operator and the
convolution values returned is shown in Figure 13 where the chunks have been
acquired for the white player. The position shown in Figure 13 is taken from the
fifth game of the 1951 World Championship match between Botvinnik and Bronstein
after white’s eighteenth move. The three chunks acquired by the convolution

57
â– 
:|:|Í;Í-Í;í;»;
m
i
*;*;*!*;*;*;*;
i
mk
i i
II
â– t
1
1
Ü!
A
-J
fc !:Í3Í$Í$!
jrj
*;=;*;*;*;sl*¿
jljlÜ:
11
jijljljljljlj
u
Si
IIH
â– SHIS
n
II
i
&
¡I
&
*j;S*5*5*|*Í;S
ü
R
li
R
L¿*TÉ| ¡',
Dvl
â– 1
1
r
0
33
162
2
16
0
8
0
224
19S
0
120
174
7
â– M
0
32 ;
di
28
r
Blanks - 0
Figure 13: Chunks on a chess board and their convolution values.

58
algorithm are outlined in the figure. The pawn structure in the middle-left of the
board could become a strong factor in the endgame and possibly create a passed
pawn. This knowledge is available from the geometric relationship of the pieces in
the chunk. The other two chunks are a standard offensive configuration (i.e., the
queen and rook) and a standard defensive configuration (i.e., the fianchettoed bishop
and castled king) that are found in many chess games.
The generality of the geometric chunk acquisition technique is demonstrated
by applying the convolution operator to the adversarial domain of the Go game.
Because only one type of piece is used, the general template from Figure 8 is the
only template used in performing the convolution. The results of the convolution
algorithm applied to a section of the Go game board are shown in Figure 14. This
chunk represents the Go formation called eyes. Eyes are an important defensive
strategic feature in Go. Other Go formations such as ladders, knots, false eyes, and
tiger’s mouths are also acquired by the adversary modeling methodology’s
convolution algorithm.
3.1.3 Acquiring Textual Chunks
The background research presented in Chapter 2 indicates that chess masters
make use of their verbal memory to recall specific opening sequences of play. Our
adversary modeling methodology acquires this facet of expert chess performance in
addition to the geometric piece chunks.
Chess masters can recall verbatim particular opening sequences of play, which
indicates the use of rote learning. We emulate the cognitive process of rote learning

59
The heavy black lime
represents the edge
of the game board.
160
35
34
98
130
24
12
Values in squares correspond
to pieces located at game
board intersections
Figure 14: Chunk acquisition in Go.
to acquire textual chunks. Each move from a previously played game is read in
textually, in modified algebraic notation, from the game record used as input. The
algebraic notation is then translated to update the current board representation
maintained by the adversary modeling methodology application. Prior to translating
the textual notation used to record the moves of an adversary’s game, specific types
of moves are saved in the "move" section of the adversary knowledge base verbatim.
Due to the availability of transpositions which lead to the same board position
through several different move sequences, we store each of the first five moves
performed by an adversary individually. A statistical representation of the frequency
that a move occurs and the average game turn is generated and stored with each
move. The statistical representation includes: the number of times a specific move

60
has been observed, the mean move number at which the move has occurred, and the
color of the pieces being played.
From the statistical knowledge of each of the moves displayed by the
adversary, the adversary model predicts opening sequences of play based on the
average over-all performance of an adversary. The knowledge of piece color is
crucial because chess masters use different modes of play for each of the two colors.
As the white player, a chess master can choose which opening sequence to play and
has greater control over the position of the pieces following the opening. However,
as the black player, a chess master plays a responsive mode instead of the controlling
mode of white. The opening moves of black are made as responses to the white
player’s move choices. Therefore, if an identical move is made by an adversary as
both a white piece and black piece player, then the two moves are considered as
different moves and are stored separately.
Other types of moves are useful for predicting the playing style of an
adversary. We save the number of pawn moves that occur during the first ten moves
of each game and the number of attacking moves made during the first ten and
twenty moves of each game to aid in inferring the playing style of an adversary. Any
other move which can be identified textually is also a candidate for the textual move
knowledge of the adversary model. A flowchart for the process of collecting textual-
based knowledge is shown in Figure 15. The frequency and direction of castling
moves are also acquired to assist in prediction of possible castle moves during an
adversary’s future games.

61
To demonstrate the flowchart, assume that the move c4--pawn to queen’s
bishop four--is made by the adversary on his fourth game turn. Since the move
occurs during the first five moves of the game, the move will be saved into the
textual chunk knowledge base. If the move already exists in the textual chunk
knowledge base, then the number of times this move has been observed is
incremented and the statistics regarding the time of the move are updated to reflect
the new knowledge. Because a pawn-the queen’s bishop pawn--is being moved, we
increment the count of pawn moves made during the first ten moves of the game.
We save the total number of games analyzed, the number of games the
adversary played as white, the number of games won by the adversary, the number

62
of games won as white, and the average length of all analyzed games. This
information is used by the adversary model to determine the frequency of a specific
move and the likelihood that the move will be repeated by the adversary.
Other adversarial domains can benefit from the textual knowledge acquisition
performed by the adversary modeling methodology. For example, the acquisition of
textual patterns will capture standard operating procedures and protocols, such as
how to accept an enemy’s surrender, that have been learned by an adversary in a
military domain.
3.2 Induction to Learn Repeated Chunk Patterns
We have acquired the geometric chunks, both standard and pawn structure
chunks, through the computer vision technique of applying convolutions to the board
image. Our goal with the adversary modeling methodology is to learn the chunks
which are used by the adversary to evaluate domain situations. We propose that the
chunks which are commonly used by an adversary can be identified through their
repeated use across several games.
Induction is used to identify groups of chunks from the collection of chunks
acquired from each of the analyzed games. Each group which contains a minimum
of two chunks, indicating the observation of the chunk in two separate games, is
identified as a cognitive chunk used by the adversary in domain evaluations. Chunks
that are found only in single games may occur by chance and are therefore not used
to predict the tactical and strategic decisions of an adversary. Because of our

63
inductive premise, we require that a minimum of two adversary games be studied
prior to performing the induction.
The induction algorithm requires chunks to be identical before being grouped
together. Our induction algorithm is noise intolerant since even small variations in
a chunk can have a large effect on the strategic evaluation value of a chunk. An
example of three chunks which only differ by the location or value of a single piece
and which have significantly different strategic meanings is shown in Figure 16. The
first chunk shows a strong pawn formation, while the second chunk shows two pawns
being attacked by an opponent’s pawn, and the third chunk shows a pawn formation
with a double pawn which may become a liability towards the end of the game. The
noise intolerance of the induction algorithm is not a real problem since most
adversarial domains keep very accurate records of the past performances of domain
entities.
Chunks are considered identical when each has the same number and type of
pieces arranged in the same geometric configurations. The strategic significance of
a chunk is usually invariant with respect to left and right alignment, so chunks are
reflected through a vertical axis of the game board while trying to identify identical

64
chunks. When the adversary is playing the black pieces, we reflect the pieces through
the mid-board horizontal axis to group chunks that occur while the adversary is
playing either color of pieces. Chunks considered to be identical, as shown in Figure
17, are grouped together and stored in the knowledge base that forms the adversary
model.
S*S*í]Í;Í;5;Í;
;:;S;:;S;S;i-S
rTvTv’T
Hi
1|
I*;:;*;* !;!
«M
â–  ill
B
fl
ill i 111
•i-S-S-JlS-t-J
¡1 11
Figure 17: Four identical chunks.
We use a frame based system to store the chunks found by the induction
algorithm in the adversary model’s knowledge base. The slots in each frame are the
actual chunk pieces and relative positions of the induced chunks, the number of times
this chunk has been observed, the earliest move number when the chunk was created,
the latest move number when the chunk was dissolved, the outcome (win or loss) of

65
the games in which the chunk occurred, and the color of the adversary’s pieces.
Although we have not implemented this slot, chunks which are observed frequently
can have a "best move" slot added which would model the scripts used by chess
masters for certain configurations.
The adversary model’s knowledge base of patterns known to an adversary can
be constructed in parts. At least two games need to be processed by the convolution
operator to acquire potential chunks. After the potential chunks have been acquired,
the induction learning mechanism processes the potential chunks of each game by
first checking to see if the chunk already exists in the adversary model’s knowledge
base. If the chunk already exists in the knowledge base then the appropriate slots
are updated to reflect the new information associated with the new potential chunk.
Should the chunk not exist in the current knowledge base, then the chunk is
compared against the chunks of every other game that has just been processed by the
convolution operation. Any identical chunks are grouped and stored in the
knowledge base as a new pattern known to the adversary.
The induction learning mechanism is applied separately to general chunks
containing different pieces of the same color and pawn structure chunks containing
only pawns of both colors. Following the induction procedure, the adversary model
has a knowledge base which identifies all the recurring textual and geometric
patterns from the previous games of a particular adversary. Multiple adversaries can
be modelled by the adversary modeling methodology with each adversary having a
separate knowledge base. The adversary model’s knowledge bases of specific

66
adversaries are used to infer the playing style of an adversary and to predict the
tactical and strategic movement choices of an adversary.
3.3 Acquiring an Adversary’s Playing Style
Knowing the general playing style of an adversary enables chess programs to
select moves that will place the adversary in a disadvantageous strategic position.
The general playing style of an adversary is inferred from the collection of patterns
contained in the adversary model knowledge base. We use a rule-based inference
engine to heuristically determine the playing style of an adversary.
Currently, we infer two different styles of play. The inference engine indicates
that an adversary prefers open positions, closed positions, or has no obvious
preference. The closed position style of play is further used to imply strong
positional play by the adversary. Likewise, an open position playing style implies a
strong tactical player.
The rule base heuristics address the following domain specific factors which
indicate an adversary’s playing style preference:
• The presence and relative size of pawn structures.
• The number of pawn moves made during the opening sequence.
• The presence of pawn chains, or inter-locked pawn structures.
• The number of opening moves that are attacking-move a piece across
the mid-board horizontal axis.
Each rule is evaluated to determine if a specific playing style is demonstrated. The

67
rules are not competitive, but instead they lend support to the adversary model’s
belief that a particular playing style is preferred by the adversary by adjusting the
likelihood value attached to each playing style. The playing style with the greatest
likelihood is then inferred as the general playing style of the adversary.
Paun Moves Pawn Structures of 5* Pawns
Figure 18: Fuzzy logic used by the heuristic rules.
The heuristic rules use fuzzy logic during the evaluation process. Graphs of
the fuzzy values for each of the heuristic rules is shown in Figure 18, with the fuzzy
regions shaded. Heuristic rules that have a value in the fuzzy areas which lie close
to the division between the two styles of play currently inferred do not support either
hypothesis. Our decision to use fuzzy logic prevents us from assigning a playing style
to an adversary that is based on statistically insignificant data with respect to the
fuzzy set values.

68
The heuristic rules used to infer an adversary’s playing style are not
competitive, new heuristic rules can be added at any location of the rule base. As
new domain specific knowledge becomes available that will infer additional playing
styles or support the current playing style inferences, the current rule base can be
easily augmented.
The determination of an adversary’s playing style preference has been added
to the generalized adversary modeling methodology to increase the adversary model’s
performance in the chess domain. The playing style inference engine makes use of
domain specific knowledge in addition to the textual and geometric patterns and for
this reason is not considered as a main part of the adversary modeling methodology.
The pattern acquisition, induction learning mechanism, and adversary action
prediction tool each use only minimal amounts of domain specific knowledge so the
adversary model implementation can be used in other adversarial domains with little
or no modifications.
3.4 Application of the Adversary Model
The adversary model consists of a knowledge base of textual and geometric,
or visual, patterns known to the adversary. The third part of our hypothesis
presented at the beginning of this chapter indicates how we are going to use this
knowledge. An adversary will attempt to reduce the complexity of evaluating the
current domain situation by maneuvering the board position to contain one or more
of the pattern chunks with which he is familiar.

69
The adversary model also contains two processes which utilize the knowledge
about patterns known by an adversary. The first process detects analogous situations
in the domain and the second process attempts to predict the most probable
adversary action. We define an analogous domain situation to exist when a chunk
in the knowledge base is almost complete in the domain. The second process then
predicts the necessary move to complete the chunk.
Currently a move window size of one is used when detecting analogous
domain situations. For chess, this means that the chunk must be able to have all of
the pieces of the chunk in their correct relational positions in exactly one move.
Board positions that are considered to be analogous to chunks in the knowledge base
may only have at most two pieces out of position, but are normally required to only
have one piece out of position. An extension to the move window size has been
implemented to permit the adversary model to eliminate interfering pieces when a
castle is the required move to complete a chunk.
We detect the analogous situations we have defined for the chess domain by
performing pattern recognition on the game board. The board representation of the
adversary’s game currently in progress is searched prior to each move to be made by
the adversary until a piece belonging to the adversary is located. Next all of the
chunks in the adversary model’s knowledge base which can fit into the current board
location are compared against the current board pattern. Inconsistencies between
the chunks in the knowledge base and the current board configuration are counted.
An inconsistency exists if a board location corresponding to the location of a piece

70
in a knowledge base chunk is empty or contains a different piece. If only one
inconsistency exists, or two if the missing pieces are the king and rook, then the
analogous situation detection process checks to see if any viable move can complete
the chunk.
Chunks in the knowledge base occupy a square area of the game board
defined by the pieces farthest to the left, top, right, and bottom of the chunk. The
square areas of chunks frequently contain blank or non-occupied squares. These
squares serve as wild cards so that the actual square on the game board may be
blank or contain any other piece. This is done to simulate the effect of hierarchically
composing larger chunks from smaller chunks.
Each chunk in the knowledge base which can be completed suggests the
appropriate move to complete the chunk as the adversary’s next move choice.
Several chunks can each have analogous situations for any particular turn of the
current game. The collection of possible moves that enable an adversary to complete
a chunk are evaluated by the adversary move prediction process to select the most
likely adversary move.
The adversary move prediction process uses a heuristic rule-based inference
engine to evaluate the collection of suggested moves that will complete chunks. The
inference engine is similar to the one described above for analyzing the playing style
of an adversary. An initial probability value is generated for each move by dividing
the number of times the chunk suggesting the move has been observed by the total
number of prior adversary games analyzed. The heuristic rules then add or subtract

71
from the base probability of each suggested move. The move with the highest
probability is then used to predict the adversary’s next move. If two or more moves
have nearly equal probabilities, then the inference engine makes multiple move
predictions.
Computer chess programs can use the probability value associated with each
move as a measure of belief. When the adversary model infers a probability for a
move that is less than some specified value, the chess program can choose to ignore
the adversary model’s prediction.
The heuristics used in the prediction inference engine attempt to follow
several psychological principles of cognitive organization and economy. The general
description of the purpose of our heuristics follows.
• Large chunks are favored over small chunks. Chunk size is dependent on the
number of pieces contained in a chunk.
• Chunks containing major pieces are favored over chunks containing only
pawns.
• Favor moves which have been suggested by more than one chunk.
• Reduce the probability of chunks that have only been observed in lost games.
• Increase the probability of chunks whose move suggestion causes a gain in
material advantage. (This heuristic is only used in quiescent positions due to
the volatile nature of material advantage in non-quiescent positions.)
• Chunks which have occurred while using the same color pieces as the current
game are favored over chunks that have only been observed for the opposite
color.

72
• Adjust the probability to account for temporal discrepancies. If the current
game move number is outside of the move range defined by the two slots for
the chunk’s time of inception and dissolution, then reduce the probability
proportional to the distance between the current move and the chunk’s
temporal range.
• Eliminate smaller chunks that attempt to borrow pieces from an existing
larger chunk.
• Reduce the probability of chunks containing only two pieces during the
middle game segment.
• If the adversary has just dissolved a chunk, then do not recreate the chunk
immediately.
• Reduce the probability of chunks that have a maximum dissolution game turn
that is within the time range covered by the textual opening move knowledge
base.
These heuristics attempt to find the move suggestion which has the greatest number
of situational factors in common with the current game condition, such as the color
of the pieces and the time or game turn within the game. Additionally, the heuristics
promote the moves that will create the largest possible chunk and therefore afford
the greatest cognitive economy to the chess player. The heuristic which supports
moves suggested by multiple chunks is simulating the construction of hierarchically
complex chunks from smaller chunks. Our general description of the heuristic rules
indicates a minimal usage of domain specific knowledge which augments our
inter-domain application capabilities.

73
The role of the adversary model as a coach needs to be remembered.
Predictions of probable adversary actions are only made when the current game
situation resembles domain situations which have already been learned inductively.
After analyzing nine and then twelve games of Botvinnik the resultant adversary
models were used to predict the actions of Botvinnik in a new game which had not
been previously analyzed. For each of models’ predictions, over forty percent of the
new game’s board configurations had no similarities to the chunks in the knowledge
base.
Geometric chunk patterns are uncommon during the opening game segment
of a chess game. The textual knowledge base of patterns known to the adversary is
used to supplement the lack of knowledge in the geometric chunk pattern knowledge
base. When a computer chess program equipped with the adversary model is playing
the white pieces, the statistical analysis of opening moves displayed by the adversary
is used to select an opening sequence of play which is unfamiliar to the adversary.
This grants a strategic advantage to the computer chess program with the first move
of the game. Predictions for the opening moves of the adversary are made by
selecting the move with the highest probability value. Probability values are assigned
by using the statistical values stored with each opening move displayed by the
adversary to attain a base probability corresponding to the frequency each move has
been observed. This base probability is then modified to account for any difference
between the current time of the game and the mean game turn of the knowledge
base textual move pattern.

74
Moves which have already been made in the current game cause the identical
move in the knowledge base to be rejected. Additionally, when the adversary is
playing the black pieces, we implemented "response code" in the prediction process
which heuristically favors textual move patterns that have previously followed the
current move just made by the white player. The effect of the response code is
examined in Chapter 4.

CHAPTER 4
IAM
In this chapter we present a detailed examination of our implementation of
our adversary modeling methodology for the domain of chess. Our program is
named IAM, an acronym for Inductive Adversary Modeler. We first present several
detailed examples that demonstrate IAM’s functionality. Next we review the
effectiveness of IAM by examining the predictive capabilities of our adversary model
while performing against an actual opponent.
4.1 Overview of IAM’s Execution
IAM is defined in two stages. The first stage performs the knowledge
acquisition of chunks and the learning by induction phases of the adversary modeling
methodology which establish the knowledge base of geometric and textual chunks
contained in the adversary model. The second stage applies the knowledge in the
knowledge base to a game in progress to predict probable adversary moves and to
identify the adversary’s playing style.
IAM has several global data structures which facilitate the algorithmic design
of the program. Chunks, both textual and geometric, and the suggested moves of the
adversary model are stored in dynamic data structures. The form of these structures
and the pointers to them are among the global data structures. Additionally,
75

76
application constants such as the size of geometric chunks and pawn structure chunks,
the number of opening moves to acquire, the internal representation of the chess
board, and various constants used by the heuristics of the two inference engines are
stored as global variables. The use of these global structures allows us to easily
change the amount, size, or type of knowledge that is to be acquired and provides
a means for fine tuning the heuristics used in predicting an adversary’s move choices.
4,1.1 Knowledge Acquisition and Inductive Learning Stage
Input to IAM, for the knowledge acquisition and learning stage, consists of
two or three game records of the adversary’s previous performances. The records are
written in algebraic notation as shown for the sample game in Figure 19. The first
line of the game record indicates the color of pieces played by the adversary, the
length of the game, and the game result, either win or loss, for the adversary. The
remaining lines are the actual moves of a particular game.
We use only the previous games of an adversary which have resulted in a win
or loss. Drawn games are often influenced by factors, like the state of health of the
adversary, that do not impart any knowledge about the strategic tendencies or
evaluation criteria of an adversary. A good example of this comes from the 1957
World Chess Championship match between Botvinnik and Smyslov. Three of the last
four games in the match ended in a draw with an average game length of thirteen
turns. Botvinnik was down by three points in this match and the first draw was
probably offered by Botvinnik so that he could rest and prepare for the next game.
The final two draws were offered and accepted because the result of the match was

77
Botvinnik vs. Keres (Moscow, May 4/5, 1948)
1 60 1
1.
d4
d5
21.
Nb4
Rd8
41.
Kd3
Ne6
2.
Nf3
Bf5
22.
Qf5
Rd6
42.
Nd5
Kc6
3.
c4
e6
23.
Rfcl
Rxc6
43.
h4
Nd8
4.
cxd5
exd5
24.
Rxc6
Rd8
44.
Nf4
Kd6
5.
Qb3
Nc6
25.
Rxb6
cxb6
45.
Nh5
Ne6
6.
Bg5
Be7
26.
Nc6
Qc7
46.
Ke3
Ke7
7.
Bxe7
Ngxe7
27.
Nxd8
Qxd8
47.
d5
Nc5
8.
e3
Qd6
28.
Qc2
Qc7
48.
Nxg7
Kd6
9.
Nbd2
0-0
29.
Qxc7
Nxc7
49.
Ne6
Nd7
10.
Rcl
a5
30.
Nbl
Kf8
50.
Kd4
Ne5
11.
a3
Rfc8
31.
Kfl
Ke7
51.
Ng7
Nc4
12.
Bd3
a4
32.
Ke2
Kd6
52.
Nf5 +
Kc7
13.
Qc2
Bxd3
33.
Kd3
Kc6
53.
Kc3
Kd7
14.
Qxd3
Nd8
34.
Nc3
Ne8
54.
g4
Ne5
15.
0-0
Ne6
35.
Na2
f6
55.
g5
fxg5
16.
Rc3
b5
36.
f3
Nc7
56.
hxg5
Nf3
17.
Qc2
Rcb8
37.
Nb4 +
Kd6
57.
Kb4
Nxg5
18.
Nel
Nc8
38.
e4
dxe4 +
58.
e5
h5
19.
Rc6
Qe7
39.
fxe4
Ne6
59.
e6 +
Kd8
20.
Nd3
Nb6
40.
Ke3
Nc7
60.
Kxb5
1:0
Figure 19: An example of a game record--the 1948 Hague-Moscow Tournament.
a forgone conclusion, similar to our example in Chapter 2 of the young Soviet chess
master and the older Soviet grandmaster.
Additional knowledge can be inferred from the win/loss result of each game
played by an adversary. When the induction procedure detects a series of lost games
by the adversary, then the possibility that the adversary will be altering his
established strategy is inferred. The adversary move prediction inference engine
decreases the probability of chunks observed only in lost games from predicting the
next adversary move. Likewise, an overall winning performance by an adversary
reinforces the likelihood of chunks being repeated in future games.

78
Each move of every game is read in as input separately. The textual chunking
mechanism then identifies if the current move has any of the qualifications for being
saved in the textual portion of the knowledge base. If so, a dynamic data structure
for holding textual moves is created and the move is stored.
Next, the textual move is translated and the internal board representation is
updated. After each move made by the adversary, potential chunks for each game
are identified by the convolution operator we discussed in Chapter 3. Chunks which
are identical to another chunk in the same game are coalesced into a single chunk
(see the pseudo-code in Figure 12). This step, for the twelve games played by
Botvinnik in the 1948 Hague-Moscow Championship tournament, reduced the 699
chunks identified by the convolution operator to 272 potential chunks that the
inductive learning algorithm must process. Chunks from different games are
considered unique at this point in processing.
Pawn structure chunks are collected by the convolution operator separately
from the standard geometric piece chunks. Because the occurrence of identical pawn
structures during different segments of a game yields knowledge about an adversary’s
playing style, pawn structure chunks are not coalesced.
After the individual geometric chunks have been collected for each game, the
induction algorithm processes the chunks to find patterns which are duplicated across
game boundaries. The induction algorithm found twenty-eight unique chunks, shown
in Figure 6, from the 272 potential chunks identified for Botvinnik in the
Hague-Moscow tournament. Each of the twenty-eight chunks induced was found in

79
at least two games. For example, the partial board representations for the three
games which produced one of the twenty-eight chunks are shown in Figure 20.
Bcrtvirmik
vs.
Reshevskij
â–¡
i-i-i-i-i-i-i-i-
ill
mm
IB!|
n
¡I
—1
III
mm*
ijijipip!
HI
lili
i
I
i
nnwiwi
Hiiiiiiii
i
mm
m
1
siiiiiji
iiiiüiiüiiü!
i
m
mm
Keres
vs.
Botvirmik
i
â– 
9
Ü1
iliiii
jijiijjijjjiji
a
iiiijjjijjlj:
c?
iijiiliij
Botvimiik
III
mm
11
I
1
vs.
Euue
Pi
1
i
R
!|!|!
Figure 20: Chunk occurrence in three different games.

80
The induction algorithm, with pseudo-code shown in Figure 21, verifies if a
potential or induced chunk already exists in the adversary pattern knowledge base
and updates the knowledge base appropriately. When a chunk is found to be
repeated in multiple games, we use the function REMOVE to eliminate multiple
instances of the chunk from the list of potential chunks produced from our
convolution algorithm. The induced chunks which do not exist in the current
knowledge base are then written into new frames, created by the function NEW, in
the knowledge base by the function WRITE.
The potential chunks collected by the convolution operator which are not
grouped into a valid chunk to be saved in the knowledge base are written to a
temporary file. This temporary file serves as a database of non-repeating chunks
which is used to identify chunk repetitions that occur outside of the current games
being induced. All chunks in this temporary database are used by the induction
algorithm to group chunks that have not been previously grouped. The database is
deleted after all games from a specific time period or tournament have been
analyzed. The deletion of the database forces the induced chunks to have a temporal
relevance, or recency, to the potential chunks being analyzed.
The textual move chunks are then analyzed to gather the statistics to be used
by the prediction stage of IAM. The statistics that are gathered include the mean
game turn that a move was executed, the number of times a particular move was
observed, the results of the games in which the move was observed, and the color of
the pieces being played. This statistical analysis requires that only one entry be made

81
For I = 1 to number of chunks found for the current games do
Begin
Flag_l <- FALSE ;; indicates when a chunk has been repeated
Flag_2 <- FALSE ;; indicates when a chunk is duplicated in the KB
For J = 1+1 to number of chunks found for the current games do
;; First check this chunk against the chunks that were just found
if q = q
Then Begin
Increment number of times q has been observed
If q.start_time > q.starttime
Then q.start_time «- q.start_time
If q.end_time < q.end_time
Then q.endtime <- q.end time
REMOVE(q)
Flag l - TRUE
End Begin
For J = 1 to number of chunks currently in the chunk Knowledge Base do
;; Next check this chunk against all chunks already in the Knowledge Base
if q = q
Then Begin ;; Update KB knowledge
Increment number of times Q has been observed
If q.start time < q.start time
Then q.start_time «- q.start_time
If q.end time > q.end time
Then q.end_time «- q.end_time
Flag_2 - TRUE
End Begin
If Flag l = TRUE AND Flag_2 = FALSE
Then Begin
NEW(geometric_frame, F)
F-q
WRITE(F, Geometric Chunk Knowledge Base)
End Begin
End Begin
Figure 21: Pseudo-code to perform induction on geometric chunks.

82
in the knowledge base for each specific move. Additional information concerning the
total number of games analyzed, number of games played as white, number of games
won, and number of games won as white is stored with the textual move patterns.
The pawn structure chunks are also analyzed statistically at the same time as
the textual move patterns. The statistical analysis gathers knowledge that is used by
the adversary playing style inference engine. This knowledge is about the presence,
quantity, and quality of pawn structures and inter-locked pawn chains. The quality
of a pawn structure is determined by the number of pawns involved in the structure.
The knowledge from this analysis is stored with the textual moves in the adversary
patterns knowledge base.
Following the induction on geometric chunks and the statistical analysis of
pawn structures and rote learned textual chunks, the knowledge base of patterns
known by an adversary is available for use by the prediction stage of the adversary
model. The adversary pattern knowledge base is constructed incrementally. As new
games of an adversary become available the chunk knowledge from those games can
be quickly incorporated into the existing knowledge base.
4.1.2 Predicting Adversary Actions Stage
After an adversary patterns knowledge base has been constructed, the
adversary model then attempts to predict adversary actions in relevant domain
situations. Simulations of a game played by an adversary are used to verify the
prediction capabilities of the adversary modeling methodology. Each of the
simulated games used to represent a future contest against an adversary has never
been analyzed by IAM.

83
Prior to each of these simulated games, the preferred playing style of the
adversary is inferred from the chunks in the knowledge base. We had IAM study
twelve games of two different chess grandmasters who display preferences for the two
different playing styles currently inferred by IAM. The two players were Botvinnik,
who prefers closed positions, and Spassky, who prefers open positions. By analyzing
the presence and frequency of pawn structures and the frequency of pawn and
attacking moves, IAM correctly identified the playing style preference of each
adversary. Knowledge about the playing style of an adversary is then used by the
current domain program which the adversary model is coaching to select specific
moves that manipulate the game board away from the adversary’s preferred style.
During the opening portion of the simulated game, IAM relies exclusively on
the textual move knowledge base which contains the statistical analysis of opening
patterns displayed by the adversary. Prior to the adversary’s move, IAM predicts the
adversary’s move from the moves currently in the knowledge base. Following the
actual move made by the adversary, the corresponding move in the knowledge base
is eliminated from further consideration. For an adversary playing the white pieces,
IAM has had accurate predictions ranging from forty percent to one hundred percent,
with a mean prediction accuracy of seventy-five percent.
IAM’s opening sequence predictions are less accurate for an adversary playing
the black pieces. This is because the adversary is playing responsively to his
opponent’s moves. Accuracy for opening move predictions against a black piece
playing adversary ranged from zero to twenty percent.

84
By modifying the textual chunk learning algorithm to remember the white
player’s move which preceded an adversary’s black move, we were able to
significantly increase IAM’s opening sequence prediction accuracy to range between
forty and one hundred percent, with a mean of sixty percent. This modification
produces results which are similar to IAM’s performance against the adversary while
playing the white pieces.
The number of textual chunks stored in the knowledge base increases
proportionately, since identical moves by the adversary which follow a different white
move are now considered as unique moves. The average increase in the knowledge
base size is forty-five percent. For example, after analyzing twenty-two games of
Botvinnik with the original textual chunking algorithm, thirty-one opening move
textual chunks were produced. With the response code modification forty-seven
opening move textual chunks were saved to the knowledge base. The actual textual
chunks are also larger to contain the adversary’s opponent’s move. Because of the
relatively small number of chunks in the knowledge base for a specific opponent, the
increase in prediction accuracy warrants the small increase in size to the knowledge
base.
All move predictions following the opening sequence are made primarily from
the contents of the geometric piece chunk knowledge base. Before each move to be
made by the adversary, the current board representation is searched to find pieces
belonging to the adversary. This is done because each of the chunks contained in the
knowledge base consists of pieces belonging to the adversary.

85
Once an adversary’s piece has been located on the board representation, each
chunk in the knowledge base is compared against the board representation to find
identical matches and analogous chunks, which have one or two pieces out of place.
Because chunks occupy a specific square area of the game board, chunks which are
to large to fit into the current board location are ignored to reduce processing time.
The detection of analogous board positions is performed via pattern matching. When
an analogous board position has been identified, IAM then checks to see if a legal
chess move would complete the analogous chunk. An example of this process is
shown in Figure 22. Four of the chunks shown in Figure 6 were used to identify
analogous positions on the current game board which is prior to Botvinnik’s
thirteenth move from the game displayed in Figure 19. The corresponding moves
that would produce the chunks are also displayed in the figure.
The collection of moves suggested by the analogous chunks is then processed
by an inference engine to heuristically determine the most probable move. The
inference engine predicted the castling move for the Figure 22 board position. This
prediction represents the result of applying the various psychologically based
heuristics concerning chunk size and relevance to the current domain. The queen
being en prise is domain specific information and is therefore not considered by the
move prediction inference engine. However, the actual move made by Botvinnik,
Qc2 or b3-c2, is one of the moves suggested by an analogous chunk and reviewed by
the inference engine. An experiment to augment the adversary pattern knowledge
base with a minimal amount of domain specific knowledge is presented in section 4.2.

86
The process of searching the current board configuration to identify analogous
positions is repeated before each move to be made by the adversary. Existing chess
programs can use IAM’s predictive capabilities by having IAM search the possible
board configuration at a specific node of the game tree and finding analogous chunk
situations in the proposed board configuration. IAM will then predict the most
probable adversary move that corresponds to our domain simplification hypothesis
of Chapter 3.
4.1.3 Integrating IAM’s Knowledge into a Chess Program
Our induction based adversary modeling methodology produces knowledge
which can be incorporated into current chess programs by two methods. Knowledge

87
about the specific opening sequences preferred by an adversary can be used to
manually optimize a chess program’s opening book. Optimization of the opening
book entails augmenting the opening book with opening sequences that are unknown
to the adversary and adjusting the preferences of the chess program to select these
new opening sequences.
Current chess programs use a move generator to limit the breadth of their
game trees. Move generators suggest a number of moves, typically five, which are
considered as the only moves available from a specific position. We describe the
reduction of tree complexity by the adversary modeling methodology in Chapter 5.
IAM predicts the move that an adversary will execute next from a particular board
position. These predictions can be used by chess programs to further limit the search
of their game trees.
A chess program would have the adversary model make predictions of the
next move in the game tree while simultaneously generating the five best moves with
its move generator. When predictions from the adversary model are available, the
chess program would use the predictions to expand its current game tree. Otherwise,
the chess program would continue to use the moves supplied by its move generator.
Combinations of adversary model predicted moves and move generator suggested
moves can be used by heuristically ordering the moves suggested by both methods
and expanding the game tree with the first five moves from the ordered move list.

88
4.2 IAM’s Performance Results
In this section we describe several experiments that have been performed with
IAM to test the overall capabilities of the adversary modeling methodology. Each
experiment is described preceding the analysis of the results.
The games of the adversary Botvinnik are all taken from his World
Championship matches from the 1948 Hague-Moscow tournament through his 1961
match against Tal. The two games which are used to simulate Botvinnik’s
performance in a future match, come from Botvinnik’s 1963 match against Petrosian,
and are shown in Figure 23 and Figure 24. The game TEST1 is the first game from
the Petrosian match resulting in a win or loss in which Botvinnik played the white
pieces. TEST2 is the next game from the match, following TEST1, in which
Botvinnik played the black pieces.
4.2.1 General Performance Results
The purpose of our first experiment is to determine the base performance
results of IAM so that we can compare these results against other experimental
results. We anticipate that the predictive capabilities of IAM will increase as more
chunks are learned. The records of every game played by Botvinnik from his World
Championship Matches, prior to 1963, are used as input.
4.2.1.1 Performance versus the adversary as the white player
TEST1 is simulated for IAM following the construction of the adversary
model’s knowledge base for each of the tournaments of Botvinnik’s career. The

89
length in game turns of TEST1 is fifty-seven moves. The performance of I AM in this
experiment is shown in Table 1.
TEST1-
Botvinnik
playing White
1.
d4
d5
21.
Re2
Nb6
41.
Nc5
Bf5
2.
c4
e6
22.
Rhel
Nc4
42.
Kg3
a4
3.
Nc3
Be7
23.
Bxc4
Rxc4
43.
Kf4
a3
4.
cxd5
exd5
24.
Rd2
Re8
44.
Ke5
Rb4
5.
Bf4
c6
25.
Re3
a6
45.
Nd3
Rb5
6.
e3
Bf5
26.
b3
Rc6
46.
Kd6
Kf7
7.
g4
Be6
27.
Na4
b6
47.
Kc6
Bxd3
8.
h3
Nf6
28.
Nb2
a5
48.
Rxd3
Rb2
9.
Bd3
c5
29.
Nd3
f6
49.
Rxa3
Rg2
10.
NO
Nc6
30.
h4
Bf7
50.
Kxd5
Rxg5 +
11.
Kfl
0-0
31.
Rxe8 +
Be6
51.
Kc6
h5
12.
Kg2
cxd4
32.
Qe3
Bf7
52.
d5
Rg2
13.
Nxd4
Nxd4
33.
g5
Be6
53.
d6
Rc2 +
14.
exd4
Nd7
34.
Nf4
Bf7
54.
Kd7
h4
15.
Qc2
Nf6
35.
Nd3
Be6
55.
f4
Rf2
16.
f3
Rc8
36.
gxf6
Qxf6
56.
Kc8
Rxf4
17.
Be5
Bd6
37.
Qg5
Qxg5 +
57.
Ra7 +
18.
Rael
Bxe5
38.
hxg5
a4
19.
Rxe5
g6
39.
bxa4
Rc4
1:0
20.
Qf2
Nd7
40.
a5
bxa5
Figure 23: A game from the 1963 Botvinnik versus Petrosian match.
The first column, #G, is the number of games currently analyzed to produce
the adversary pattern knowledge base. The #Ch and #M columns are the number
of geometric chunks and textual opening move chunks respectively. The first number
in the #M column represents the total number of chunks and the second number is
the chunks pertaining to white moves, for TEST1, or black moves, for TEST2. The
next three columns are the number of predictions made by IAM (#P), the number
of predictions which exactly matched the adversary’s ensuing move (#C), and the

90
TEST2-
Botvinnik
playing Black.
1.
d4
Nf6
21.
Rhgl
Kh7
41.
Ral
Kg7
2.
c4
g6
22.
Nb5
Rf7
42.
Ra6
Rb7
3.
Nc3
d5
23.
Nd4
Re8
43.
Ra8
Kf6
4.
Qb3
dxc4
24.
NO
Bh6
44.
Rc8
Ne5
5.
Qxc4
Bg7
25.
Ng5 +
Bxg5
45.
Ke3
Nd7
6.
e4
0-0
26.
Rxg5
Nc4
46.
Rc6 +
Kf7
7.
Be2
Nc6
27.
Rdgl
Rg8
47.
e5
Nf8
8.
Nf3
Nd7
28.
Kc2
b6
48.
Rf6 +
Kg7
9.
Be3
Nb6
29.
b3
Nd6
49.
Ke4
b5
10.
Qc5
Bg4
30.
f3
Rd7
50.
Rc6
Kf7
11.
d5
Nd7
31.
R5g2
Rdd8
51.
Rxc5
Ne6
12.
Qa3
Bxf3
32.
a4
Nf7
52.
Rd5
Ke7
13.
Bxf3
Nd4
33.
Bel
e5
53.
Be3
Rb8
14.
0-0-0
Nxf3
34.
Be3
exf4
54.
Rd6
b4
15.
gxf3
Nb6
35.
Bxf4
Rd7
55.
Ra6
Rb5
16.
Qb3
Qd7
36.
Rd2
Rxd2 +
56.
Ra7 +
Ke8
17.
h4
h5
37.
Kxd2
Rd8 +
57.
f4
Kf8
18.
f4
e6
38.
Ke2
c5
58.
f5
19.
dxe6
Qxe6
39.
a5
Rd7
20.
Qxe6
fxe6
40.
axb6
axb6
1:0
Figure 24: Another game from the 1963 Botvinnik versus Petrosian match.
number of predictions which identified the correct piece to move, but selected the
wrong square on the board as the destination (CPWL). The final three columns
reveal the statistical measures for the number of analogous positions found (%G),
the accuracy of the predictions that were made (%C), and the percentage of
predictions identifying the correct piece to be moved (%C+P). The percentage of
correct predictions for the entire game can be obtained by multiplying the
appropriate column (%C or %C+P) by the percentage of analogous positions. We
will use the percentage of correct predictions and correct piece identifications for the
entire game in our analysis of IAM’s performance.

91
Table 1: I AM Performance Measures for TEST1
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C + P
12
28
24/10
31
5
8
54.3
16.1
41.9
22
53
31/14
47
6
7
82.4
12.7
27.6
36
66
38/16
46
6
9
80.7
13.0
32.6
45
68
39/16
38
5
3
66.7
13.1
21.0
80
86
52/22
36
6
4
63.1
16.6
27.8
The games analyzed by IAM in Table 1 are for the adversary Botvinnik and
are from the following tournaments:
• #G = 12, games are from the 1948 Hague-Moscow Tournament.
• 22, new games are from the 1951 Bronstein Match.
• 36, new games are from the 1954 Smyslov Match.
• 45, new games are from the 1957 Smyslov Match.
• 80, new games are from the 1958 Smyslov Match and the 1960 and
1961 Tal Matches.
The sixth and ninth columns of Table 1 are significant in demonstrating that the
adversary model is focusing on the correct area of the game board only using
knowledge of the pattern chunks known by an adversary.
IAM’s analysis of the first twelve games played by Botvinnik produces correct
predictions of adversary moves for nine percent of the total moves and correctly

92
identifies the piece to be moved in twenty-three percent of the total moves in the
game. The size of the geometric chunk knowledge base nearly doubles with IAM’s
analysis of the next ten games played by Botvinnik. We can see from the second row
of the table that the correct piece is still identified for twenty-three percent of the
total moves. However, the percentage of correct predictions increases to almost
eleven percent of the total game moves. The new chunks enable the adversary
model to convert one of the correct piece identifications into a correct prediction.
IAM’s induction based learning technique continues to improve the
performance of the adversary model following the next fourteen games that were
analyzed. The third row of Table 1 shows the correct predictions remaining constant,
but the percentage of correct piece identifications is now over twenty-six percent of
the total moves.
We also notice that the number of predictions or analogous positions which
showed a fifty percent increase between the first and second rows of the table starts
to drop in the third row. This curtailment in the quantity of predictions made by the
adversary model continues throughout the rest of the table. Our inference engine
tries to maintain existing chunks on the game board. The reduction in analogous
positions is due to the newest chunks in the knowledge base already existing in the
location of an earlier chunk that was only partially completed. Because the new
chunk exists in its entirety, the analogy which produced the previous chunk’s
predictions is discarded by the inference engine making the predictions.

93
The last two rows of Table 1 demonstrate an interesting problem revealed in
our experiments. First the correct prediction percentage fluctuates back to nine
percent before regaining its optimal eleven percent rate in the last row. The
percentage of correct piece identifications for both rows drops below twenty percent.
The cause of this apparent decline in performance is similar to the cause for the
reduced prediction rates described above. The presence of newly learned chunks on
the game board prohibits the suggestion of moves for completing other chunks that
require borrowing a piece from the existing chunk. Competition between the chunks
for the pieces on the game board reduces the total number of predictions made by
the adversary model and likewise the number of correct piece identifications.
To understand chunk competition, we must first realize that once a chunk is
in the knowledge base it will always identify the same board configurations of a
particular game as being analogous. The chunk will then suggest the appropriate
move or moves to maneuver the missing piece into the correct location to complete
the chunk. Pseudo-code which suggests the appropriate move to complete a chunk
is shown in Figure 25. As new chunks are learned and added to the knowledge base,
the likelihood that the missing piece required to complete the earlier chunk is
contained in a new chunk increases. If a new chunk which also contains the missing
piece of the earlier chunk exists on the board and owns the specific piece required
to complete the earlier chunk, then the earlier chunk’s move suggestion is rejected
by the inference engine due to the existing chunk.

94
The pseudo-code in Figure 25 requires that the board position (i, j)
correspond to the upper-leftmost square of the defined chunk. First we determine
the piece that is required to complete the chunk and then we search the board
representation to locate the required piece. If the piece can be moved from its
current square to the square which completes the chunk without violating the
movement rules of chess, then the corresponding move is suggested. An individual
If number of pieces needed to complete the chunk C > 1
Then see if either a king-side or queen-side castle will complete C
Else Begin
For A = 1 to the width of C do ;; the chunk to be completed
For B = 1 to the length of C do
If CA,B < > 0 AND Xi+Aj+B < > CA B
Then Begin
P - cA,B
;; P is the piece needed to complete chunk C
a «- A
b - B
End Begin
For I = 1 to 8 do
For J = 1 to 8 do
If Xu = P
Then If piece P can be moved from board location (I,J)
to location (i + a, j + b)
Then suggest this move as a possible next move
End Begin
"igure 25: Pseudo-code to suggest possible adversary moves.

95
chunk can produce multiple suggestions if more than one of the required piece to
complete the chunk exists on the board and can be moved to the proper location.
IAM’s accuracy with respect to the entire game is fairly consistent, with
correct predictions ranging between nine and eleven percent of the total moves of
the game. A slightly wider range, fourteen to twenty-three percent of the total
moves, is obtained for predictions identifying the correct piece to be moved. While
being able to correctly predict approximately ten percent of an adversary’s moves is
an extraordinary accomplishment, the usefulness of our methodology for coaching
existing chess programs is dependent on the reliability of the predictions or the ratio
of the correct predictions to the total number of predictions made for a game.
42.1.2 Reducing incorrect predictions through likelihood
The percentage of correct predictions to total predictions made by IAM,
shown in Table 1, falls between twelve and seventeen percent. This means that a
chess program’s evaluation function can rely on IAM’s predictions one sixth of the
time. In Chapter 3, we noted that IAM calculates a probability or likelihood which,
after being modified by the heuristic inference rules, is used to determine the move
that the adversary will make next.
By disregarding predictions with low likelihood values, we improve the
performance ratio of correct prediction to total predictions. The likelihood values
and corresponding prediction ratios and percentages for the results shown in Table
1 are displayed in Table 2. The top row of Table 2 displays the likelihood values
used to select predictions. Entries in the table accrue the predictions from the next

96
highest likelihood value so that the second column for likelihood values less than
twenty-five percent are identical to the total prediction ratios shown in Table 1. The
percentage of correct predictions to total predictions is shown beneath each ratio
pair.
We can see that by limiting the predictions to be considered by the existing
chess program to have a likelihood value of greater than fifty percent, the predictions
of IAM are from seventy-five to one hundred percent correct. This enables a chess
program to use IAM’s knowledge without fear of blundering into the trap of utilizing
an incorrect prediction. The only problem which arises from requiring predictions
to have a high likelihood is that correct predictions with smaller likelihoods are
ignored. From Table 2 we see that the percentage of game moves which are
correctly predicted drops to just over five percent.
Table 2: Effect of Likelihood on Prediction Ratios
Games
< 25%
25-30%
30-40%
40-50%
> 50%
12
5/31
4/9
3/6
3/5
3/4
16.1%
44.4%
50 %
60 %
75 %
22
6/47
3/7
3 / 6
3/4
3/3
12.7%
42.8%
50 %
75 %
100 %
80
6/36
3 / 6
3/3
3/3
3/3
16.6%
50 %
100 %
100 %
100 %
4.2.1.3 Performance versus the adversary as the black player
The second part of our initial experiment to determine the base performance
levels of IAM uses TEST2 to establish IAM’s capabilities against an adversary
playing the black pieces. TEST2 is the game from Botvinnik’s championship match

97
against Petrosian which followed the TEST1 game. The length of the TEST2 game
is 58 turns. The performance results for IAM during the simulated TEST2 game are
displayed in Table 3. The meanings of each of the headings in Table 3 are the same
as for Table 1. Because the knowledge acquisition and inductive learning segments
have already been completed the first three columns of Table 3 are identical to
Table 1, except the textual opening move knowledge base numbers represent the
total chunks and the chunks for the black opening moves of the adversary.
Table 3: IAM Performance Measures for TEST2
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C+P
12
28
24/14
17
0
4
29.3
0.0
23.5
22
53
31/17
17
1
3
29.3
5.9
23.5
36
66
38/22
17
1
4
29.3
5.9
29.4
45
68
39/23
17
1
4
29.3
5.9
29.4
80
86
52/30
23
1
5
39.7
4.3
26.0
We immediately see that Table 3 demonstrates learning or an increase in
performance as more games are analyzed by IAM. The correct predictions rise from
zero to a consistent two percent (5.9 percent multiplied by 29.3 percent) of the total
moves of the game. Identification of the piece to be moved increases gradually from
seven percent to over ten percent of the game moves.
We previously mentioned a modification to the textual chunk knowledge base
to capture the responsive nature of black’s play. The results of this modification are
presented in Table 4. The differences between Table 3 and Table 4, other than the

98
size of the textual chunk knowledge base, are the results of the opening move
predictions which rely on the textual move knowledge base.
The modification to the textual chunk knowledge base, which simulates the
responsive nature of human black piece play, increases the prediction capabilities of
IAM. The last two rows of Table 4 show a one hundred to two hundred percent
increase in correct predictions from the original model. To verify this result, we
randomly selected another game from the Petrosian match in which Botvinnik played
the black pieces. Again, the original adversary model made one or two correct
predictions for each size of the knowledge bases. However, the adversary model with
the modified textual knowledge base produced an additional correct prediction,
effectively doubling the performance characteristics of the adversary model.
Table 4: Performance with Responsive Modification
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C + P
12
28
32/22
17
0
4
29.3
0.0
23.5
22
53
42/28
17
1
3
29.3
5.9
23.5
36
66
57/41
17
1
3
29.3
5.9
23.5
45
68
64/48
17
3
4
29.3
17.6
41.2
80
86
89/67
23
2
4
39.7
8.7
26.1
Even with the modification to account for the responsive nature of black play,
IAM’s performance against an adversary playing the black pieces is only fifty percent
of the performance against an adversary playing white. The reason for this is that
the quality of knowledge available about the adversary as a black player is inferior
to the knowledge available about the adversary as a white player. The first several

99
knowledge bases, produced from the twelve and twenty-two game analyses, have a
maximum of twenty-three percent of the games played as black, or five of twenty-two
games. The larger knowledge bases have a greater quantity of black games and a
more even distribution with forty-six percent of the eighty game knowledge base
being black games.
The quality of the knowledge about the adversary as a black player is still
inferior. Sixty-seven percent of the white games played by the adversary resulted in
won games. Only forty percent of the black games resulted in a winning outcome.
Remember that we normally use a series of lost games to indicate an imminent
change in the adversary’s strategy or playing style and chunks which are only found
in lost games have their move suggestion probabilities reduced by the prediction
inference engine. Therefore, the poor quality of information available about the
adversary as a black player is directly responsible for the reduced performance of the
adversary model against the future games of the adversary in which the adversary
plays the black pieces.
4.2.2 Performance with Respect to Time
We conducted a two part experiment to reveal temporal dependencies of the
adversary model. In the first part of our experiment, the geometric and textual
chunk knowledge base of the adversary model is created from observations of the
past performances of an adversary which occur in time near to the TEST1 game.
This is to detect any reliance on the recency of knowledge.

100
Botvinnik’s 1961 chess match versus Tal is used to analyze games that are
recent to the 1963 TEST1 game. IAM’s performance is shown in Table 5. The
overall correct predictions are fewer, only seven percent of total game moves, and
the identification of the correct piece to be moved is comparable to the forty-five
game knowledge base. The only noticeable change produced by analyzing recent
games is that the number of analogous situations detected by the adversary model
is lower than any of the previous knowledge bases using older knowledge. The
reduction in predictions corresponds to an increase in the correct prediction to total
prediction ratio, thus increasing the reliability of IAM’s predictions.
Table 5: Effect of Recency on IAM Performance for TEST1
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C + P
12
16
26/13
16
4
4
28.0
25.0
50.0
15
29
26/13
14
4
4
24.5
28.5
57.1
The second part of the temporal experiment tries to simulate the effect of
forgetting on an adversary’s pattern knowledge by aging the geometric pattern
chunks. Once a knowledge base exists in the adversary model for a particular
adversary, the geometric pattern chunks have their number of observances halved
prior to incrementally adding new chunks or reinforcement of existing chunks from
analyzing additional games played by the adversary. If a chunk’s number of
observations is less than one, then the chunk is deleted from the knowledge base.
This has the effect of removing chunks learned from earlier/older games if they are
not repeated in more recent games.

101
The performance of IAM against TEST1 with aging geometric chunks is
shown in Table 6. The textual chunks column is identical to Table 1, since textual
chunks are not being aged and therefore has been replaced by a column, REM,
indicating the number of chunks removed from the knowledge base due to aging.
The 1961 Tal games of Botvinnik were not used in this experiment hence, the last
row indicates only sixty-five games were analyzed to produce the adversary model
used by IAM.
Table 6: Effect of Aging/Forgetting on IAM’s Performance
#G
#Ch
REM
#P
#C
CPWL
%G
%C
%C+P
12
28
N/A
31
5
8
54.3
16.1
41.9
22
53
0
47
6
7
82.4
12.7
27.6
36
48
18
36
4
4
63.2
11.1
22.2
45
32
25
11
3
0
19.2
27.2
27.2
65
53
20
22
2
3
38.5
9.0
22.7
We see from Table 6 that some of the geometric chunks are consistently
repeated and therefore not deleted from the knowledge base. The geometric
patterns of pieces that are surviving in the adversary model are primarily chunks
consisting of pawn structures.
The introduction of forgetting via an aging mechanism for the geometric
chunks produces a steadily decreasing overall performance of IAM against TEST1.
While we do not propose that chess masters never forget, a better factor for
removing knowledge from the knowledge base would involve a combination of the
age of a specific piece of knowledge and a size limitation on the number of chunks

102
which can be stored in the knowledge base. In Chapter 2, research indicated that
chess masters store 10,000 to 100,000 patterns. The size of the adversary model
knowledge base after analyzing eighty games is one hundred and seventy-five
patterns, half of which are geometric chunks and the other half are textual patterns.
Many thousands of games, similar to the life experiences of a chess master, would
need to be analyzed to produce a knowledge base size approaching 10,000 patterns.
4,2.3 Performance with Respect to Textual Chunk Size
The results for IAM already shown have all been for a textual move pattern
knowledge base which learns the statistical patterns associated with the first five
opening moves. Other textual move patterns are learned, but are not used by IAM
for predicting the opening move sequence of the adversary. We modify the textual
pattern learning algorithm to learn and analyze the first ten moves and the first
fifteen moves of every game. The performance results of IAM operating with the ten
opening moves textual pattern learning is presented in Table 7 and the fifteen
opening moves knowledge base results are presented in Table 8. Each table shows
the knowledge base obtained from analyzing Botvinnik’s three earliest tournaments.
We feel that the results are sufficient to demonstrate the effect of changing the
opening move knowledge base size.
IAM’s performance with the larger textual knowledge bases is better for the
case of analyzing twelve games, with a twenty percent increase in correct predictions
and a fifteen to twenty-three percent increase in correct piece to be moved
identification, for the ten and fifteen move knowledge bases respectively. However,

103
Table 7: Performance (TEST1) with 10 Opening Moves
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C+P
12
28
60/31
32
6
9
56.1
18.7
46.8
22
53
84/46
48
6
7
84.2
12.5
27.0
36
66
112/57
49
6
8
85.9
12.2
28.5
Table 8: Performance (TEST1) with 15 Opening Moves
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C + P
12
28
102/52
34
6
10
59.6
17.6
47.0
22
53
152/81
50
6
6
87.7
12.0
24.0
36
66
213/110
51
6
7
89.5
11.7
25.4
the performance returns to the five opening moves textual knowledge base level for
the adversary model produced by analyzing twenty-two games and actually drops
below the five move knowledge base performance level for the thirty-six game
adversary model.
The increase in performance for the twelve game adversarial model textual
knowledge base results from predictions covering the sixth through tenth or fifteenth
moves which were not found to contain positions analogous to the geometric chunks.
The responsibility for prediction shifts from the geometric chunks to the textual
chunks. The return to normal performance levels and the drop in performance
experienced for the twenty-two and thirty-six game adversary models occurs from the
subsequent geometric chunks which are learned and able to find analogous positions
during the extended opening time period. The newly learned geometric chunks are

104
filling in the gaps for the sixth and later moves that are present in the smaller twelve
game adversary model.
The number of textual patterns stored in the adversary model’s knowledge
base increases significantly with the change to capture additional opening moves as
textual patterns. For the ten opening moves textual pattern modification the textual
portion of the knowledge base is tripled and the fifteen move modification exceeds
quintupling the number of textual chunks. Since no apparent benefit is produced by
the extension of the textual patterns, we do not recommend this modification except
when the number of geometric chunks is small (e.g., less than fifty chunks).
4.2,4 Performance with Respect to Geometric Chunk Size
Our next experiment is to demonstrate the adequacy of the four-by-four or
sixteen square geometric chunk size. As shown in Appendix B, the majority of the
chunks induced by IAM could have been acquired with a three-by-three or nine
square chunk size, but we use the geometric chunks to save the ending piece count
and this requires a sixteen element chunk. It is possible that since we discard chunks
which exceed the maximum size, larger chunks which will improve the adversary
model’s performance may be lost.
We research the performance of the adversary model with respect to
geometric chunk size by altering the chunk acquisition algorithm following the
convolution operator to accept chunks of a twenty-five square area and again to
accept chunks of a thirty-six square area. The results for the first three adversary
models are shown in Table 9, for the five-by-five area chunks, and Table 10, for the

105
six-by-six area chunks. The ‘nc’ in the textual chunks column indicates ‘no change’
from the Table 1 values.
Table 9: IAM Performance (TEST1) with Five-by-Five Chunks
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C + P
12
28
nc
31
5
7
54.3
16.1
38.7
22
54
nc
46
6
6
80.7
13.0
26.0
36
67
nc
45
6
6
78.9
13.3
26.6
Table 10: IAM Performance (TEST1) with Six-by-Six Chunks
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C + P
12
29
nc
31
5
7
54.3
16.1
38.7
22
55
nc
45
6
6
78.9
13.3
26.6
36
68
nc
45
6
6
78.9
13.3
26.6
The number of new chunks induced can be obtained by subtracting the
number of geometric chunks found by the four-by-four algorithm shown in Table 1
from the number of chunks shown in the above tables. One five-by-five geometric
chunk and one six-by-six geometric chunk are produced for the thirty-six game
analysis of row three in the tables. The five-by-five chunk is shown in Figure 26
along with another chunk that is induced for both the five-by-five and four-by-four
chunk sizes.
The performance is identical for the twelve game adversary model regardless
of the geometric chunk size. We see in Figure 26, that the smaller chunk is
contained within the larger chunk and is capable of making all predictions involving

106
Figure 26: Chunk acquired with 5x5 size and corresponding smaller chunk.
the seven pieces of the chunk. However, the five-by-five chunk requires that an
additional two pieces, the queen and extra pawn, be in place to use this chunk for
predicting the identical moves of the four-by-four chunk. IAM’s performance actually
suffers for the next two larger adversary models due to interference from the
five-by-five chunk induced during the twenty-two game analysis.
The new chunks, which cover a larger board area, already exist on the game
board during a time that the smaller chunks identify analogous positions. The
concurrent existence of the new larger chunk causes the prediction inference engine
to prohibit the smaller chunk move suggestions, some of which were able to predict
the correct piece to be moved. Because IAM’s inference engine simulates the

107
hierarchical construction of complex chunks from smaller less complex chunks, the
overall performance of IAM is better when we use the smallest chunk size of
four-by-four. Based on IAM’s performance we define a primitive chunk to contain
three to nine pieces and cover a board area of no more than sixteen squares.
4.2.5 The Effect of Domain Specific Knowledge
Our adversary modeling methodology uses only minimal amounts of domain
knowledge, such as the movement capabilities of each piece, to predict the actions
of an adversary. Current chess programs use a myriad of domain specific knowledge
to evaluate board positions while selecting moves. Our final experiment is to
determine the effect on performance when domain specific knowledge is contained
in the adversary model.
We augment the heuristic rules used by the prediction inference engine to
include knowledge concerning pieces being en prise. The purpose of the heuristic
is to invalidate move suggestions when a major piece is en prise, unless the suggested
move will change the board location of the en prise piece. Our heuristics are simply
rejecting inferior move choices and not attempting to suggest any alternate moves
that will save the en prise piece.
IAM’s performance for the twelve game and twenty-two game adversary
models against TEST1 is shown in Table 11. The performance for the adversary
model created from the analysis of twelve of Botvinnik’s games is identical to the
original results given in Table 1. However, the performance for the twenty-two game
adversary model shows a reduction in the number of predictions made by the

108
adversary model from forty-seven to thirty-four. This corresponds to an increase in
the reliability of the predictions and further increases the utility of the adversary
model’s predictions for existing chess programs, as discussed in section 4.2.2.
Table 11: Effect of Domain Specific Knowledge
#G
#Ch
#M
#P
#C
CPWL
%G
%C
%C + P
12
28
24/10
31
5
8
54.3
16.1
41.9
22
53
31/14
34
6
6
74.5
17.6
35.3
The addition of the small amount of knowledge concerning en prise pieces
increases the reliability of IAM’s correct move predictions by five percent and is
higher than all of the reliability values displayed in Table 1. An explicit example of
IAM’s performance improvement is seen in the chunks’ predictions for the position
in Figure 22. Since the queen is en prise, only the fourth chunk’s suggestion which
predicts the correct move is permitted by the new heuristics, thus increasing the
correct predictions for that game by one. IAM’s improved performance with the
addition of a minuscule amount of domain specific knowledge indicates that further
improvements in reliability can be expected with the continued addition of greater
amounts of domain specific knowledge.
4,2.6 Repeatability of the Demonstrated Results
The performance results given in Section 4.2, have been obtained from the
analysis of TEST1 and TEST2 by IAM. To verify these results and gain confidence
in the statistical significance of the displayed results, we had IAM predict adversary
moves in five additional games. Two of these games are from the end of the

109
Hague-Moscow Tournament; therefore, only the games preceding these test games
were used to create the adversary model knowledge base and the other three are
additional games from the 1963 Petrosian match.
IAM’s performance in all five of the additional games is very similar to the
results which have already been reported. The two white games had durations of
forty-eight and sixty-three moves with IAM making five and six correct predictions
respectively, or approximately ten percent of the total moves, and identification of
the piece to be moved approximately twenty percent of the total game moves.
Performance in each of the three additional black games is similar to the
demonstrated results with correct predictions ranging from one to three per game
and slightly higher correct piece identification.
For the total seven games, IAM’s performance is consistent across the three
games in which the adversary played the white pieces and the four games in which
the adversary played the black pieces. The ability of IAM to repeatedly obtain
similar performance levels increases the statistical significance of IAM’s performance
results.

CHAPTER 5
RESULTS
We have successfully acquired geometric and textual chunks that have been
repeated by an adversary in more than one game. Our assumption that these chunks
form the foundations of an adversary’s evaluation criteria for the domain is validated
by the actual predictions made by the adversary model. Approximately ten percent
of all game moves, when the adversary plays the white pieces, are correctly predicted.
Furthermore, for over twenty percent of the moves in a game we can identify the
piece which is about to be moved.
The identification of which piece is to be moved indicates that our adversary
model is focussing attention in the same locale as the adversary’s attention.
Additional support for the similarities between the adversary modeling methodology’s
perception of the board and the adversary’s board evaluation is found by looking at
the two moves following the adversary model’s predictions. For the TEST1 game,
within two moves of a prediction made by IAM, the predicted move was executed
once and the piece which IAM predicted to be moved was used four times.
Although IAM’s predictions were a little premature, we can see that the adversary
model is cognizant of the area of the board where action is about to take place.
The performance of the adversary model is exceptional when we account for
the quality of the information analyzed. Table 12 shows the win to loss ratio for the
110

Ill
Botvinnik games analyzed by IAM. High quality information is obtained from won
games and lost games provide us with questionable-quality knowledge. Again, we see
that IAM’s lack of performance against the adversary as the black player results from
sixty percent of the analyzed black games having ended in a loss.
Table 12: Ratio of Won and Lost Games for White and Black
White Games
Black Games
Total Games Analyzed
43
37
Won Games
29
15
Lost Games
14
22
Our research has resulted in three specific advancements for intelligent
programs operating in adversarial domains. Through use of the adversary modeling
methodology we can reduce opening book sizes while maintaining comparable
performance capabilities, reveal critical paths in the game tree not previously
considered, and reduce search complexity by pruning unnecessary nodes from the
game tree. The advancements are detailed individually with respect to their
application in the chess domain.
5.1 Reducing Opening Book Size
Most current chess programs make use of opening books ranging in size from
5,000 to 60,000 bytes (Newborn & Kopec, 1990). The size of an opening book affects
the speed of the program’s play with larger books taking more time to process.
Research is being performed to reduce the size of opening books for chess programs

112
by taking advantage of redundancies among various opening lines (White, 1990).
White’s method affords a thirty-eight percent savings in space, reducing a 5,000 byte
opening book to 1900 bytes.
The collection of textual chunks referring to opening moves simulates a
human master’s analysis of an opponent and requires substantially less space than
standard opening books. If we assume the same memory constraints used by White
of two bytes per move, then the opening moves learned by the adversary model
require 312 bytes. This is the maximum size for the opening book from Table 4,
using the responsive code for black moves by the adversary after analyzing eighty
games. We also need three additional bytes per move to save the number of times
a move has been observed, the color of pieces played by the adversary for each
move, and the mean occurrence time of each move, increasing our storage
requirements to 468 bytes. Even with our additional storage requirements, the
adversary model’s textual knowledge base provides over a ninety percent savings in
the space requirements for an opening book.
To utilize the adversary model, a chess program should analyze the collected
opening sequences saved in the textual knowledge base to identify opening sequences
that are unfamiliar to the adversary. Using our definition of five game turns for the
standard duration of a known opening, we can still store fourteen complete opening
sequences, with moves for both players, and maintain twice the space savings of
previous opening book reduction methods. This will enable chess programs to gain
a strategic advantage over their opponents by using unfamiliar openings while still
reducing their opening book size by eighty percent.

113
5.2 Revealing New Search Paths in the Game Tree
Adversarial domains, especially games, frequently use tree models to represent
important information in the planning process. Domains using tree representations
must use a search algorithm for selecting the optimal solution from the tree.
Adversarial domain tree representations are very large, 1043 nodes for chess, which
prohibits an exhaustive search to find the solution. Heuristics are used to prune
branches from the tree with only partial knowledge of what is contained down the
pruned branches and to evaluate current board positions to determine the end result
of a path even though a leaf node has not yet been reached by the search algorithm.
We improve the performance of search algorithms by revealing branches or
paths that would not normally be considered. In Figure 27, we present a subtree
from the larger game tree for a hypothetical chess game. The computer chess
program at the MAX node level is attempting to select the optimal move from
node A through use of the standard minimax search with alpha-beta pruning. The
evaluation function values for each of the nodes at the bottom of the search range
from negative five, for a sure loss to the computer, through positive five, for a sure
win to the computer, with zero representing a drawn outcome.
The left subtree returns a value of zero to the adversary’s MIN level decision
node. The MAX player, the computer, now knows that a draw can be obtained by
selecting the move corresponding to the left branch of this game tree. Likewise, the
right branch returns a value of negative one, or a slight chance of a loss, as the MIN
player’s movement choice. The first node evaluated for the center branch has a

114
negative three value. Alpha-beta pruning would cut the remaining two branches
from node B because a better value is already available to the MAX player from the
right subtree. Therefore, negative three becomes the value associated with selecting
the center branch of the tree. The minimax algorithm will choose the maximum
node value from the MIN level from the right subtree, resulting in a probable draw
as the outcome of the game.
With the appropriate knowledge in the adversary model, a much better
outcome could be obtained by the chess program. The move that is represented by
the center branch of the game tree extending from node A corresponds to the
proffering of a gambit by the MAX player and the right-most branch from node B
corresponds to the acceptance of the gambit by the MIN player. The adversary

115
model of this adversary, the MIN player, shows that the adversary has never accepted
a gambit in previous play. Therefore, the negative three valued node should not be
considered as a viable move choice for the adversary to consider. The new game
tree, minus the gambit acceptance node, now returns a value of positive two to the
B node as MIN’s choice. The new best move for the MAX player using knowledge
of the adversary’s past performances is to choose the center branch of the tree and
proffer the gambit.
A specific example of the ability of our adversary model to reveal adversary
move choices that were not previously considered by a chess programs search
algorithm comes from the game shown in Figure 19 and using the geometric chunks
displayed in Figure 6. Figure 28 shows the board position prior to the twenty-first
move to be made by Botvinnik and the chunk used in predicting the move. Two
commercial chess programs were asked to analyze this position and select the next
move to be made.
The SARGON 4 program, searching to a depth of six ply, selected the Rcl
move and the GNUCHESS program, searching to a depth of eight ply, selected the
Ne5 move for the position shown in Figure 28. IAM predicted that Botvinnik would
move the knight located at d3 to either the b4 or e5 square. Both move choices were
predicted with equal likelihood since each move will create the same chunk. Each
of the commercial programs, which have a program option to reveal their search
trees, never considered the Nb4 move in their respective searches. Botvinnik actually
made the Nb4 move choice that was one of the two IAM predictions. This example

116
HEMM\
jjijjjjjjlij;
III ¡M 1
mm
tin
â– HI:
^4, i'j’i
m i
¡Hi i ill i (
i1:*;*;*!* i!:;*
ÍÍÍiii|iÍíÍííí*i
í ill HI
*!*;«:*:*;=;*!
ü ill ñ l
a ill
S &3
B! a H
Hli
figure 28: Prediction of Botvinnik’s next move.
Chunk used
â– to predict
NM and NeB
illustrates that the adversary model is capable of alerting existing chess programs to
move choices that are either not considered at all, as above, or not considered to be
optimal selections.
5.3 Reducing Search Complexity Through Automatic Pruning
Another improvement to artificial intelligence search strategies performance
is made by the adversary model through eliminating the need to perform search at
certain levels of a game tree. Providing a unique descending branch from a node has
the same effect as pruning all of the other branches from that node, which reduces
the search complexity.

117
As we have shown in Chapter 4, IAM is capable of predicting the exact move
to be made by an adversary ten percent of the time. For an additional fifteen
percent of the game, IAM predicts the correct piece to be moved by the adversary.
The contributions of this knowledge in the adversarial domain of football are
obvious. If the coach of one team knows the exact play to be called by the adversary
every tenth down and which player will end up with the ball on one and one half of
the remaining downs, then the coach has a significant strategic and tactical advantage
over the opposing team.
For adversarial domains using tree representations of domain knowledge, the
knowledge acquired by the adversary model succeeds in reducing the tree complexity
which corresponds to a deeper search for the same number of evaluated tree nodes.
When a search is performed to a depth of twenty ply with a branching factor of N,
the maximum number of branches which descend from all nodes of the tree, then
with a prediction rate of ten percent we will be able to accurately predict the
individual branch which will be followed by an adversary from one of the nodes.
This effectively prunes N minus one branches from the tree at that node and all of
the associated sub-branches descending from the pruned nodes. By adding a single
branch to the tree, we can now search to a depth that is one greater than the original
search. A comparison of the two searches with N equal to four is demonstrated
graphically in Figure 29, which shows the standard search tree and the extended tree
obtained by using the adversary model’s knowledge for the same number of expanded
nodes.

118

119
The time order for the standard search algorithm is
0(standard search) = kl * Nm = 0(Nm),
where ‘m’ is the depth of the search and ‘kl’ is a constant greater than one
corresponding to the time required to actually perform the search and evaluation
algorithms. With ‘p’ representing the percentage of correct predictions made by the
adversary model and the constant ‘k2’ corresponding to the time required to use the
adversary model’s knowledge base and prediction algorithm, the new time order of
the search algorithm is
0(prediction search) = k2 + (kl * Nm’(1'p)) = 0(Nm’(1'p)).
This shows that the time order can be reduced, since ‘p’ is less than or equal to one,
by an exponential factor with the addition of a linear constant to the time equation.
Essentially, a search algorithm can now afford to search (m * p) extra ply
deeper in the game tree with only a linear increase in time. In Chapter 4, we
mentioned that search algorithms would require reliable predictions. The
mathematical equations above assume perfect reliability. We have seen that a one
hundred percent reliability can be obtained by only considering chunks from the
adversary model knowledge base that have a fifty percent of greater likelihood value.
The use of high likelihoods correspondingly decreases the percentage of correct
predictions from ten percent to five percent of the total game moves. With a five
percent prediction rate, a tree search of forty ply will have its time order reduced
from 0(N40) to 0(N(40 ’(1' (0 05/2)))) 0r 0(N39). A forty-one ply depth search can be
accomplished in the same time as the previous forty ply search. The reason for

120
dividing the ‘p’ prediction percentage value by two is that only every other branch in
the game tree is predictable by the adversary model.
Because we require perfect reliability and we are using induction so that only
domain situations that are analogous to situations which have been induced will
produce predictions, the effect of gaining one ply of search every twenty game turns
is a reasonable result. In section 4.2.5, we showed that the inclusion of even a
minimal amount of domain specific knowledge in the adversary model significantly
decreases the number of wrong predictions that are forecast by the model. With
greater amounts of domain specific knowledge, we believe that the ten percent
correct prediction performance rate can be obtained with nearly perfect reliability
hence, doubling the exponential savings to the time order of adversarial domain
search algorithms.
We now assume that the addition of a larger quantity of domain specific
knowledge will enable the adversary model to make only accurate predictions or
predictions that identify the correct piece to be moved. Even greater time savings
are obtained if we use the predictions which identify the correct piece to be moved.
Utilizing the prediction heuristics of correct pieces, in addition to correct moves,
requires that the search algorithm still performs a search at each depth level of the
game tree. The improved search is still faster than the standard search algorithm,
since only those moves that are going to move the predicted piece need to be
examined. For example, in a search tree which has a breadth of search ‘N’, of five,
only two or three of the suggested moves will actually use the piece predicted by the

121
adversary model. An example of the improved search tree from using the adversary
model’s correct piece identifications is shown in Figure 30.
Figure 30: Search tree obtained from Correct Piece identification.
The time order equation uses ‘q’ to represent the percentage of actual moves
for which the adversary model can predict either the correct move or identify the
correct piece to be move. We must remember from Chapter 4, that ‘q’ is twice as
large as ‘p’. The new reduced branching factor for nodes where predictions are made
is ‘R’ and we can guarantee that ‘R’ is less than ‘N’. The new time order is
O(improved prediction search) = (k2 * Rm”q) + (kl * Nm (1'q))
= 0(Nm*(1-q)) or 0(Rm ’q).

122
The actual time order depends on the value of ‘q\ Because of the need to actually
perform a reduced search at each of the tree nodes associated with adversary model
predictions, the resultant additive factor is also an exponential instead of a linear
constant. If ‘q’ is less than one half, or fifty percent, then the order is the first
alternative. Otherwise, the second alternative is larger and will become the time
order of the search algorithm. With the twenty percent correct piece identification
performance demonstrated by IAM in Chapter 4, the improved algorithm reduces the
time order of the standard search algorithm from 0(N20) to 0(N18) or a two ply
increase in the search capabilities of the standard search algorithm. This is one
order of magnitude better than the prediction based search which uses only the
correct move predictions, although the reduction in complexity is not as intuitively
appealing.
5.4 Summary of Research Contributions
In Chapter 2, background research claimed that the speed and depth of search
are critical factors in the playing ability of intelligent game programs which use tree
representations of the game domain. We have reduced the size requirements of
opening books for chess programs with a corresponding reduction in the amount of
time required to use the smaller opening book. The predictions of an adversary’s
move by the adversary model has resulted in an exponential savings in time for
standard game search algorithms. Time savings realized from the use of the
adversary modeling methodology correspond to increases in the depth of search that

123
is accomplished by a search algorithm for the same amount of time. Deeper
searches in the game tree provide better tactical solutions to chess problems.
The textual chunks which are used to reduce the opening book size have a
correlative strategic contribution to adversary domain programs. A chess program
can use the knowledge in the textual portion of the adversary model’s knowledge
base to select opening sequences of play that have not been experienced by the
adversary. This produces a strategic benefit to the chess program by placing the
adversary in unfamiliar territory.
Improvements in playing ability of current adversarial programs are realized
through the revelation of previously unconsidered search paths. The adversary model
enables a chess program to determine the most likely move to be made by an
adversary based on the adversary’s past performance.
As increases in search speed through technological advancements in hardware
capabilities approach their physical limitations, further increases in intelligent
program performance must be accomplished by using knowledge. The adversary
model provides a methodology for simulating the cognitive preparations that experts
in adversarial domains use when preparing to meet an adversary. Previous methods
have limited themselves to small stages of the game. We have produced a generic
method which can be used to augment intelligent program performance at all stages
of a game.

CHAPTER 6
CONCLUSIONS AND FUTURE RESEARCH
6.1 Conclusions
We have developed and implemented an adversary modeling methodology
which acquires the cognitive chunks that are used by an adversary to evaluate
complex domain situations. The chunks are acquired by employing a technique
frequently used by adversarial domain experts which is to analyze the previous
performances of a specific adversary. Chunks are comprised of both textual and
visual patterns that are repeated at different times in the adversarial domain. For
chess, this means a pattern that occurs in two or more separate games.
Our performance results indicate that we can accurately predict ten percent
of an adversary’s moves when high quality information, or a high percentage of won
games, has been analyzed. Additionally, we can predict the piece to be moved for
over twenty percent of the adversary’s move choices in a game. These results reveal
that adversarial domain experts frequently use basic cognitive techniques that are
domain independent to choose their moves. The cognitive technique that we rely on
is the simplification of complex domain situations to situations that have been
encountered previously.
124

125
' Current chess program capabilities result from a combination of search speed
and use of domain specific knowledge. Our adversary model is not exempt from the
use of domain specific, since we have seen in Chapter 4 that the addition of domain
specific knowledge increases the reliability of the adversary model’s predictions. We
have demonstrated that by using our domain independent knowledge the complexity
and time order of a search algorithm can be reduced. The use of our adversary
modeling methodology’s domain independent knowledge increases the performance
capabilities of current chess programs. We believe that the continued increase in
performance of intelligent chess programs requires the use of additional knowledge
including domain independent knowledge similar to an adversary model.
6.2 Future Research Directions
We have tried to indicate, throughout this dissertation, meaningful extensions
to the adversary modeling methodology’s current research. A summary of these
extensions and some additional thoughts for future research directions are presented
below.
6.2.1 Acquire More Geometric Chunks
We have seen that the performance of the adversary model increases with the
addition of more chunks as long as the chunks do not interfere with each other. The
convolution algorithm identifies all pieces that are horizontally or vertically proximal
or can defend another piece diagonally. Chunks which exceed a sixteen square size
are ignored. Due to the hierarchical composition of complex chunks from smaller

126
simpler chunks, we feel that chunks which are originally to big to be evaluated by the
induction algorithm can be broken down into smaller chunks that maintain the
meaning of the original chunk. Trying to eliminate pieces from a chunk by ignoring
all pieces beyond a particular horizontal or vertical axis of the game board may
destroy the strategic and tactical significance of the chunk. A better method would
be to devise heuristic rules which could eliminate fringe pieces from a chunk while
maintaining the strategic and tactical importance of the new chunk’s pieces.
6.2.2 Increase the Knowledge Contained in Each Chunk
The knight is currently included in chunks without regard to its movement
capabilities. The convolution algorithm can be modified to use a five-by-five
template when a knight is contained in a chunk to capture the movement capabilities
of the knight. Increasing the template size, with suitable modifications for each
individual piece’s movement capabilities as shown in Figure 31, enables the adversary
modeling methodology to acquire chunks that are meaningful, but not proximal. The
original templates, shown in Figure 8, used by the convolution algorithm returns a
bitmap of values that are independent of the center piece. To save space since the
new bitmap values will need to store values up to 256, the new templates are
dependent on the central piece and further processing by the chunk acquisition
algorithm will need to refer back to the central piece at each board location. If the
enlarged templates tried to use the general piece independent approach of the
original templates then bitmap values up to 2255, or one bit in each bitmap for every
position of the fifteen by fifteen templates of the bishop, rook, and queen.

127
3
16
4
32
/
/
2
64
1
128
Knight Template
Blanks have zero value
512
32
2
256
16
1
/
/
4
64
IK
8
128
Part of the
15 x 15 Rook Template
Figure 31: Knight template and partial Rook template.
Additional knowledge can be obtained by dropping our similarity constraint
of chunks containing one color of pieces. Permitting chunks to contain both colors
of pieces simultaneously has already been used to acquire knowledge about the pawn
structures and pawn chains found in an adversary’s games. Along with the increased
template size mentioned above, chunks with both color pieces enables the adversary
modeling methodology to acquire knowledge concerning the use of forks, pins, and
skewers. In addition to the vision based convolution method for acquiring chunks,
another computer vision technique lends itself to the acquisition of non-proximal
relationships between pieces. The Hough transform is used to identify the beams of
light which pass through a specific point in an image. While Hough transforms are
cumbersome to implement in vision domains, the discrete nature of board games
would permit an efficient implementation of the Hough transform which would
identify pieces that both lay on a particular line of the board and would thus capture

128
distance relationships without the need for over-sized templates in the convolution
operator.
Storing the knowledge acquired by either a Hough transform or the
convolution operator with fifteen by fifteen templates would require a sixty-four
square chunks size, or quadruple the current chunk size. Although we can
implement the sixty-four square chunks, the additional space required would make
the adversary model inefficient and less desirable as a slower coach. The use of
semantic networks would enable the adversary modeling methodology to store all
chunks in a space efficient manner. Semantic networks would require a change in
the functional description of our implementation of the adversary modeling
methodology, but would not change the theoretical implications of the methodology.
6.2.3 Place More Adversarial Knowledge in the Adversary Model
We have already seen in Chapter 4, the inclusion of domain specific
knowledge increases the reliability of the adversary model’s predictions. Increasing
the domain specific knowledge available to the adversary model will further augment
the capabilities of the adversary modeling methodology. For game domains which
already have a great amount of domain specific knowledge in their evaluation
functions, cooperation between the adversary model and the evaluation function
might serve to eliminate inferior predictions. However, many adversarial domains
do not have a large amount of domain specific information contained in their
application programs. We can easily modify the prediction inference engine heuristic
rules to increase the quantity of domain specific knowledge available to the adversary
model.

129
Other types of global knowledge are also available to increase the adversary
model’s capabilities. In Chapter 2, we mentioned that cultural and educational
differences affect the strategic and tactical planning of adversaries. Kotov and
Yudovich (1961) state that large scale instruction and training in chess is backed by
the Soviet government via the Trade Unions. The sociological and cultural traits of
the idealized Soviet man are the primary traits of the chess school. The desire to win
corresponds to ardent patriotism. Specific strengths and weaknesses are displayed
by similar cultural groups. A statistical analysis of the 1973 international chess
tournaments revealed several statistically significant chess weaknesses produced by
a Soviet chess education (Mednis, 1978). Table 13 displays some of the weaknesses
that were found by Mednis’s analysis. The global knowledge about Soviet chess
players displayed in Table 13 would a chess program which knew that its opponent
was a Soviet player. A chess program playing against a Soviet opponent could use
the fourth row of the table to realize that time pressure is relevant in one fourth of
Soviet chess losses and correspondingly a chess program could reduce its search
depth from a possible high of thirteen ply to a maximum of ten ply to force the
Soviet player into a time sensitive situation.
Our current research focuses on the perceptual evaluation of domain
situations by a domain expert. The next step performed by chess masters after
evaluating the board position is to plan their attack or defense. The prediction
capabilities of the adversary model correspond to the evaluation mechanisms used
by the adversary. Suggesting moves which are strategically and tactically superior

130
against a specific adversary is the next logical step. The adversary model already
infers the general playing style of an adversary and acquires the opening sequences
displayed by an adversary. These two pieces of knowledge can be used to form plans
against the opening and playing styles of the adversary. The inclusion of the global
cultural knowledge and domain specific knowledge would enable the adversary model
to suggest move sequences that would be strategically advantageous. The ability to
make move suggestions would have a further impact on search complexity since both
the MIN, which is already affected, and the MAX node levels would be coached by
the adversary model.
Table 13: Statistical Analysis of Soviet Chess Weaknesses.
STRATEGIC
KNOWLEDGE
GAMES LOST FROM
EFFECT OF KNOWLEDGE
PERCENTAGE
Lose more with Black.
39 of 68
57%
Weak middle-game.
44 of 68
65%
More errors made from
defending than attacking
positions.
52 of 68
76%
Suffer from time pressure.
18 of 68
26%
More errors due to
strategy than tactics.
41 of 66
62%
Our adversary modeling methodology successfully predicts many of the moves
that an adversary will make. The predictions are based on domain independent
knowledge and demonstrate that the use of domain independent knowledge as well
as domain dependent knowledge are the critical elements in continuing to improve
adversarial domain application program performance. The current model focuses

131
primarily on the cognitive and perceptual processes used by an adversary to evaluate
the domain. The textual and visual chunks, which form the foundations of the
evaluation mechanisms employed by experts, are the domain independent knowledge
that enables the adversary model’s predictions. By further increasing the knowledge
contained in the adversary model with domain dependent and global knowledge, the
adversary model will be able to suggest strategically advantageous moves as well as
predicting likely adversary moves. The use of the adversary model as a coach will
greatly enhance the performance capabilities of existing adversarial domain planning
programs.

APPENDIX
The following glossary of chess and artificial intelligence terms is meant to
assist the reader in understanding the examples and explanations of this dissertation.
alpha-beta pruning:
algebraic notation:
A method of reducing the complexity of search trees
which utilize minimax processing. The general idea is
that once a value has been established at the MIN level
of a search tree, then subsequent values of other
subtrees to be backed up to the current MIN level which
are less than the established value will force all other
branches of that subtree to be pruned, or alpha pruned.
Likewise, once a value has been established for a MAX
level of the tree then any subsequent subtrees which
have a value greater than the current MAX level value
will force the remaining branches of that subtree to be
beta pruned.
A method for describing the moves of a chess game and
the position of the pieces. Each column of the chess
board is designated by a letter from ‘a’ to ‘h’ and each
row is designated by a number from 1 to 8, as shown
below.
a8
b8
c8
d8
e8
f8
g8
h8
a7
b7
c7
d7
e7
f7
g7
h7
a6
b6
c6
d6
e6
f6
g6
h6
a5
b5
c5
d5
e5
£5
g5
h5
a4
b4
c4
d4
e4
f4
g4
h4
a3
b3
c3
d3
e3
f3
g3
h3
a2
b2
c2
d2
e2
£2
g2
h2
al
bl
cl
dl
el
fl
gl
hi
132

133
Moves are described by giving one of the letters N, B, R,
Q, or K to designate the piece that is being moved (no
letter translates to a pawn move) followed by the letter
and number of the destination square. For example, Nf3
describes the move of a knight to the king’s-bishop-three
square. The one row is always white’s home row and
correspondingly the eight row is always black’s home
row.
en prise:
A piece is "en prise" if it can be captured by the
opponent on the next move.
evaluation function:
Evaluation functions are used by chess games to
determine the value, or goodness, of a specific board
position. Typical evaluation functions use domain
dependent tactical and positional knowledge, such as
material balance and control of the center squares of the
game board, in determining a position’s value.
fork:
When a knight has the positional ability to capture two
different pieces.
heuristic:
A "rule of thumb" which will produce a solution to a
problem. However, the solution produced by the
heuristic is not guaranteed to be optimal.
major piece:
Any chess piece other than a pawn. A king, queen,
rook, bishop, or knight piece.
material balance:
minimax:
A number which represents the difference between the
number and quality of pieces currently on the board for
each side. Each piece has its own value with the value
of the stronger pieces being greater than the value of the
weaker pieces, such as the value nine for a queen and
the value three for a knight. The sum of the value for
each piece remaining on the board for each side is
obtained. Then the opponent’s material sum is
subtracted from your material sum to obtain the
material balance.
An algorithm that decides the optimal tree path to
follow by alternately backing up the minimum and
maximum of all leaf nodes. The MAX player tries to
obtain the maximum possible value of all branches.
minimax:

134
pawn structure:
pin:
ply:
quiescent:
skewer:
Any group of pawns of the same color which are either
diagonally or horizontally adjacent to each other.
When a queen or rook is attacking a piece which lies
between the queen or rook and another piece of greater
value than the intervening piece.
A ply refers to a single level of a game tree or one half
of a game turn. Each game turn consists of one move
by each of the players involved. A ply is therefore, the
move of one of the two players.
A board position in which no captures can take place
and the king is not in check. Quiescence is also used to
refer to board positions in which a capture can occur,
but no subsequent captures are possible.
Similar to a pin except the threatening piece is attacking
along a diagonal (i.e. a bishop).

REFERENCES
Adelson-Velsky, G. M., Arlazarov, V. L., & Donsky, M. V. (1975). Some Methods
of Controlling the Tree Search in Chess Programs. Artificial Intelligence.
6141. 361-371.
Adelson-Velsky, G. M., Arlazarov, V. L., & Donsky, M. V. (1988). Algorithms for
Games. Springer-Verlag, New York.
Aho, A. V., Hopcroft, J. E., & Ullman, J. D. (1974). The Design and Analysis of
Computer Algorithms. Addison-Wesley, Reading, MA.
Aliston, W., & Weiskopf, D. (1984). The Complete Baseball Handbook. Allyn and
Bacon, Boston.
Anderson, J. R. (1980). Cognitive Psychology and Its Implications. W. H. Freeman,
New York.
Anderson, J. R. (1987). Causal Analysis and Inductive Learning. In Proceedings of
the Fourth International Workshop on Machine Learning. Irvine, CA, 288-299.
Andriole, S. J., Black, H. H., & Hopple, G. W. (1986). Intelligent Aids for Tactical
Planning. IEEE Transactions on Systems. Man, and Cybernetics. SMC-16(6),
854-864.
Angluin, D., & Smith, C. H. (1983). Inductive Inference: Theory and Methods.
Computing Surveys. 15(3). 237-269.
Barr, A., & Feigenbaum, E. A. (Eds.). (1981). The Handbook of Artificial
Intelligence Volume 1. Addison-Wesley, Reading, MA.
Bell, A. G. (1978). The Machine Plavs Chess? Pergamon Press, New York.
Berliner, H. (1979). On Construction of Evaluation Functions for Large Domains.
In Proceedings Sixth International Joint Conference on Artificial Intelligence.
Tokyo, Japan, 53-55.
Berliner, H. (1988). HITECH Becomes First Computer Senior Master. AI
Magazine. 9(3). 85-87.
135

136
Berliner, H., & Ebeling, C. (1988). Pattern Knowledge and Search: The SUPREM
Architecture. Technical Report CMU-CS-88-109, Carnegie-Mellon University,
Pittsburgh.
Berliner, H., & Goetsch, G. (1984). A Quantitative Study of Search Methods and
the Effect of Constraint Satisfaction. Technical Report CMU-CS-84-147,
Carnegie-Mellon University, Pittsburgh.
Blum, L., & M. Blum, M. (1975). Toward a Mathematical Theory of Inductive
Inference. Information and Control. 28(21. 125-155.
Bole, L. (Ed.). (1987). Computational Models of Learning. Springer-Verlag,
New York.
Bonasso, R. P. (1988). What AI Can Do for Battle Management. AI Magazine,
9(31. 77-83.
Bond, M. H. (Ed.). (1986). The Psychology of the Chinese People. Oxford
University Press, Oxford, UK.
Boring, E. G. (Ed.). (1945). Psychology for the Armed Services. The Infantry
Journal, Washington, DC.
Bratko, I., & Michie, D. (1980). An Advice Program for a Complex Chess
Programming Task. The Computer Journal. 23(41. 353-359.
Bratko, I., Tancig, P„ & Tancig, S. (1986). Detection of Positional Patterns in Chess.
In Advances in Computer Chess 4. 113-126. Pergamon Press, New York.
Campbell, M. S. (1988). Chunking as an Abstraction Mechanism. PhD thesis,
Carnegie-Mellon University, Pittsburgh.
Candland, D. K. (1980). Speaking Words and Doing Deeds. American Psychologist,
35(21. 191-198.
Carbonell, J. G. (1981). Counterplanning: A Strategy-Based Model of Adversary
Planning in Real-World Situations. Artificial Intelligence. 16(31, 295-329.
Case J., & Smith, C. (1983). Comparison of Identification Criteria for Machine
Inductive Inference. Theoretical Computer Science. 25(21. 193-220.
Chase, W. G., & Simon, H. A. (1973). Perception in Chess. Cognitive Psychology,
4(11. 55-81.

137
Chase, W. G., & Simon, H. A. (1988). The Mind’s Eye in Chess. In Readings in
Cognitive Science. 461-494. Morgan Kaufmann, San Mateo, CA.
Christensen, J., & Korf, R. E. (1986). A Unified Theory of Heuristic Evaluation
Functions and Its Application to Learning. In Proceedings AAAI-86 The
Fifth National Conference on Artificial Intelligence. Philadelphia, 148-152.
Cimbala, S. J. (1987). Artificial Intelligence and National Security. Lexington Books,
Lexington, MA.
Crookall, D., Greenblat, C. S., Coote, A., Klabbers, J., Watson, D. R. (Eds.). (1987).
Simulation-Gaming in the Late 1980’s. Pergamon Press, New York.
Daley, R. P., & Smith, C. H. (1986). On the Complexity of Inductive Inference.
Information and Control. 69(1-3). 12-40.
Davis, B. (1956). Gray Fox Robert E. Lee and the Civil War. Fairfax Press, Lehigh
Valley, PA.
Davis, P. K. (1988a). Applying Artificial Intelligence Techniques to Strategic-Level
Gaming and Simulation. Technical Report N-2752-RC, Rand Strategy
Assessment Center, Santa Monica, CA.
Davis, P. K. (1988b). A new analytic technique for the study of deterrence,
escalation control, and war termination. Simulation. 50(5). 195-202.
Davis, P. K., Bankes, S. C., & Kahan, J. P. (1986). A New Methodology for
Modeling National Command Level Decisionmaking in War Games and
Simulations. Technical Report R-3290-NA, Rand Strategy Assessment Center,
Santa Monica, CA.
DeJong, K. A., & Schultz, A. C. (1988). Using Experience-Based Learning in Game
Playing. In Proceedings of the Fifth International Conference on Machine
Learning. Ann Arbor, MI, 284-290.
Department of the Army (1986). Field Manual 7-7J The Mechanized Infantry
Platoon and Squad (Bradley). Washington, DC.
Dunnigan, J. F. (1982). How To Make War. William Morrow, New York.
Ebeling, C. (1986). All the Right Moves: A VLSI Architecture for Chess.
PhD thesis, Carnegie-Mellon University, Pittsburgh.

138
Elithorn, A., & Banerji, R. (Eds.). (1984). Artificial and Human Intelligence.
Elsevier, New York.
Erickson, M. D., & Zytkow, J. M. (1988). Utilizing Experience for Improving the
Tactical Manager. In Proceedings of the Fifth International Conference on
Machine Learning. Ann Arbor, MI, 444-450.
Ermarth, F. W. (1978). Contrasts in American and Soviet Strategic Thought.
International Security. 3(21. 138-155.
Evans, L. (1970). Chess Catechism. Simon and Schuster, New York.
Findler, N. V., & Meltzer, B. (1971). Artificial Intelligence and Heuristic
Programming. Elsevier, New York.
Frey, P. W. (1983). Chess Skill in Man and Machine. Springer-Verlag, New York.
Geitner, P. (1989). Computer is Pawn to Chess Master. The Gainesville Sun.
114(1091. 1A,6A.
Gelo, J. H. (1988). Chess World Championships All the Games 1834-1984.
McFarland and Company, Jefferson, NC.
Gruner, W. P. (1988). Intelligence Information for Submarines. The Submarine
Review. (Jan). 21-28.
Hamilton, J. (1990). Kasparov’s Visit to Boston. Chess Life. 45(1). 30-31,92.
Holding, D. H. (1985). The Psychology of Chess Skill. Lawrence Erlbaum
Associates, Hillsdale, NJ.
Horgan, D. D„ Millis, K., & Neimeyer, R. A. (1989). Cognitive Reorganization and
the Development of Chess Expertise. International Journal of Personal
Construct Psychology. 2, 15-36.
Horowitz, A. (1973). The World Chess Championship A History. Macmillan,
New York.
Kodratoff, Y., & Michalski, R. S. (Eds.). (1990). Machine Learning An Artificial
Intelligence Approach Volume III. Morgan Kaufmann, San Mateo, CA.
Kopec, D., & Newborn, M. (1987). Belle and Mephisto Dallas Capture Computer
Chess Titles at FJCC. Communications of the ACM. 30(7). 640-645.

139
Kotov, A., & Yudovich, M. (1961). The Soviet School of Chess. Dover Publications,
New York.
Laird,J., Rosenbloom, P., & Newell, A. (1986). Universal Subgoaling and Chunking.
Kluwer Academic, Norwell, MA.
Langer, W. C. (1972). The Mind of Adolf Hitler. Basic Books, New York.
Langley, P. (1985). Learning to Search: From Weak Methods to Domain Specific
Heuristics. Cognitive Science. 9(2). 217-260.
Lasker, E. (1973). Modern Chess Strategy. David McKay Company, New York.
Lee, K., & Mahajan, S. (1988). A Pattern Classification Approach to Evaluation
Function Learning. Artificial Intelligence. 36(1). 1-25.
Lenat, D. B. (1982). The Nature of Heuristics. Artificial Intelligence. 19(2).
189-249.
Lenat, D. B., & Feigenbaum, E. A. (1987). On the Thresholds of Knowledge.
Technical Report AI-126-87, Microelectronics and Computer Technology
Corporation (MCC), Austin, TX.
Lenat, D. B., Hayes-Roth, F., & Klahr, P. (1979). Cognitive Economy in Artificial
Intelligence Systems. In Proceedings of the Sixth International Joint
Conference on Artificial Intelligence. Tokyo, Japan, 531-536.
Levy, D. (1984). The Joy of Computer Chess. Prentice Hall, Englewood Cliffs, NJ.
Levy, D. (Ed.). (1988). Computer Games I. Springer-Verlag, New York.
McMillen, D. H. (Ed.). (1984). Asian Perspectives on International Security. St.
Martin’s Press, New York.
Mednis, E. (1978). How to Beat the Russians. David McKay Company, New York.
Mednis, E. (1990). Move Orders in the Opening: The Modern Master’s Tool. Chess
Life. 45(6). 14-16.
Michalski, R. S., Carbonell, J. G., & Mitchell, T. M. (Eds.). (1983). Machine
Learning An Artificial Intelligence Approach. Morgan Kaufmann, San Mateo,
CA.

140
Michalski, R. S., Carbonell, J. G., «fe Mitchell, T. M. (Eds.). (1986). Machine
Learning An Artificial Intelligence Approach Volume II. Morgan Kaufmann,
San Mateo, CA.
Miller, G. A. (1956). The Magical Number Seven, Plus or Minus Two: Some Limits
on Our Capacity for Processing Information. The Psychological Review.
63(21. 81-97.
Mitchell, T. M., Carbonell, J. G., Learning A Guide to Current Research. Kluwer Academic, Norwell, MA.
Muggleton, S. (1990). Inductive Acquisition of Expert Knowledge. Addison-Wesley,
Reading, MA.
Narendra, K„ «fe Thathachar, M. (1989). Learning Automata: An Introduction.
Prentice Hall, Englewood Cliffs, NJ.
Newborn, M., Computer Chess Championship. Communications of the ACM. 32(10).
1225-1230.
Newborn, M., «fe Kopec, D. (1990). The Twentieth Annual ACM North American
Computer Chess Championship. Communications of the ACM. 33(7). 92-104.
Nunn, J., New York.
Pachman, L. (1975). Complete Chess Strategy: First Principles of the Middle Game.
Doubleday, New York.
Polya, G. (1973). How To Solve It. Princeton University Press, Princeton, NJ.
Pomerantz, J. R. (1986). Visual Form Perception: An Overview. In Pattern
Recognition by Humans and Machines Volume 2 Visual Perception. 1-30.
Academic Press, Orlando, FL.
Pritchard, D. B. (1973). Go: A Guide to the Game. Faber and Faber, Winchester,
MA.
Reiss Games (1974). Go: The Ancient Oriental Game. New York.
Riggins, J., Football. Santa Barbara Press, Santa Barbara, CA.

141
Ritter, G. X., Wilson, J. N., & Davidson, J. L. (1988). Image Algebra: An Overview.
Technical Report TR-88-05, University of Florida, Gainesville, FL.
Robbins, J. (1988). America’s Red Army. The New York Times Magazine. (Apr).
38-46.
Rosch, E. H. (1973). Natural Categories. Cognitive Psychology. 4(3). 328-350.
Rosenbloom, P. S. (1982). A World-Championship-Level Othello Program. Artificial
Intelligence. 19(3). 279-320.
Roycroft, A. J. (1988). Expert Against Oracle. In Machine Intelligence 11. 347-373.
Oxford University Press, Oxford, UK.
Ryan, P. J. (1988). Some Commandments of Submarine Warfare. The Submarine
Review. (Jan). 60-63.
Saariluoma, P. (1984). Coding Problem Spaces in Chess A Psychological Study.
Societas Scientiarum Fennica, Helsinki, Finland.
Samuel, A. L. (1959). Some Studies in Machine Learning Using the Game of
Checkers. IBM Journal of Research and Development. 3(3). 210-229.
Samuel, A. L. (1967). Some Studies in Machine Learning Using the Game of
Checkers II - Recent Progress. IBM Journal of Research and Development.
11(6). 601-617.
Schonberg, H. C. (1973). Grandmasters of Chess. J. B. Lippincott Company,
Philadelphia.
Shannon, C. E. (1950). Programming a Computer for Playing Chess. The
Philosophical Magazine. 41(314). 256-275.
Shapiro, A. D. (1987). Structured Induction in Expert Systems. Addison-Wesley,
Reading, MA.
Simon, H. A., & Gilmartin, K. (1973). A Simulation of Memory for Chess Positions.
Cognitive Psychology. 5(1). 29-46.
Sleeman, D„ & Brown, J. S. (1982). Intelligent Tutoring Systems. Academic Press,
Orlando, FL.
Snow, W. P. (1867). Lee and His Generals. Richardson and Company, New York.

142
Sokolovskiy, V. D. (1968). Soviet Military Strategy. Macdonald and Jane’s, Stanford,
CA.
Soltis, A. (1976). Pawn Structure Chess. David McKay Company, New York.
Tadepalli, P. (1989). Knowledge Based Planning in Games. Technical Report
CMU-TR-89-135, Carnegie-Mellon University, Pittsburgh.
Utgoff, P. E. (1986). Machine Learning of Inductive Bias. Kluwer Academic,
Norwell, MA.
Utgoff, P. E., & Saxena, S. (1987). Learning a Preference Predicate. In Proceedings
of the Fourth International Workshop on Machine Learning. Irvine, CA,
115-121.
Von Neumann, J., & Morgenstern, O. (1953). Theory of Games and Economic
Behavior. Princeton University Press, Princeton, NJ.
Wallace, J. (1990). Bloody Chosin: The Blind Lead the Brave. U. S. News & World
Report. 108(251. 37-43.
Waterman, D. A. (1970). Generalization Learning Techniques for Automating the
Learning of Heuristics. Artificial Intelligence. 1(1-2). 121-170.
White, J. F. (1990). The Amateurs’ Book Opening Routine. ICCA Journal. 13(1),
22-26.
Wilcox, B. (1985). Reflections on Building Two Go Programs. SIGART Newsletter.
(94). 29-43.
Wilkins, D. E. (1980). Using Patterns and Plans in Chess. Artificial Intelligence. 14,
165-203.
Wilkins, D. E. (1982). Using Knowledge to Control Tree Searching. Artificial
Intelligence. 18(1). 1-51.
Witkin, G. (1990). New Drug Warriors: Lasers, Labs, and Coke-Eating Bugs. U. S.
News & World Report. 108(7). 20-21.
Woolf, B. P. (1984). Context Dependent Planning In A Machine Tutor. PhD thesis,
University of Massachusetts, Amherst, MA.

143
Young, P. R., & Lehner, P. E. (1986). Applications of a Theory of Automated
Adversarial Planning to Command and Control. IEEE Transactions on
Systems. Man, and Cybernetics. SMC-16(6). 806-812.
Zagare, F. C. (1984). Game Theory Concepts and Applications. Sage Publications,
Newbury Park, CA.

BIOGRAPHICAL SKETCH
Steven Michael Walczak received his Bachelor of Science degree with a major
field of study in mathematics from the Pennsylvania State University in 1981.
Following his undergraduate education, Mr. Walczak was employed by the
Department of Defense where he was responsible for the design and implementation
of various software projects including a signal analysis expert system. While working
for the government, he acquired his Master of Science in computer science from the
Johns Hopkins University. In 1986, Mr. Walczak moved to Florida and began
working for two different government contractors. His primary responsibilities were
the development of an automated knowledge acquisition system, the development of
an intelligent tutoring system, and the design of a message forwarding system.
Mr. Walczak returned to academia in 1988 as a full-time student working for his
Doctor of Philosophy degree in computer and information sciences from the
University of Florida.
144

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and
quality, as a dissertation for the degree of Doctor of Philosophy.
Paul A Fishwick, Chairman
Assistant Professor of Computer and
Information Sciences
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and
quality, as a dissertation for the degree of Doctor of Philosophy.
t
7
Manuel E. Bermudez
Associate Professor of Computer and
Information Sciences
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and
quality, as a dissertation for the degree of Doctor of Philosophy.
Douglas D. Dankel II
Assistant Professor of Computer and
Information Sciences
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and
quality, as a dissertation for the degree of Doctor of Philosophy.
// y/ „/? . "
Walter R. Cunningharrj/
Professor of Psychology

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and
quality, as a dissertation for the degree of Doctor of Philosophy. /
r.
Zjl.
JL
// -
If'a S. Fischler
Professor of Psychology
This dissertation was submitted to the Graduate Faculty of the College of
Engineering and to the Graduate School and was accepted as partial fulfillment of
the requirements for the degree of Doctor of Philosophy.
December 1990
— Winfred M. Phillips
Dean, College of Engineering
Madelyn M. Lockhart
Dean, Graduate School

UNIVERSITY OF FLORIDA
3 1262 08553 8451




xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ENUY1NVE2_F3F8WW INGEST_TIME 2017-07-13T21:25:28Z PACKAGE AA00003762_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 2

86,1* ,1'8&7,9( ,1)(5(1&( 2) 3$67 3(5)250$1&( 72 %8,/' 675$7(*,& &2*1,7,9( $'9(56$5< 02'(/6 %< 67(9(1 0,&+$(/ :$/&=$. $ ',66(57$7,21 35(6(17(' 72 7+( *5$'8$7( 6&+22/ 2) 7+( 81,9(56,7< 2) )/25,'$ ,1 3$57,$/ )8/),//0(17 2) 7+( 5(48,5(0(176 )25 7+( '(*5(( 2) '2&725 2) 3+,/2623+< 81,9(56,7< 2) )/25,'$

PAGE 3

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

PAGE 4

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

PAGE 5

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

PAGE 6

&RPSXWHU &KHVV DQG $UWLILFLDO ,QWHOOLJHQFH -XVWLILFDWLRQ IRU &KHVV 'RPDLQ $SSOLFDWLRQ $'9(56$5< 02'(/,1* 0(7+2'2/2*< .QRZOHGJH $FTXLVLWLRQ RI &KXQNV 'HILQLQJ WKH &KXQNV WR EH $FTXLUHG $FTXLULQJ *HRPHWULF &KXQNV $FTXLULQJ 7H[WXDO &KXQNV ,QGXFWLRQ WR /HDUQ 5HSHDWHG &KXQN 3DWWHUQV $FTXLULQJ DQ $GYHUVDU\fV 3OD\LQJ 6W\OH $SSOLFDWLRQ RI WKH $GYHUVDU\ 0RGHO ,$0 2YHUYLHZ RI ,$0fV ([HFXWLRQ .QRZOHGJH $FTXLVLWLRQ DQG ,QGXFWLYH /HDUQLQJ 6WDJH 3UHGLFWLQJ $GYHUVDU\ $FWLRQV 6WDJH ,QWHJUDWLQJ ,$0fV .QRZOHGJH LQWR D &KHVV 3URJUDP ,$0f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

PAGE 7

5()(5(1&(6 %,2*5$3+,&$/ 6.(7&+ YL

PAGE 8

/,67 2) 7$%/(6 ,$0 3HUIRUPDQFH 0HDVXUHV IRU 7(67 (IIHFW RI /LNHOLKRRG RQ 3UHGLFWLRQ 5DWLRV ,$0 3HUIRUPDQFH 0HDVXUHV IRU 7(67 3HUIRUPDQFH ZLWK 5HVSRQVLYH 0RGLILFDWLRQ (IIHFW RI 5HFHQF\ RQ ,$0 3HUIRUPDQFH IRU 7(67 (IIHFW RI $JLQJ)RUJHWWLQJ RQ ,$0fV 3HUIRUPDQFH 3HUIRUPDQFH 7(67f ZLWK 2SHQLQJ 0RYHV 3HUIRUPDQFH 7(67f ZLWK 2SHQLQJ 0RYHV ,$0 3HUIRUPDQFH 7(67f ZLWK )LYHE\)LYH &KXQNV ,$0 3HUIRUPDQFH 7(67f ZLWK 6L[E\6L[ &KXQNV (IIHFW RI 'RPDLQ 6SHFLILF .QRZOHGJH 5DWLR RI :RQ DQG /RVW *DPHV IRU :KLWH DQG %ODFN ,OO 6WDWLVWLFDO $QDO\VLV RI 6RYLHW &KHVV :HDNQHVVHV 9OO

PAGE 9

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

PAGE 10

$Q H[DPSOH RI D JDPH UHFRUGWKH +DJXH0RVFRZ 7RXUQDPHQW &KXQN RFFXUUHQFH LQ WKUHH GLIIHUHQW JDPHV 3VHXGRFRGH WR SHUIRUP LQGXFWLRQ RQ JHRPHWULF FKXQNV 3UHGLFWLRQV IURP IRXU FKXQNV $ JDPH IURP WKH %RWYLQQLN YHUVXV 3HWURVLDQ PDWFK $QRWKHU JDPH IURP WKH %RWYLQQLN YHUVXV 3HWURVLDQ PDWFK 3VHXGRFRGH WR VXJJHVW SRVVLEOH DGYHUVDU\ PRYHV &KXQN DFTXLUHG ZLWK [ VL]H DQG FRUUHVSRQGLQJ VPDOOHU FKXQN +\SRWKHWLFDO JDPHVHDUFK WUHH 3UHGLFWLRQ RI %RWYLQQLNfV QH[W PRYH 6WDQGDUG VHDUFK WUHH DQG WKH ,$0 LPSURYHG VHDUFK WUHH 6HDUFK WUHH REWDLQHG IURP &RUUHFW 3LHFH LGHQWLILFDWLRQ .QLJKW WHPSODWH DQG SDUWLDO 5RRN WHPSODWH ,;

PAGE 11

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fV HYDOXDWLRQ FULWHULD DQG REMHFWLYHV LV GLIILFXOW WR REWDLQ VLQFH DQ DGYHUVDU\ ZLOO JXDUG WKLV LQIRUPDWLRQ LQ RUGHU WR SURPRWH D IDYRUDEOH RXWFRPH IRU KLPVHOI :H SUHVHQW DQ DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ ZKLFK LV FDSDEOH RI DFTXLULQJ WKH HYDOXDWLRQ FULWHULD RI DQ DGYHUVDU\ WKURXJK REVHUYDWLRQV RI WKH DGYHUVDU\fV SDVW SHUIRUPDQFHV LQ WKH GRPDLQ 2XU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ XVHV LQGXFWLRQ WR FDSWXUH SHUFHSWXDO FKXQNV WKDW DUH UHSHDWHGO\ GLVSOD\HG E\ D VSHFLILF DGYHUVDU\ 7KH DFTXLUHG FKXQNV DUH WKHQ XVHG WR SUHGLFW WKH DGYHUVDU\fV DFWLRQV GXULQJ IXWXUH HQFRXQWHUV 3DWWHUQ UHFRJQLWLRQ LV XVHG WR LGHQWLI\ SHUFHSWXDO FKXQNV LQ WKH FXUUHQW DGYHUVDULDO GRPDLQ [

PAGE 12

VLWXDWLRQ ZKLFK DUH DQDORJRXV WR FKXQNV WKDW KDYH EHHQ SUHYLRXVO\ OHDUQHG IURP WKH FXUUHQW DGYHUVDU\ :H KDYH LPSOHPHQWHG RXU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ LQ WKH SURJUDP ,$0 ,QGXFWLYH $GYHUVDU\ 0RGHOHU ZKLFK KDV EHHQ DSSOLHG WR WKH GRPDLQ RI FKHVV :KHQ KLJK TXDOLW\ NQRZOHGJH LV DYDLODEOH DERXW DQ DGYHUVDU\fV SDVW SHUIRUPDQFHV WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ DFFXUDWHO\ SUHGLFWV WHQ SHUFHQW RI DQ DGYHUVDU\fV IXWXUH DFWLRQ SODQV 7KH DELOLW\ WR DFFXUDWHO\ SUHGLFW DQ DGYHUVDU\fV PRYH FKRLFHV UHGXFHV WKH FRPSOH[LW\ RI WKH JDPH WUHH VHDUFK ZKLFK HQKDQFHV WKH SOD\LQJ DELOLW\ RI FXUUHQW FKHVV SURJUDPV [L

PAGE 13

&+$37(5 ,1752'8&7,21 $GYHUVDULDO VLWXDWLRQV VXUURXQG XV LQ YDULHG IRUPV RQ D GDLO\ EDVLV ([DPSOHV RI DGYHUVDULDO VLWXDWLRQV LQFOXGH Y\LQJ IRU D SURPRWLRQ DW RXU SODFH RI HPSOR\PHQW 0RQGD\ QLJKWf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fV SUHYLRXV SHUIRUPDQFH SUHIHUDEO\ LQ VLWXDWLRQV DQDORJRXV WR WKH XSFRPLQJ FRQWHVW +XPDQ FRPSHWLWRUV SHUIRUPLQJ LQ DGYHUVDULDO GRPDLQV VWXG\ WKH SDVW SHUIRUPDQFHV RI WKHLU DGYHUVDULHV WR LGHQWLI\ VWUHQJWKV DQG ZHDNQHVVHV ZKLFK FDQ EH

PAGE 14

H[SORLWHG UHVSHFWLYHO\ WR JDLQ D VWUDWHJLF RU WDFWLFDO DGYDQWDJH 7KLV LQIRUPDWLRQ LV SUHVHUYHG E\ WKH FRPSHWLWRU DV D PRGHO RI WKH OLNHO\ DFWLRQV DQ DGYHUVDU\ ZLOO SHUIRUP LQ SDUWLFXODU VLWXDWLRQV 7KH PRGHO RI WKH DGYHUVDU\fV SUREDEOH DFWLRQV LV WKHQ H[SORLWHG ZKHQ DQ DSSURSULDWH VLWXDWLRQ RFFXUV GXULQJ WKH FRXUVH RI WKH FRPSHWLWLRQ 6XFK VLWXDWLRQV DULVH ZKHQ WKH FXUUHQW JDPH VWDWHVLWXDWLRQ LQ WKH DGYHUVDULDO GRPDLQLV DQDORJRXV WR D JDPH VWDWH ZKLFK RFFXUUHG LQ WKH DGYHUVDU\fV SDVW 3UREOHP 6WDWHPHQW &XUUHQW DUWLILFLDO LQWHOOLJHQFH DSSOLFDWLRQ HIIRUWV LQ DGYHUVDULDO GRPDLQV DUH QRW SHUIRUPHG RSWLPDOO\ $ IDFWRU ZKLFK NHHSV DGYHUVDU\EDVHG SURJUDPV IURP SHUIRUPLQJ DW H[SHUW OHYHOV LV WKH ODFN RI NQRZOHGJH DERXW WKH SRWHQWLDO DGYHUVDU\ /HQDW DQG )HLJHQEDXP f FODLP WKDW GRPDLQ VSHFLILF NQRZOHGJH LV UHTXLUHG WR VROYH GLIILFXOW SUREOHPV DQG WKDW WKH SHUIRUPDQFH RI LQWHOOLJHQW SURJUDPV LPSURYHV DV DGGLWLRQDO GRPDLQ NQRZOHGJH LV XWLOL]HG $FTXLULQJ NQRZOHGJH IRU SURJUDPV LQ DGYHUVDULDO GRPDLQV LV KLQGHUHG E\ WKH IDFW WKDW DGYHUVDULHV DUH XQZLOOLQJ WR UHYHDO WKHLU WDFWLFDO DQG VWUDWHJLF JRDOV VLQFH WKLV ZRXOG EH FRXQWHUSURGXFWLYH WR WKH DWWDLQPHQW RI WKHLU JRDOV &DQGODQG f DQG %ROH f KDYH FODLPHG WKDW NQRZOHGJH DERXW DQ\ GRPDLQ LV DYDLODEOH IURP D ZLGH YDULHW\ RI VRXUFHV /HQDW f SURSRVHV WKH XVH RI WKH REVHUYDEOH DFWLRQV RI WKH DGYHUVDU\ ZKLOH SHUIRUPLQJ LQ WKH DGYHUVDULDO GRPDLQ 7KH SURSHU VHOHFWLRQ RI DGYHUVDULDO GRPDLQ VLWXDWLRQV WR REVHUYH JXDUDQWHHV WKH DFTXLVLWLRQ RI KLJK TXDOLW\ NQRZOHGJH

PAGE 15

)RU WKH GRPDLQ RI FKHVV ODFN RI NQRZOHGJH DERXW DQ DGYHUVDU\f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fV WDFWLFV DQG VWUDWHJLHV WR FRQWLQXH WR LPSURYH WKHLU SHUIRUPDQFH :H QHHG WR FRQVLGHU WZR VHSDUDWH IDFWRUV WKDW DIIHFW WKH VWUDWHJLF GHFLVLRQV DQ DGYHUVDU\ ZLOO PDNH $Q DGYHUVDU\f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

PAGE 16

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

PAGE 17

DQG 3RPHUDQW] f WR HPXODWH WKH SULPDU\ HYDOXDWLRQ FULWHULD XVHG E\ DQ DGYHUVDU\ .RGUDWRII DQG 0LFKDOVNL f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f :H DFTXLUH WKH FRJQLWLYH FKXQNV XVHG E\ DQ DGYHUVDU\ WR HYDOXDWH WKH FXUUHQW GRPDLQ VLWXDWLRQ 7KH DFTXLUHG FRJQLWLYH FKXQNV RI WKH DGYHUVDU\ KHOS XV WR YLHZ WKH GRPDLQ WKH VDPH DV RXU DGYHUVDU\ DQG WR DYRLG LQIOXHQFLQJ GRPDLQ HYDOXDWLRQV ZLWK SHUVRQDO ELDVHV :H QRZ KDYH D ODUJH FROOHFWLRQ RI FKXQNV WKDW KDYH EHHQ GLVSOD\HG E\ WKH DGYHUVDU\ 7KH FKXQNV DUH HLWKHU WH[WXDO SDWWHUQV FRUUHVSRQGLQJ WR WKH DGYHUVDU\fV YHUEDO PHPRU\ RU JHRPHWULF SDWWHUQV FRUUHVSRQGLQJ WR WKH DGYHUVDU\fV YLVXDO PHPRU\ 7KH FKXQNV ZH KDYH DFTXLUHG FRQWDLQ SDWWHUQV WKDW KDYH EHHQ XVHG RQO\ RQFH DQG PD\ KDYH RFFXUUHG E\ FKDQFH :H XVH LQGXFWLRQ WR HOLPLQDWH WKHVH FKDQFH FKXQNV DQG FUHDWH D YDOLG DQG XVHIXO DGYHUVDU\ PRGHO 7KH LQGXFWLYH SKDVH LGHQWLILHV FKXQNV ZKLFK DUH UHSHDWHG LQ WZR RU PRUH JDPHV 7KH UHSHWLWLRQ RI SDWWHUQV LV XVHG WR LQGLFDWH WKH SUREDELOLW\ WKDW DQ DGYHUVDU\ ZLOO XVH WKH DFTXLUHG FKXQN LQ IXWXUH

PAGE 18

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f QRWH WKDW VSHHG LV D FULWLFDO IDFWRU IRU HYDOXDWLRQ IXQFWLRQV LQ FKHVV DV ZHOO DV RWKHU DGYHUVDULDO GRPDLQV 7KH NQRZOHGJH WKDW ZH DUH OHDUQLQJ DERXW DQ DGYHUVDU\ LV DYDLODEOH DV KLVWRULFDO UHFRUGV SULRU WR WKH FXUUHQW SHUIRUPDQFH RI WKH DGYHUVDULDO GRPDLQ SURJUDP 7KHUHIRUH WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\fV NQRZOHGJH DFTXLVLWLRQ FDQ EH SHUIRUPHG SULRU WR LWV XVH E\ WKH GRPDLQ SURJUDP 7KLV SUHSURFHVVLQJ HIIHFWLYHO\ HOLPLQDWHV DQ\ WLPH RU FRPSOH[LW\ FRVWV

PAGE 19

JHQHUDWHG E\ WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ IURP EHLQJ DGGHG WR WKH FXUUHQW SURJUDPfV HYDOXDWLRQ IXQFWLRQ FRVW $KR f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

PAGE 20

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

PAGE 21

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f GHVFULEHV WKH ZD\ KXPDQV IRUP FRQFHSWV DQG FDWHJRULHV DERXW WKHLU QDWXUDO ZRUOG 3RO\D f LQ WKH HDUO\ V IRUPDOL]HG WKH FRQFHSW RI LQGXFWLRQ ,QGXFWLRQ LV WKH SURFHVV RI GLVFRYHULQJ JHQHUDO ODZV RU FRQFHSWV IURP REVHUYDWLRQV DQG FRPELQDWLRQV RI SDUWLFXODU LQVWDQFHV $Q H[DPSOH RI LQGXFWLRQ IRU D QDWXUDO ZRUOG FRQFHSW ZRXOG EH WKH JHQHUDOL]DWLRQ WKDW DOO UDYHQV DUH EODFN 7KLV JHQHUDOL]DWLRQ LV EDVHG RQ WKH REVHUYDWLRQ RI D OLPLWHG QXPEHU RI LQVWDQFHV RI D UDYHQ :LWKRXW VHHLQJ DOO WKH SRVVLEOH H[DPSOHV RU LQVWDQFHV RI UDYHQ ZH DUH FRQILGHQW WR LQGXFH WKDW WKH FRORU RI

PAGE 22

D UDYHQ LV WKH VDPH DV WKH FRORU RI WKH LQVWDQFHV ZH KDYH VHHQEODFN 7KH VWUHQJWK RI RXU EHOLHI LQFUHDVHV DV PRUH DQG PRUH H[DPSOHV RI EODFN UDYHQV DUH REVHUYHG +RZHYHU RQO\ RQH LQVWDQFH RI D QRQEODFN UDYHQ QHHGV WR EH LGHQWLILHG WR LQYDOLGDWH RXU LQGXFWLRQ )URP WKH DERYH H[DPSOH ZH FDQ VHH WKDW LQGXFWLYH LQIHUHQFH LV OHDUQLQJ LQ WKH OLPLW 'DOH\ t 6PLWK f :H QHHG WR VHH VXIILFLHQW REVHUYDWLRQV RI WKH ULJKW W\SH WR IRUP D FRUUHFW JHQHUDOL]HG FRQFHSW 6LQFH ZH FDQQRW JXDUDQWHH WKDW WKH ULJKW W\SH RI REVHUYDWLRQV DUH EHLQJ SURYLGHG WR DQ LQGXFWLYH LQIHUHQFH V\VWHP ZH PXVW UHO\ RQ WKH XVH RI D ODUJH QXPEHU RI H[DPSOHV RI FDWHJRU\ LQVWDQFHV WR DFTXLUH WKH GHVLUHG FRQFHSW 7KLV PHDQV WKDW WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ ZH KDYH GHYHORSHG VKRXOG SHUIRUP PRUH DFFXUDWHO\ DV ZH SURYLGH JUHDWHU QXPEHUV RI H[DPSOHV RI DQ DGYHUVDU\fV GRPDLQ SHUIRUPDQFH 6LQFH 3RO\DfV IRUPDOL]DWLRQ RI WKH FRQFHSW RI LQGXFWLRQ WKHRUHWLFDO UHVHDUFK RQ LQGXFWLRQ KDV SULPDULO\ IRFXVHG RQ WKH IRUPDWLRQ RI FRQFHSWV RU FODVVLILFDWLRQV $QJOXLQ DQG 6PLWK f DQG &DVH DQG 6PLWK f SUHVHQW ULJRURXV GLVFXVVLRQV RI WKH PDWKHPDWLFDO SULQFLSOHV RI LQGXFWLRQ 7KLV WKHRUHWLFDO EDFNJURXQG SURYLGHV D VROLG EDVH IRU WKH XVH RI LQGXFWLYH LQIHUHQFH DV D FRPSXWDWLRQDO WRRO IRU OHDUQLQJ FODVVHV RI REMHFWV RU LGHDV 7KH H[WUDFWLRQ RI JHQHUDO SULQFLSOHV IURP VSHFLILF H[DPSOHV LV D PDMRU UHVHDUFK JRDO LQ PDFKLQH OHDUQLQJ $QGHUVRQ f ,QGXFWLRQ KDV EHHQ DSSOLHG WR PDQ\ GRPDLQV ([DPSOHV RI DSSOLHG LQGXFWLYH LQIHUHQFH DOJRULWKPV DUH 0LFKDOVNLfV ,1'8&( DQG 4XLQODQfV ,' 0LFKDOVNL HW DO DQG f (DFK RI WKHVH

PAGE 23

H[DPSOH DSSOLFDWLRQV DWWHPSWV WR FODVVLI\ LQVWDQFHV RI D GRPDLQ SUREOHP HJ VR\EHDQ GLVHDVHV IRU WKH ,1'8&( SURJUDPf LQWR PD[LPDOO\VSHFLILF JURXSV WKDW ZLOO LGHQWLI\ DOO NQRZQ LQVWDQFHV DV ZHOO DV DFFRXQWLQJ IRU XQNQRZQ LQVWDQFHV 3ULRU WR WKH WKHRUHWLFDO GLVFXVVLRQV RI LQGXFWLYH LQIHUHQFH PHQWLRQHG DERYH %OXP DQG %OXP f KDG DOUHDG\ GHPRQVWUDWHG WKDW WKH DELOLW\ WR SOD\ JDPHV RU DW OHDVW WR PDNH WHFKQLFDOO\ FRUUHFW PRYHV FDQ EH OHDUQHG WKURXJK LQGXFWLYH LQIHUHQFH ,QGXFWLRQ KDV QRW EHHQ XWLOL]HG LQ JDPH GRPDLQV GXH WR WKH ODFN RI KLJK TXDOLW\ SOD\ SURGXFHG E\ %OXPVf DOJRULWKP 2XU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ IRFXVHV RQ D QDUURZHU VOLFH RI WKH JDPH GRPDLQV QDPHO\ SUHGLFWLQJ DGYHUVDU\ PRYH FKRLFHV DQG SOD\LQJ VW\OH DQG LV DEOH WR OHDUQ KLJK TXDOLW\ NQRZOHGJH $GYHUVDULDO GRPDLQV SUHVHQW VHYHUDO VSHFLDOL]HG SUREOHPV WR WKH VWDQGDUG LQGXFWLYH LQIHUHQFH DOJRULWKPV OLNH ,' /DQJOH\ f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

PAGE 24

LQGXFWLRQ PHWKRGV 6WUXFWXUH LQGXFWLRQ PHWKRGV DUH XVHG WR SURGXFH FODVVLILFDWLRQ V\VWHPV 0XJJOHWRQ f SURSRVHV DQ LQGXFWLYH PHWKRG FDOOHG VHTXHQFH LQGXFWLRQ ZKLFK KDV D WLPH HIIHFW UHODWLRQVKLS WR WKH GRPDLQ 6HTXHQFH LQGXFWLRQ XVHV D VHULHV RI GRPDLQ GHVFULSWLRQV WKDW DUH DOWHUHG E\ DFWLRQV RSHUDWLQJ RQ WKH GRPDLQ WR SURGXFH FRQWURO V\VWHPV 0LWFKHOO HW DO f UHIHU WR D SURFHVV RI PHFKDQL]HG LQGXFWLRQ ZKLFK UHFHLYHV GRPDLQ HYHQWV DV LQSXW DQG SURGXFHV FRQFHSWV IRU SUHGLFWLQJ IXWXUH HYHQWV )RU DGYHUVDULDO GRPDLQV VHTXHQFH LQGXFWLRQ SURYLGHV D PHWKRG WR SURGXFH HYDOXDWLRQ FULWHULD IRU SUHGLFWLQJ DQ DGYHUVDU\fV VWUDWHJLF GHFLVLRQV 7UHH 6HDUFK 3URJUDPV LQ WKH DGYHUVDULDO GRPDLQV RI JDPHV LQYDULDEO\ XVH WUHHV DV WKHLU NQRZOHGJH UHSUHVHQWDWLRQ VFKHPH )LQGLQJ VROXWLRQV RU PRYH FKRLFHV LQ JDPH WUHHV LQYROYHV VHDUFKLQJ WKH WUHH IRU WKH RSWLPDO ILQLVKLQJ YDOXH 9RQ 1HXPDQQ DQG 0RUJHQVWHUQ f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

PAGE 25

H[KDXVWLYH VHDUFK ,QVWHDG JDPH SURJUDPV ZLOO VHDUFK GRZQ WR D SUHVHOHFWHG GHSWK DQG DSSO\ DQ HYDOXDWLRQ IXQFWLRQ WKDW ZLOO GHWHUPLQH D EHVW JXHVV IRU WKH WUXH ILQLVKLQJ YDOXH REWDLQDEOH IURP WKDW QRGH (YDOXDWLRQ IXQFWLRQV XVH GRPDLQ NQRZOHGJH WR SUHGLFW WKH RXWFRPH RI WKH JDPH IURP D VSHFLILF SRVLWLRQ $OO GRPDLQ VSHFLILF NQRZOHGJH LV FRQWDLQHG LQ WKH HYDOXDWLRQ IXQFWLRQ 7KH TXDOLW\ RI NQRZOHGJH KDV D GLUHFW HIIHFW RQ WKH SOD\LQJ DELOLW\ RI WKH JDPH SURJUDP %HUOLQHU t *RHWVFK DQG /HH t 0DKDMDQ f 1RUWKZHVWHUQ 8QLYHUVLW\fV &+(66 ZDV DEOH WR GHIHDW RWKHU FKHVV SURJUDPV ZKLFK VHDUFKHG VHYHUDO SO\ OHYHOVf GHHSHU LQWR WKH JDPH WUHH EDVHG RQ WKH VWUHQJWK RI LWV HYDOXDWLRQ IXQFWLRQ (OLWKRUQ t %DQHUML DQG :LONLQV f 7KH SULPDU\ SUREOHP IDFHG E\ VHDUFK DOJRULWKPV XVLQJ WUHH VWUXFWXUHV LV WKH KRUL]RQ HIIHFW 7KH KRUL]RQ HIIHFW RFFXUV ZKHQ WKH VROXWLRQ WR D VHDUFK SUREOHP OLHV DW D GHSWK LQ WKH WUHH WKDW LV EH\RQG WKH FXUUHQW VHDUFK GHSWK 7KLV LV D UHDOLVWLF SUREOHP IRU FRPSOH[ DGYHUVDULDO GRPDLQV ,Q FKHVV WKH .5.1 HQGJDPH NLQJ DQG URRN YHUVXV NLQJ DQG NQLJKW FDQ UHTXLUH D WZHQW\VHYHQ PRYH VROXWLRQ RU D JDPH WUHH VHDUFK GHSWK RI ILIW\WZR SO\ %UDWNR t 0LFKLH f 6LPLODUO\ WKH .%%.1 HQGJDPH KDV VROXWLRQV RI VL[W\VL[ PRYHV RU RQH KXQGUHG WKLUW\WZR SO\ VHDUFK GHSWK 5R\FURIW f &XUUHQW FKHVV SURJUDPV VHDUFK EHWZHHQ HLJKW DQG IRXUWHHQ SO\ EHIRUH VHOHFWLQJ D PRYH 7ZR GLIIHUHQW VFKRROV RI WKRXJKW H[LVW FRQFHUQLQJ PHWKRGV IRU FRQWLQXLQJ WR LPSURYH JDPH SURJUDP SHUIRUPDQFH 7KH ILUVW VFKRRO EHOLHYHV WKDW WKH GHSWK RI WKH VHDUFK LV WKH VROH FULWHULRQ DIIHFWLQJ WKH SOD\LQJ DELOLW\ RI JDPH SURJUDPV

PAGE 26

$GHOVRQ9HOVN\ HW DO f VWDWH WKDW LQFUHDVLQJ WKH GHSWK RI VHDUFK RI D JDPH WUHH LQFUHDVHV WKH TXDOLW\ DQG UHOLDELOLW\ RI JDPH SOD\ %HUOLQHU DQG (EHOLQJ f VSHFLILFDOO\ VWDWH WKDW D GHHS WUHH VHDUFK ZLOO RXWSHUIRUP D VKDOORZHU VHDUFK ZKLFK XVHV PRUH NQRZOHGJH 9DULRXV WHFKQLTXHV DUH XVHG WR LQFUHDVH WKH GHSWK RI VHDUFK IRU D SURJUDP /LPLWLQJ WKH EUHDGWK RI VHDUFK DQG SUXQLQJ RI XQGHVLUDEOH EUDQFKHV WKURXJK DOSKDEHWD SUXQLQJ DQG LWHUDWLYH GHHSHQLQJ SHUPLW D GHHSHU VHDUFK IRU D VWDWLF QXPEHU RI HYDOXDWHG WUHH QRGHV %DUU t )HLJHQEDXP f 6SHFLDOL]HG KDUGZDUH FDQ DOVR EH XVHG WR LQFUHDVH WKH VSHHG DQG WKHUHE\ WKH GHSWK RI VHDUFK (EHOLQJ f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

PAGE 27

/(0(/ )LJXUH 6DPSOH JDPH WUHH ZLWK DOSKDEHWD SUXQLQJ RI NQRZOHGJH LQ WKH FKHVV GRPDLQ LV QLQH SO\ IRU WDFWLFDO NQRZOHGJH WZHQW\ILYH SO\ IRU SRVLWLRQDO NQRZOHGJH DQG IRUW\ILYH SO\ IRU VWUDWHJLF NQRZOHGJH LQ (OLWKRUQ t %DQHUML f &XUUHQW JDPH SURJUDPV XVH PRVWO\ WDFWLFDO NQRZOHGJH DORQJ ZLWK VRPH SRVLWLRQDO NQRZOHGJH )RU DGYHUVDULDO GRPDLQV WKH XVH RI GRPDLQ NQRZOHGJH FDQ EH GDQJHURXV E\ OHDGLQJ XV LQWR D IDOVH VHQVH RI VHFXULW\ 6DPXHO f ZDUQV WKDW SURJUDPV XVLQJ WKH 0LQLPD[ DOJRULWKP PXVW WDNH LQWR FRQVLGHUDWLRQ WKH LQWHQW RI WKH DGYHUVDU\ &HFFKLQL LQ &URRNDO HW DO f FODLPV WKDW WKH DFWXDO LQWHQW RI WKH DGYHUVDU\ ZLOO UHTXLUH SURJUDPV WR XVH GLIIHUHQW KHXULVWLFV LQ RUGHU WR ZLQ +XPDQ WUHH VHDUFK LV JRDO GLUHFWHG DQG WKH GLUHFWLRQ DQG PHWKRG RI VHDUFK PD\ FKDQJH DV QHZ LQIRUPDWLRQ

PAGE 28

EHFRPHV DYDLODEOH )UH\ f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f LQ WKH IRUP RI VLJQDWXUH WDEOHV DQG LQ WKH 5XVVLDQ FKHVV SURJUDP .$,66$ $GHOVRQ9HOVN\ HW DO f WR SHUIRUP FXWRIIV RU SUXQLQJ 7KH WKHRUHWLFDO LQFUHDVH LQ VHDUFK GHSWK ZKLFK ZH KDYH DFKLHYHG WKURXJK WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ LV GHWDLOHG LQ &KDSWHU 6HYHUDO DXWKRUV )LQGOHU t 0HOW]HU DQG 8WJRII f KDYH QRWHG WKDW WKH HFRQRP\ RI D VHDUFK KHXULVWLF FDQ EH PHDVXUHG E\ WKH UHGXFWLRQ LQ VHDUFK HIIRUW DQG FRPSOH[LW\ ZKLFK SURGXFHV D GHHSHU VHDUFK $WKOHWLF $GYHUVDULDO 'RPDLQV 7KH PRGHOLQJ RI DQ DGYHUVDU\ E\ KXPDQV LV ZHOO GRFXPHQWHG LQ DWKOHWLF GRPDLQV &RDFKHV RI EDVHEDOO $OLVWRQ t :HLVNRSI f IRRWEDOO 5LJJLQV t :LQWHU f DQG RWKHU VSRUWV XVH VFRXWLQJ UHSRUWV DQG ILOPV RI WKHLU RSSRQHQWV WR VWXG\ WKHLU RSSRQHQWfV VWUDWHJLHV 7KLV HQDEOHV WKH FRDFKHV RI DWKOHWLF WHDPV RU

PAGE 29

LQGLYLGXDOV WR GHYHORS FRXQWHUVWUDWHJLHV WR IRLO WKH JRDOV RI WKH DGYHUVDU\ DQG SURPRWH WKHLU RZQ JRDOV ,Q D SHUVRQDO FRPPXQLFDWLRQ 2ZHQ +RO\RDN RI WKH ([HUFLVH DQG 6SRUW 6FLHQFHV 'HSDUWPHQW DW WKH 8QLYHUVLW\ RI )ORULGD VWDWHG WKDW WKH DPRXQW RI SUHSDUDWLRQ SHUIRUPHG E\ DWKOHWLF WHDP FRDFKHV FRUUHVSRQGV GLUHFWO\ WR WKH SHUFHLYHG WKUHDW RI WKH DGYHUVDU\ :KHQ WKH WHFKQLTXH RI VWXG\LQJ DQ DGYHUVDU\fV SDVW SHUIRUPDQFH LV QRW IXOO\ XVHG VXUSULVLQJ UHVXOWV FDQ RFFXU $W WKH KRPHFRPLQJ IRRWEDOO JDPH IRU WKH 8QLYHUVLW\ RI )ORULGD 8)f WKH RSSRQHQW ZDV :HVW 7H[DV 6WDWH 8QLYHUVLW\ :76f $OWKRXJK :76 ZDV FRQVLGHUHG WR EH DQ XQGHUGRJ DQG ZDV QRW H[SHFWHG WR VFRUH DQ\ SRLQWV WKH\ REWDLQHG D IRXUWHHQ SRLQW OHDG HDUO\ LQ WKH JDPH E\ XVLQJ VRPH SOD\V WKDW 8) KDG QRW DQWLFLSDWHG 7KH ODFN RI SUHSDUDWLRQ RQ 8)fV SDUW ZDV UHVSRQVLEOH IRU WKH VXUSULVLQJ VWDUW RI WKH KRPHFRPLQJ JDPH (GXFDWLRQ 'RPDLQ 7KH DUWLILFLDO LQWHOOLJHQFH GRPDLQ RI LQWHOOLJHQW WXWRULQJ V\VWHPV ,76Vf RU LQWHOOLJHQW FRPSXWHUDLGHG LQVWUXFWLRQ KDV DOUHDG\ PDGH VXFFHVVIXO XVH RI PRGHOV RI KXPDQ WKLQNLQJ 7KHVH PRGHOV DUH EDVHG RQ WKH DYDLODELOLW\ RI NQRZOHGJH WR WKH KXPDQ DQG WKDW KXPDQfV PHQWDO VHW DW D SDUWLFXODU SRLQW LQ WLPH 7KH SHRSOH EHLQJ PRGHOHG LQ ,76 SURJUDPV DUH WKH VWXGHQWV ZKR DUH DWWHPSWLQJ WR OHDUQ 7KH VWXGHQW PRGHOV DUH XVHG WR SUHGLFW WKH VRXUFHV RI HUURU LQ VWXGHQW WKLQNLQJ ZKLFK FDXVH LQFRUUHFW UHVSRQVHV WR WHVW LWHPV SURGXFHG E\ WKH ,76

PAGE 30

:RROI f KDV GHPRQVWUDWHG DQ ,76 ZKLFK XVHV D PRGHO RI WKH SUREDEOH NQRZOHGJH FXUUHQWO\ KHOG E\ WKH VWXGHQW IRU GLVFRXUVH SODQQLQJ 7KLV PRGHO DGDSWV LWV XQGHUVWDQGLQJ RI WKH VWXGHQWfV NQRZOHGJH GXULQJ WKH OHVVRQ WR FDSWXUH FKDQJHV LQ WKH VWXGHQWfV FRJQLWLYH VWDWH SUHFLSLWDWHG E\ WKH DFTXLVLWLRQ RI QHZ NQRZOHGJH 2XU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ RSHUDWHV VLPLODUO\ E\ FRQVWUXFWLQJ D PRGHO RI WKH SUREDEOH HYDOXDWLRQ PHFKDQLVPV NQRZQ WR DQ DGYHUVDU\ WR SUHGLFW IXWXUH VWUDWHJLF GHFLVLRQV RI WKH DGYHUVDU\ 2WKHU H[DPSOHV RI WKLV PRGHOLQJ DSSURDFK DV XVHG LQ ,76V FDQ EH IRXQG LQ 6OHHPDQ DQG %URZQ f $GYHUVDULDO *DPH 'RPDLQV %HFDXVH RI WKHLU ILQLWH VL]H JDPHV SURYLGH XV ZLWK D FRQYHQLHQW DEVWUDFWLRQ RI PRUH GLIILFXOW UHDO ZRUOG SUREOHPV 7KH IRUPHU -DSDQHVH XVH RI WKH JDPH RI *R WR WUDLQ WKHLU PLOLWDU\ OHDGHUV 5HLVV *DPHV f GHPRQVWUDWHV WKH SUDFWLFDO H[WHQVLRQ RI D JDPH GRPDLQ WR D PLOLWDU\ GRPDLQ 3ULWFKDUG f HODERUDWHV IXUWKHU E\ FODLPLQJ WKDW *R LV JXHULOOD ZDUIDUH RQ D JUDQG VFDOH :H ZLOO EH XVLQJ WZR SHUVRQ ]HURVXP SHUIHFW LQIRUPDWLRQ JDPHV =DJDUH f WR GHPRQVWUDWH WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ =HURVXP JDPHV DUH DOZD\V DGYHUVDULDO LQ QDWXUH VLQFH WKH ZLQV RI RQH VLGH DUH HTXLYDOHQW WR WKH ORVVHV RI WKH RWKHU VLGH 2XU VHOHFWLRQ RI SHUIHFW LQIRUPDWLRQ JDPHV ZDV PDGH WR SHUPLW DQ DOJRULWKPLF HYDOXDWLRQ RI WKH UHVXOWV 7KH GRPDLQ RI FKHVV ZKLFK ZH ZLOO XVH IRU RXU DSSOLFDWLRQ RI WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ LV UHYLHZHG LQ VHFWLRQ

PAGE 31

2WKHU W\SHV RI JDPHV VXFK DV JDPHV RI FKDQFH WKDW GR QRW ILW RXU VHOHFWLRQ RI ]HURVXP SHUIHFW LQIRUPDWLRQ JDPHV FDQ DOVR EHQHILW IURP WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ 2XU FRQWULEXWLRQ ZKLFK UHGXFHV WKH FRPSOH[LW\ RI VHDUFK LV XVDEOH E\ DQ\ JDPH SURJUDP WKDW XWLOL]HV JDPH WUHH VHDUFK DQG KHXULVWLFV WR VROYH GRPDLQ SUREOHPV %HUOLQHUfV %.* EDFNJDPPRQ SURJUDP LQ /HY\ f LV DQ H[DPSOH RI D JDPH RI FKDQFH ZKLFK XVHV KHXULVWLF PHWKRGV DQG JDPH WUHH VHDUFK WR VHOHFW EHWZHHQ PRYHV &KULVWHQVHQ DQG .RUI f FODLP WKDW KHXULVWLF PHWKRGV IRU RQH SHUVRQ SX]]OHV DUH HTXLYDOHQW WR KHXULVWLF PHWKRGV IRU WZR SHUVRQ JDPHV 7KH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ LV FDSDEOH RI SUHGLFWLQJ PRYHV LQ RQH SHUVRQ JDPHV WKDW ZLOO UHGXFH WKH FRPSOH[LW\ RI WKH FXUUHQW JDPH SRVLWLRQ WR RQH ZKLFK KDV D NQRZQ VROXWLRQ 2WKHU UHVHDUFK XVLQJ LQGXFWLRQ RI SUHYLRXV JDPHV KDV EHHQ SHUIRUPHG IRU RQH SHUVRQ JDPHV 35()(5 8WJRII t 6D[HQD f VROYHV WKH SX]]OH E\ SHUIRUPLQJ LQGXFWLRQ RQ SUHYLRXV SX]]OH VROXWLRQV 6DPXHO DQG f ZDV WKH ILUVW UHVHDUFKHU WR XVH PDFKLQH OHDUQLQJ WR FUHDWH DQ LQWHOOLJHQW JDPH SOD\LQJ SURJUDP 6DPXHOfV FKHFNHU SOD\LQJ SURJUDP XVHV URWH OHDUQLQJ WHFKQLTXHV WR OHDUQ KRZ WR SOD\ FKHFNHUV 7KLV SURJUDP LV FDSDEOH RI LPLWDWLQJ WKH VW\OH RI DQ DGYHUVDU\ E\ PLPLFNLQJ WKH ZLQQLQJ PRYHV RI WKH DGYHUVDU\ SUHVHQWHG LQ WKH VHW RI WUDLQLQJ LQVWDQFHV 3URJUDPV ZKLFK UHO\ RQ URWH OHDUQLQJ WHQG WR EHFRPH LQIOH[LEOH DQG UHTXLUH DQ HTXDOO\ ODUJH QXPEHU RI UHWUDLQLQJ LQVWDQFHV WR DGDSW WR D GLIIHUHQW VW\OH RI SOD\ 7KH PHWKRG RI PLPLFNLQJ DQ DGYHUVDU\ KDV EHHQ VKRZQ WR IDLO IRU WKH JDPH RI *R 3ULWFKDUG f DQG VKRXOG QRW EH FRQVLGHUHG DV D JHQHUDOL]HG PHWKRG RI VWUDWHJLF SODQQLQJ DJDLQVW DGYHUVDULHV

PAGE 32

6DPXHO GHYHORSHG VLJQDWXUH WDEOHV WR UHGXFH VHDUFK FRPSOH[LW\ 6LJQDWXUH WDEOHV DUH XVHG WR UHPHPEHU WKH HYDOXDWLRQ IXQFWLRQ YDOXH IRU VSHFLILF SRVLWLRQV :KHQ LGHQWLFDO SRVLWLRQV DUH LGHQWLILHG WKURXJK D SDWWHUQ PDWFKLQJ PHFKDQLVP WKH VLJQDWXUH WDEOHV DUH XVHG WR UHWULHYH WKH FRUUHVSRQGLQJ QRGH YDOXH :LOFR[ f KDV DOVR DSSOLHG SDWWHUQ UHFRJQLWLRQ WHFKQLTXHV WR WKH JDPH RI *R WR FRQVWUXFW 1(0(6,6 D VWURQJ *R SOD\LQJ SURJUDP ,$*2 5RVHQEORRP f XVHV WKH VWDQGDUG VHDUFK WHFKQLTXHV RI DOSKDEHWD SUXQLQJ DQG LWHUDWLYH GHHSHQLQJ WR SOD\ FKDPSLRQVKLSOHYHO 2WKHOOR 5RVHQEORRP SRLQWV RXW WKDW WKH KHXULVWLFV IRU ,$*2 QHHG WR FKDQJH GXULQJ GLIIHUHQW SKDVHV RI WKH JDPH WR FDSWXUH VKLIWV LQ WKH VWUDWHJLF LPSRUWDQFH RI YDULRXV JDPH SULQFLSOHV )RU H[DPSOH LQ WKH PLGGOH JDPH RI 2WKHOOR PRELOLW\ LV D FULWLFDO IDFWRU EXW LQ WKH HQG JDPH SRUWLRQ RI 2WKHOOR WKH TXDQWLW\ RI VWDEOH SLHFHV LV WKH SULPDU\ FRQFHUQ %HUOLQHU f QRWHV VLPLODU VKLIWV LQ VWUDWHJLF UHDVRQLQJ IRU WKH GRPDLQ RI FKHVV 2XU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ FRPSHQVDWHV IRU WKH VWUDWHJLF VKLIWV DW YDULRXV SRLQWV LQ WKH JDPH E\ DVVLJQLQJ WLPH GXUDWLRQV WR WKH GRPDLQ HYDOXDWLRQ FKXQNV WKDW DUH DFTXLUHG IRU HDFK DGYHUVDU\ *,1$ 'H-RQJ t 6FKXOW] f WULHV WR OHDUQ WR SOD\ 2WKHOOR E\ XVLQJ H[SHULHQFH JDLQHG IURP SUHYLRXV JDPHV 7KLV DSSURDFK LV YHU\ VLPLODU WR WKH RQH XVHG E\ 6DPXHO IRU KLV FKHFNHUV SURJUDP *,1$ DWWHPSWV WR H[SORLW WKH ZHDNQHVVHV RI D VSHFLILF RSSRQHQW 6HYHUDO RI *,1$fV LPSOHPHQWDWLRQ IHDWXUHV DUH DSSOLFDEOH DFURVV DGYHUVDULDO GRPDLQV 7KHVH IHDWXUHV DUH

PAGE 33

f 6HUYH DV DQ H[WHUQDO FRDFK WR DQ H[LVWLQJ SURJUDP LQ WKH GRPDLQ f 2QO\ XVH REVHUYDEOH EHKDYLRU WR OHDUQ DERXW DQ DGYHUVDU\ )RU *,1$ WKH REVHUYDEOH EHKDYLRU LV WKH PRYHV PDGH E\ DQ DGYHUVDU\ LQ SULRU JDPHV f 'R QRW XVH IHHGEDFN WR PRGLI\ OHDUQLQJ H[FHSW IRU WKH UHVXOWV RI WKH JDPHV WKDW KDYH EHHQ VWXGLHG 7KH EHVW H[DPSOH RI VWUDWHJLF SODQQLQJ EDVHG RQ DQ DGYHUVDU\fV PRGXV RSHUDQGL LV :DWHUPDQfV f SRNHU SOD\HU :DWHUPDQ KDV LPSOHPHQWHG D GUDZ SRNHU SOD\LQJ SURJUDP ZKLFK DQDO\]HV DQ RSSRQHQWfV VW\OH RI SOD\ WR GHFLGH ZKHQ D EHW LV DSSURSULDWH 7KLV SURJUDP OHDUQV RYHU D SHULRG RI VHYHUDO KDQGV ZKLFK KHXULVWLFV WR XVH DJDLQVW D SDUWLFXODU DGYHUVDU\ :H H[SDQG XSRQ :DWHUPDQfV LGHD E\ QRW OLPLWLQJ WKH SURJUDP WR D VWDWLF QXPEHU RI SUHGHILQHG KHXULVWLFV /DZ (QIRUFHPHQW 'RPDLQV 3ROLFH GHSDUWPHQWV DQG WKH )HGHUDO %XUHDX RI ,QYHVWLJDWLRQ HDFK PDNH XVH RI DGYHUVDU\ PRGHOLQJ $V DQ H[DPSOH D SROLFH GHWHFWLYH ZLOO WU\ WR SUHGLFW ZKHQ ZKHUH DQG WR ZKRP D VHULDO NLOOHU ZLOO SHUIRUP KLV QH[W FULPH 7KLV LV GRQH SULPDULO\ E\ DQDO\]LQJ WKH FULPLQDOfV SUHYLRXV SHUIRUPDQFHV RI FULPHV 6LPLODU WHFKQLTXHV DUH EHLQJ FRQVLGHUHG IRU SUHGLFWLQJ WKH RXWFRPH RI SRVVLEOH WHUURULVW DFWLYLWLHV LQ WKH 5,6. SURMHFW VSRQVRUHG E\ /DZUHQFH /LYHUPRUH 1DWLRQDO /DERUDWRULHV

PAGE 34

7KH NH\ WR VXFFHVVIXO GUXJ HQIRUFHPHQW LV LQWHOOLJHQFH JDWKHULQJ DQG LQIRUPDWLRQ SURFHVVLQJ :LWNLQ f 5DZ GDWD LV YROXPLQRXV DQG PXVW EH FRQYHUWHG WR D XVHDEOH IRUP 2XU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ LV FDSDEOH RI ILQGLQJ SDWWHUQV LQ UDZ GDWD WKDW FRUUHVSRQG WR UHVRXUFH PRYHPHQW SUREOHPV HJ WKH DFWXDO GUXJV DQG WKH PRQHWDU\ FRPSHQVDWLRQVf DQG FDQ EH XVHG E\ ODZ HQIRUFHPHQW RIILFLDOV WR LQFUHDVH WKHLU HIIHFWLYHQHVV 0LOLWDU\ 'RPDLQV 'HYDVWDWLQJ UHVXOWV RFFXU RQ WKH EDWWOHILHOG ZKHQ VXIILFLHQW SUHSDUDWLRQ DJDLQVW DQ DGYHUVDU\ LV QRW SHUIRUPHG 'XH WR GLIIHUHQW VRFLRORJLFDO JHRJUDSKLFDO DQG HGXFDWLRQDO EDFNJURXQGV SRWHQWLDO DGYHUVDULHV ZLOO KDYH XQLTXH EHKDYLRUDO UXOHV %RQG DQG %RULQJ f 7KHVH EHKDYLRUDO UXOHV JRYHUQ YDU\LQJ EHOLHIV DQG JRDOV FRQFHUQLQJ ZDU DQG WKH XVH RI QXFOHDU ZHDSRQV 0F0LOOHQ DQG 6RNRORYVNL\ f 'XQQLJDQ f DQG (UPDUWK f VWDWH WKDW LI ZH DVVXPH WKDW DQ HQHP\ ZLOO DFW VWULFWO\ DFFRUGLQJ WR RXU RZQ EHOLHIV WKHQ ZH ZLOO SURGXFH D WDFWLFDO EOLQGQHVV 7DFWLFDO EOLQGQHVV SURGXFHV WKH LQDELOLW\ WR SUHGLFW RU IDWKRP WKH SODQV RI RXU DGYHUVDULHV +LVWRULF H[DPSOHV RI WKH WDFWLFDO DQG VWUDWHJLF EOLQGQHVV SURGXFHG E\ FRQVFULSWLQJ RXU RZQ EHOLHIV RQWR DQ DGYHUVDU\ DUH WKH ERPELQJ RI 3HDUO +DUERU E\ WKH -DSDQHVH LQ WKH GHIHDW RI WKH 8QLWHG 6WDWHV WDFWLFDO IRUFHV DW WKH &KRVLQ 5HVHUYRLU GXULQJ WKH .RUHDQ :DU :DOODFH f DQG WKH 7HW 2IIHQVLYH RI WKH 1RUWK 9LHWQDPHVH LQ 0DQ\ VROGLHUfV OLYHV PLJKW KDYH EHHQ VDYHG HDUO\ LQ :RUOG :DU ,, LI WKH 8QLWHG 6WDWHV VROGLHUV KDG XQGHUVWRRG WKH -DSDQHVH DWWLWXGH WRZDUG VXUUHQGHU %RULQJ f

PAGE 35

+XPDQ $GYHUVDU\ 0RGHOLQJ (IIRUWV LQ 0LOLWDU\ 'RPDLQV 0LOLWDU\ OHDGHUV *UXQHU DQG 5\DQ f FODLP WKDW YLFWRU\ LV RIWHQ FRQWLQJHQW XSRQ NQRZLQJ \RXU HQHP\ 7KLV PHDQV WKDW PLOLWDU\ SODQQHUV PXVW EH DEOH WR HYDOXDWH WKH FXUUHQW VLWXDWLRQ XVLQJ WKH VDPH FULWHULD DV WKHLU DGYHUVDU\ $ YLWDO FRPSRQHQW RI LQWHOOLJHQFH LQIRUPDWLRQ LV EULHILQJV RQ DQ DGYHUVDU\fV RSHUDWLRQDO FKDUDFWHULVWLFV RU WDFWLFV *UXQHU f 5REHUW ( /HH ZKR LV RIWHQ FLWHG DV WKH JUHDWHVW PLOLWDU\ VWUDWHJLVW % 'DYLV f LV D SUDFWLFDO H[DPSOH RI WKH SRZHU DFFRPSDQ\LQJ GHWDLOHG NQRZOHGJH DERXW DQ DGYHUVDU\ /HH ZDV HGXFDWHG DW :HVW 3RLQW ZKHUH DOO RI WKH 8QLRQ FRPPDQGHUV ZHUH DOVR WUDLQHG DQG VHUYHG DORQJ VLGH PDQ\ RI WKH DGYHUVDULHV KH ZRXOG IDFH LQ WKH &LYLO :DU GXULQJ KLV WHQXUH DV DQ RIILFHU LQ WKH 8QLWHG 6WDWHV $UP\ DQG &DYDOU\ 6QRZ f /HHfV GHWDLOHG NQRZOHGJH RI WKH WUDLQLQJ DQG WDFWLFV RI KLV DGYHUVDULHV HQDEOHG KLP WR IUHTXHQWO\ RXWSHUIRUP EHWWHU HTXLSSHG DQG ODUJHU IRUFHV 7KH 8QLWHG 6WDWHV PLOLWDU\ PDNHV XVH RI WKLV FRQFHSW LQ WUDLQLQJ RXU VROGLHUV (DFK VHUYLFH EUDQFK PDLQWDLQV D FDGUH RI SHUVRQQHO ZKR DUH WUDLQHG LQ 6RYLHWVW\OH ZDU WDFWLFV DQG VWUDWHJ\ 5REELQV f 7KH WUDLQLQJ FHQWHUV IRU WKH $UP\ $LU )RUFH DQG 1DY\ DUH WKH 1DWLRQDO 7UDLQLQJ &HQWHU 5HG )ODJ DQG WKH 7RS *XQ VFKRRO UHVSHFWLYHO\ 7KH SXUSRVH RI WKHVH 6RYLHW DQDORJRXV ILJKWLQJ IRUFHV LV WR SURYLGH D UHDOLVWLF WUDLQLQJ HQYLURQPHQW WR EHWWHU SUHSDUH WKH PLOLWDU\ IRU WKH SRVVLELOLW\ RI DQ DFWXDO ZDU DJDLQVW 6RYLHW IRUFHV RU IRUFHV WKDW XVH 6RYLHWVW\OH WDFWLFV $ KLVWRULF H[DPSOH RI DGYHUVDU\ PRGHOLQJ IRU PLOLWDU\ SXUSRVHV RFFXUUHG GXULQJ :RUOG :DU ,, $ GHWDLOHG SV\FKRORJLFDO SURILOH RI +LWOHU ZDV FRQVWUXFWHG E\

PAGE 36

/DQJHU /DQJHU f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fV LQWHQWLRQV DQG VWUDWHJLF VW\OH RWKHUZLVH WDFWLFDO EOLQGQHVV ZLOO UHVXOW 7KH WDFWLFDO WURRS PRYHPHQW SODQQHU 7$&3/$1 $QGULROH HW DO f LV W\SLFDO RI PRVW PLOLWDU\ SODQQLQJ WRROV 7$&3/$1 RQO\ DFFRXQWV IRU VWDWLF HOHPHQWV RI WKH GRPDLQ VXFK DV WKH HIIHFW RI WHUUDLQ RQ HTXLSPHQW PRYHPHQW .QRZOHGJH DERXW DGYHUVDULHV LV QRQH[LVWHQW 7KH QH[W VWHS WRZDUGV DQ DGYHUVDU\ PRGHO LV SURJUDPV VLPLODU WR $5(6
PAGE 37

SHUPLWWLQJ D FKHVV DGYHUVDU\ WR PDNH IRUZDUG PRYHV ,QILQLWHO\ PDQ\ VWUDWHJLF FKRLFHV FDQ EH PDGH E\ DQ DGYHUVDU\ WKDW ZLOO QRW EH FRYHUHG E\ WKH SUHVFULEHG DFWLRQ FKRLFHV WKDW DUH SHUPLWWHG E\ WKH WDFWLFDO SODQQLQJ DOJRULWKP 7KH 32/,7,&6 V\VWHP &DUERQHOO f SURFHHGV LQ WKH GHYHORSPHQW SODQ E\ LQWURGXFLQJ JRDOV WKDW DQ DGYHUVDU\ LV WU\LQJ WR DFFRPSOLVK VXFK DV EORFNLQJ WKH FXUUHQW JRDO RI WKH SODQQLQJ SURJUDP 7KH DGYHUVDULDO JRDOV XVHG LQ 32/,7,&6 DUH VWDWLFDOO\ GHILQHG DQG DUH EDVHG RQ SUHFRQFHLYHG QRWLRQV RI DQ DGYHUVDU\fV LQWHQW %DVLQJ WKH DFWLRQV RI DQ DGYHUVDU\ RQ VWDWLF SUHGHILQHG JRDOV UXQV WKH ULVN RI FUHDWLQJ WDFWLFDO EOLQGQHVV PHQWLRQHG LQ WKH LQWURGXFWLRQ WR VHFWLRQ 7KH DFWXDO JRDOV RI DQ DGYHUVDU\ PXFK OHVV KRZ WKH DGYHUVDU\ LQWHQGV WR DFFRPSOLVK WKRVH JRDOV PD\ QRW EH NQRZQ GXULQJ WKH FRQWHVW RU PD\ EH FRQWUDU\ WR WKH H[SHFWHG JRDOV &URRNDOO HW DO DQG 1DUHQGUD t 7KDWKDFKDU f 32/,7,&6 GRHV SURYLGH D JRRG H[DPSOH RI D SURJUDP LQ ZKLFK D WKLQNLQJ DGYHUVDU\ DWWHPSWV WR DFKLHYH VSHFLILF JRDOV 70 (ULFNVRQ t =\WNRZ f LV DQRWKHU JRDO EDVHG V\VWHP IRU PDNLQJ WDFWLFDO GHFLVLRQV 7KH SURJUDP %2*(< LV DQ H[WHUQDO OHDUQLQJ SURJUDP ZKLFK JHQHUDWHV ODUJH QXPEHUV RI VLPXODWLRQV RI WDFWLFDO HQJDJHPHQWV ZLWK VSHFLILF JRDOV %2*(< OLNH *,1$ LQ VHFWLRQ LV DQRWKHU H[DPSOH RI WKH XWLOLW\ RI FRDFKLQJ SURJUDPV WKDW VXSSRUW H[LVWLQJ GRPDLQ VSHFLILF SURJUDPV WKURXJK DQ H[WHUQDO OHDUQLQJ IXQFWLRQ 7KH 0$5. DQG 0$5. ,, V\VWHPV 'DYLV D DQG E DQG 'DYLV HW DO f HPERG\ WKH PLQG VHW RI DQ DGYHUVDU\ LQ '(:7 'HWHUUHQFH (VFDODWLRQ

PAGE 38

FRQWURO DQG :DU 7HUPLQDWLRQ PRGHOV &XUUHQWO\ WKH 0$5. V\VWHPV KDYH WZR GHFLVLRQ PRGHOV (DVW DQG :HVW :KLOH 0$5. DQG 0$5. ,, DWWHPSW WR FDSWXUH GLIIHUHQFHV LQ WKH PLQG VHWV EHWZHHQ DGYHUVDULHV WKH PRGHOV DUH VWDWLFDOO\ GHILQHG OLNH WKH JRDOV RI WKH 32/,7,&6 SURJUDP 7KHVH VWDWLF PLQG VHWV DUH EDVHG RQ SUHGHILQHG FRQFHSWLRQV RI DQ DGYHUVDU\fV LQWHQWLRQV ZKLFK PD\ QRW EH DFFXUDWH DQG ZLOO QRW DFFRXQW IRU WKH PRGLILFDWLRQ RYHU WLPH WR GHFLVLRQ KHXULVWLFV XVHG E\ DQ DGYHUVDU\ 2QH RI WKH FRQWULEXWLRQV WKDW DUWLILFLDO LQWHOOLJHQFH FDQ SURYLGH WR PLOLWDU\ SODQQLQJ LV UHDVRQLQJ WKDW DFFRXQWV IRU WKH WHPSRUDO UHOHYDQFH RI DFWLRQV %RQDVVR f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f 7\SLFDO ERDUG ZDUJDPHV PDNH XVH RI D KH[DJRQDO PDS IRU SORWWLQJ WKH PRYHPHQW RI UHVRXUFHV %\ XVLQJ VXFK PDS

PAGE 39

UHSUHVHQWDWLRQV WKH XVH RI VSHFLILF IRUPDWLRQV E\ D SDUWLFXODU FRPPDQGHU FDQ EH DFTXLUHG E\ RXU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ 3ROLWLFDO DQG (FRQRPLF 'RPDLQV 7RS OHYHO SROLF\ GHFLVLRQV PDGH E\ JRYHUQPHQW RIILFLDOV DUH DOVR DIIHFWHG E\ SULRU NQRZOHGJH DERXW DQ DGYHUVDU\fV JRDOV DQG VWUDWHJLHV &LPEDOD f *RYHUQPHQW QHJRWLDWRUV DWWHPSW WR GLVFRYHU WKH OLPLWDWLRQV IDFHG E\ DQ DGYHUVDU\ WR DFFRPSOLVK WKH RSWLPDO RXWFRPH RI DQ DJUHHPHQW RU WUHDW\ EHWZHHQ WKH WZR DGYHUVDULHV 7KH XVH RI NQRZOHGJH RI WKH SROLWLFDO PLQG VHW IRU D JHRJUDSKLF UHJLRQ SHUPLWV QHJRWLDWRUV WR LQWHUSUHW LQIRUPDWLRQ IURP WKH SURSHU SHUVSHFWLYH &LPEDOD f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

PAGE 40

5HOLJLRQ 'RPDLQ 7KH LPSRUWDQFH RI NQRZLQJ WKH LQWHQWLRQV DQG VWUDWHJLF VW\OH RI DQ DGYHUVDU\ KDV EHHQ UHFRUGHG WKURXJKRXW KLVWRU\ $Q HDUO\ H[DPSOH RI WKHVH ZULWLQJV FRPHV IURP WKH ILUVW FHQWXU\ $' ZKHQ WKH DSRVWOH 3DXO &RULQWKLDQV f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

PAGE 41

+XPDQ 8VH RI $GYHUVDU\ 0RGHOLQJ LQ &KHVV %HIRUH SOD\LQJ IRU WKH :RUOG &KHVV &KDPSLRQVKLS DJDLQVW &DSDEODQFD ZKR ZDV FRQVLGHUHG WR EH LQYLQFLEOH DQG ZDV DSSURSULDWHO\ FDOOHG 7KH &KHVV 0DFKLQH $OHNKLQH VWXGLHG DQG DQDO\]HG &DSDEODQFDfV SULRU JDPHV 6FKRQEHUJ f )URP KLV VWXG\ $OHNKLQH ZDV DEOH WR GHWHUPLQH D ZHDNQHVV LQ &DSDEODQFDfV SOD\ WKDW KH ZDV DEOH WR H[SORLW WR EHFRPH WKH QHZ :RUOG &KHVV &KDPSLRQ 2WKHU FKHVV PDVWHUV LQFOXGLQJ WKH :RUOG &KHVV &KDPSLRQV %RWYLQQLN 7DO DQG .DVSDURY SUHSDUH IRU FKHVV PDWFKHV E\ ULJRURXVO\ VWXG\LQJ WKH SUHYLRXV JDPHV RI WKHLU RSSRQHQWV +RURZLW] f 1XQQ 1XQQ t *ULIILWKV f GHVFULEHV D JDPH DJDLQVW VW+DQVHQ LQ ZKLFK ERWK SOD\HUV EDVHG WKHLU JDPH VWUDWHJ\ RQ H[SHFWDWLRQV RI WKHLU DGYHUVDU\fV SOD\LQJ VW\OH 7KHVH H[SHFWDWLRQV ZHUH IRUPHG IURP SULRU H[SHULHQFH DJDLQVW WKH DGYHUVDU\ $OWKRXJK QR RQH ZDV DEOH WR FDSLWDOL]H XSRQ WKH LQIRUPDWLRQ LW KDV EHHQ QRWHG WKDW SULRU WR KLV :RUOG &KDPSLRQVKLS PDWFK DJDLQVW 6SDVVN\ )LVFKHU YDULHG KLV VWDQGDUG RSHQLQJ IRU ZKLWH IURP H SDZQ WR .LQJ IRXU RQO\ WKUHH WLPHV GXULQJ WRXUQDPHQW OHYHO SOD\ (YDQV f 6HYHUDO RI WKH FKHVV JUDQGPDVWHUV LQFOXGLQJ /DVNHU DQG 7DO +RURZLW] DQG 6FKRQEHUJ f KDYH EHHQ GHVFULEHG DV SV\FKRORJLFDO SOD\HUV 7KHVH SOD\HUV ZRXOG LQWHQWLRQDOO\ PDNH WKH PRYH WKDW ZRXOG EH PRVW GLVWXUELQJ WR WKHLU RSSRQHQWV HYHQ WKRXJK VXFK PRYHV ZHUH IUHTXHQWO\ QRW WKH EHVW WDFWLFDO FKRLFH /HY\ f KDV GHVFULEHG KLV VW\OH RI SOD\ DJDLQVW FRPSXWHU FKHVV SURJUDPV VLPLODUO\ VWDWLQJ WKDW KLV PRYH FKRLFHV DUH RIWHQ PDGH WR FRQIXVH WKH FRPSXWHU

PAGE 42

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f KDV KDG ZLWK WKH FRPSXWHU FKHVV FRPPXQLW\ ,Q WKHVH EHWV /HY\ FODLPV WKDW QR FRPSXWHU FKHVV SURJUDP ZLOO EH DEOH WR EHDW KLP 7KH ILUVW EHW ZKLFK KDG D WHQ \HDU GXUDWLRQ ZDV ZRQ E\ /HY\ KRZHYHU '((3 7+28*+7 KDV GHIHDWHG /HY\ LQ WKH ODWHVW RI WKH EHWV $ UHFHQW GLVFXVVLRQ RQ WKH UHFJDPHVFKHVV QHZV QHWZRUN E\ )HQJ+VLXQJ +VX LQGLFDWHV DQ LQFUHDVH LQ WKH QXPEHU RI KXPDQ RSSRQHQWV RI FRPSXWHU FKHVV SURJUDPV XVLQJ DQWLFRPSXWHU SOD\LQJ VW\OHV DJDLQVW SURJUDPV HQWHUHG LQ PDMRU FKHVV WRXUQDPHQWV :KHQ 0LNH 9DOYR GHIHDWHG '((3 7+28*+7 WKH UHLJQLQJ FRPSXWHU FKHVV SURJUDP FKDPSLRQ IRU WKH SDVW WKUHH \HDUV LQ D WZR JDPH SRVWDO FKHVV PDWFK YLD HOHFWURQLF PDLO LW ZDV QRWHG 1HZERUQ t .RSHF f WKDW 9DOYR KDG REVHUYHG '((3 7+28*+7 SOD\ D QXPEHU RI JDPHV DQG KDG DFTXLUHG WKH SOD\LQJ VW\OH RI '((3 7+28*+7 7KH FKHVV SURJUDP GLG QRW KDYH D VLPLODU RSSRUWXQLW\ %HUOLQHU f FODLPV WKDW YDULRXV FKHVV SOD\HUV KDYH DVNHG IRU WKH SUHYLRXV JDPH UHFRUGV RI +,7(&+ WKH FRPSXWHU FKHVV SURJUDP FRFKDPSLRQ VR WKDW WKH\ PLJKW EH SUHSDUHG IRU D SRVVLEOH PDWFK

PAGE 43

,PSRUWDQFH RI 3DZQ 6WUXFWXUHV DQG &KHVV 2SHQLQJV :H SD\ VSHFLDO DWWHQWLRQ WR SDZQ IRUPDWLRQV GXULQJ RXU FKXQN OHDUQLQJ SURFHVV 3KLOLGRU KDV VWDWHG WKDW WKH ZLQ RU ORVV RI D JDPH GHSHQGV RQ WKH TXDOLW\ RI WKH SDZQ SRVLWLRQV GXULQJ WKH JDPH 3DFKPDQ f 7KH VWUHQJWK DQG ZHDNQHVV RI SDZQ IRUPDWLRQV LV RI SDUDPRXQW LPSRUWDQFH /DVNHU f 3DZQ IRUPDWLRQV FDQ EH XVHG WR UHYHDO LQIRUPDWLRQ FRQFHUQLQJ SRVLWLRQDO DGYDQWDJHV 5HLQIHOG f )XUWKHUPRUH WKUHH RI WKH VL[ FULWHULD IRU GHWHUPLQLQJ WKH FKDUDFWHU RI D SRVLWLRQ JLYHQ E\ &DSDEODQFD UHODWH WR WKH HIIHFWLYHQHVV RI WKH SDZQV LQ WKH JDPH 3DFKPDQ f 3DZQ IRUPDWLRQV FRQWDLQ LQIRUPDWLRQ DERXW WKH SRVLWLRQDO QDWXUH RI WKH FKHVV JDPH 6ROWLV f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f 2XU RZQ DQDO\VLV RI WKH JDPHV RI %RWYLQQLN DQG 6SDVVN\ KDV LQGLFDWHG WKDW HDFK RI WKHVH FKHVV JUDQGPDVWHUV SUHIHUV D SDUWLFXODU VW\OH FORVHG DQG RSHQ UHVSHFWLYHO\ 2SHQLQJV DOVR SOD\ DQ LPSRUWDQW UROH LQ GHWHUPLQLQJ WKH SRVLWLRQDO DGYDQWDJH RI WKH JDPH &XUUHQW FRPSXWHU FKHVV SURJUDPV XWLOL]H D ERRN RU GDWDEDVH RI RSHQLQJ PRYHV ZKLFK KDYH D GXUDWLRQ IURP WZR PRYHV WR PRUH WKDQ WHQ PRYHV 7KH HYDOXDWLRQ DOJRULWKPV ZKLFK HQDEOH FKHVV SURJUDPV WR HVWLPDWH WKH VWUHQJWK RI WKH

PAGE 44

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f KDV GHPRQVWUDWHG WKDW SDUWLFXODU RSHQLQJ IRUPDWLRQV FDQ EH DFKLHYHG WKURXJK D ZLGH YDULHW\ RI WKH DFWXDO PRYH RUGHU $ VWDWLVWLFDO DQDO\VLV RI WKH RSHQLQJ PRYH VHTXHQFHV GLVSOD\HG E\ DQ DGYHUVDU\ DFFRXQWV IRU RSHQLQJ PRYH RUGHU YDULDWLRQV E\ ILQGLQJ WKH XSSHU ERXQG ORZHU ERXQG DQG PHDQ WLPH RI VSHFLILF PRYHV 6HOHFWLQJ DQ RSHQLQJ OLQH RI SOD\ LV XVXDOO\ SHUIRUPHG DW UDQGRP RU LQ UHVSRQVH WR WKH DGYHUVDU\fV FKRLFH RI RSHQLQJ PRYH %\ VWXG\LQJ WKH SOD\LQJ VW\OH DQG SUHYLRXV JDPHV RI DQ DGYHUVDU\ VSHFLILF RSHQLQJ OLQHV WKDW DUH ZHOO NQRZQ WR DQ DGYHUVDU\ DUH LGHQWLILHG 7KLV NQRZOHGJH FDQ EH XVHG E\ D FKHVV SURJUDP WR HPXODWH KXPDQ FKHVV H[SHUWV E\ FKRRVLQJ RSHQLQJV WKDW DUH QRW DV ZHOO NQRZQ WR WKH DGYHUVDU\ 1XQQ 1XQQ t *ULIILWKV f DYRLGV WKH 3HOLNDQ YDULDWLRQ RI WKH 6LFLOLDQ 'HIHQVH LQ RQH RI KLV JDPHV DJDLQVW $QWKRQ\ EHFDXVH 1XQQ NQRZV IURP H[SHULHQFH WKDW $QWKRQ\ LV D 3HOLNDQ VSHFLDOLVW 2WKHU H[DPSOHV RI VHOHFWLQJ RSHQLQJ OLQHV RI SOD\ EDVHG RQ WKH SULRU SHUIRUPDQFH RI DQ DGYHUVDU\ DUH JLYHQ E\ 1XQQ $FTXLVLWLRQ RI WKH RSHQLQJ VHTXHQFH NQRZOHGJH KHOG E\ WKH DGYHUVDU\ ZLOO HQDEOH FXUUHQW FKHVV FRPSXWHU SURJUDPV WR VHOHFW OLQHV RI SOD\ ZKLFK DUH VWUDWHJLFDOO\

PAGE 45

DGYDQWDJHRXV +ROGLQJ f VXSSRUWV RXU SHUVSHFWLYH E\ VWDWLQJ WKDW ZH VKRXOG XVH LQIRUPDWLRQ DERXW SOD\HUV SOD\LQJ VW\OH WR GLVUXSW WKHLU QRUPDO SOD\ 3KLORVRSK\ 'HQQHWW f VWDWHV WKDW FXUUHQW FKHVV SURJUDPV FRQFHQWUDWH WKHLU WLPH RQ EUDQFKHV RI D JDPH WUHH ZKLFK UHSUHVHQW WKH DGYHUVDU\f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fV FKHVV UDWLQJ ZLWKRXW GDPDJLQJ KLV RZQ UDWLQJ DQG PLJKW LPSURYH WKH \RXQJ PDVWHUfV FRQILGHQFH DQG IXWXUH SOD\LQJ DELOLW\ 7KHUHIRUH WKH 6RYLHW JUDQGPDVWHU ZKR KDV DOUHDG\ ZRQ WKH WRXUQDPHQW E\ YLUWXH RI KLV SUHYLRXV YLFWRULHV RIIHUV WKH \RXQJHU PDVWHUOHYHO SOD\HU D GUDZ PLGZD\ WKURXJK WKH JDPH $QRWKHU H[DPSOH RI WKH SUREOHPV IDFHG E\ FRPSXWHU SURJUDPV LQ GHDOLQJ ZLWK WKH LQWHQWLRQDOLW\ RI KXPDQ DGYHUVDULHV LV SURYLGHG E\ /HY\ f ZKR FODLPV WKDW KH ZLOO LQWHQWLRQDOO\ FKRRVH LQIHULRU PRYHV ZKHQ SOD\LQJ DJDLQVW D FRPSXWHU 7KH LQIHULRU PRYH FKRLFHV DUH VHOHFWHG E\ /HY\ EHFDXVH KH EHOLHYHV WKDW FRPSXWHU FKHVV

PAGE 46

SURJUDPV KDYH JUHDWHU GLIILFXOW\ LQ SOD\LQJ D ZLQQLQJ JDPH DJDLQVW VXFK DQ LUUDWLRQDO VW\OH RI SOD\ 7KH /HY\ H[DPSOH DERYH GHPRQVWUDWHV RQH RI WKH WUDLWV RI KXPDQ DGYHUVDULHV +XPDQV HVSHFLDOO\ ZKHQ FRPSHWLQJ DJDLQVW D PDFKLQH DWWHPSW WR PLVOHDG WKHLU DGYHUVDU\ DERXW WKHLU LQWHQWLRQV DQG FDSDELOLWLHV :H PXVW EH DEOH WR GLVWLQJXLVK EHWZHHQ UHOHYDQW LQIRUPDWLRQ DQG PLVLQIRUPDWLRQ DERXW DQ DGYHUVDU\fV SOD\LQJ VW\OH ,QFRUSRUDWLQJ WKH GHWHFWLRQ RI PLVLQIRUPDWLRQ IURP DQ DGYHUVDU\ LQWR WKH LQGXFWLYH OHDUQLQJ PHFKDQLVP DGGV DQ DGGLWLRQDO KHXULVWLF HOHPHQW 7KH DGGLWLRQ RI DQRWKHU KHXULVWLF HOHPHQW UHGXFHV WKH SUREDELOLW\ RI DFTXLULQJ D YDOLG DGYHUVDU\ PRGHO 7KH HIIHFW RI DQ DGYHUVDU\fV PLVLQIRUPDWLRQ DWWHPSWV LV QHJDWHG E\ VHOHFWLQJ SUHYLRXV GRPDLQ SHUIRUPDQFHV WKDW OLPLW WKH SRVVLELOLW\ RI PLVLQIRUPDWLRQ %\ RQO\ XVLQJ WKH SUHYLRXV JDPHV RI DQ DGYHUVDU\ IURP WRXUQDPHQW SOD\ LQ ZKLFK WKH DGYHUVDU\ KDV VRPHWKLQJ DW VWDNH PLVLQIRUPDWLRQ LV HOLPLQDWHG )RU RXU DSSOLFDWLRQ LQ WKH FKHVV GRPDLQ ZH XVH WKH :RUOG &KDPSLRQVKLS 0DWFK JDPHV RI DGYHUVDULHV WDNHQ IURP WKH FRPSHQGLXP E\ *HOR f 3V\FKRORJ\ &KDVH DQG 6LPRQ f KDYH H[WHQGHG WKH UHVHDUFK SHUIRUPHG E\ GH *URRW ZKLFK FODLPV WKDW WKH GHSWK RI JDPH WUHH VHDUFK DQG WKH WRWDO QXPEHU RI PRYHV FRQVLGHUHG LV DSSUR[LPDWHO\ WKH VDPH IRU ERWK QRYLFHV DQG FKHVV PDVWHUV +RZHYHU WKH DELOLW\ WR UHFDOO FKHVV SRVLWLRQV DIWHU D ILYH VHFRQG SUHVHQWDWLRQ RI WKH SRVLWLRQV LV PDUNHGO\ GLIIHUHQW &KHVV PDVWHUV DUH DEOH WR UHFRQVWUXFW FKHVV SRVLWLRQV LQYROYLQJ WZHQW\ RU PRUH SLHFHV ZLWK FRPSOHWH DFFXUDF\ &KHVV QRYLFHV DUH RQO\ DEOH WR SODFH

PAGE 47

IRXU RU ILYH RI WKH SLHFHV FRUUHFWO\ 7KH GLIIHUHQFH EHWZHHQ WKH PDVWHUV DQG WKH QRYLFHV GLVDSSHDUV ZKHQ WKH VXEMHFWV DUH DVNHG WR UHSURGXFH UDQGRP SDWWHUQV RI FKHVV SLHFHV (DFK JURXS FDQ RQO\ SODFH WKUHH RU IRXU RI WKH SLHFHV IURP WKH UDQGRP ERDUG FRQILJXUDWLRQV 7KH DSSDUHQW GHFUHDVH LQ DELOLW\ E\ WKH PDVWHU OHYHO SOD\HUV LV DWWULEXWHG WR WKH IDFW WKDW WKH UDQGRP FRQILJXUDWLRQV RI SLHFHV FRQWDLQHG QR LQKHUHQW GRPDLQ NQRZOHGJH %DVHG RQ 0LOOHUfV f K\SRWKHVLV WKH FKHVV PDVWHUV DUH IRUPLQJ KLJKHU OHYHO FKXQNV FRPSRVHG RI JURXSV RI SLHFHV IURP WKH JDPH SRVLWLRQV 7KHVH FKXQNV FRQWDLQ GRPDLQ VSHFLILF NQRZOHGJH IURP WKH SULRU H[SHULHQFH RI WKH FKHVV PDVWHUV &KDVH DQG 6LPRQ f UHSRUW WKDW KXPDQ FKHVV PDVWHUV VWRUH IURP WR FKHVV SDWWHUQV 7KH VNLOO RI FKHVV PDVWHUV LQFUHDVHV DV PRUH SDWWHUQV DUH DFTXLUHG 1RYLFH FKHVV SOD\HUV PRYH DZD\ IURP VORZ GHGXFWLYH UHDVRQLQJ WRZDUGV WKH IDVW SHUFHSWXDO SURFHVVLQJ XVHG E\ PDVWHUV DV PRUH DQG PRUH SDWWHUQV DUH OHDUQHG &KDVH t 6LPRQ f 7KH FKXQNLQJ WKHRU\ RI OHDUQLQJ /DLUG HW DO f VXSSRUWV WKLV YLHZSRLQW LQ VWDWLQJ WKDW SHUIRUPDQFH LPSURYHV YLD WKH DFTXLVLWLRQ RI NQRZOHGJH DERXW SDWWHUQV DQG FKXQNV 5LFKDUG 6HOW]HU WKH IDWKHU RI IRXUWHHQ \HDU ROG FKHVV PDVWHU %REE\ 6HOW]HU LQ D SHUVRQDO FRPPXQLFDWLRQ UHODWHV WKDW %REE\fV LQFUHDVH LQ VNLOO LV UHIOHFWHG E\ DQ LPSURYHG SHUFHSWLRQ RI WKH ERDUG SRVLWLRQ DQG VXEVHTXHQW SRVLWLRQ HYDOXDWLRQ %UDWNR HW DO f VWDWH WKDW QDWXUDO SRVLWLRQDO PRYHV DUH FORVHO\ UHODWHG WR WKH FKXQN VWUXFWXUHV RI D JDPH SRVLWLRQ 6SHFLILF FKXQNV FDXVH FKHVV SOD\HUV WR JHQHUDWH WDFWLFDO DQG VWUDWHJLF SODQV 7KH FKXQN GLVSOD\HG LQ )LJXUH ZLOO FDXVH DQ

PAGE 48

DGYHUVDU\ WR LPPHGLDWHO\ FRQVLGHU SODQV IRU D EDFNURZ PDWH ZLWK HLWKHU WKH TXHHQ RU RQH RI WKH URRNV %DFNURZ PDWH SODQV UHVXOW IURP WKH NQRZOHGJH FRQWDLQHG LQ WKH FKXQN FRQFHUQLQJ WKH RSSRQHQWfV NLQJfV OLPLWHG PRELOLW\ FRQILJXUDWLRQV 6LPRQ DQG *LOPDUWLQ f GHYHORSHG 0$33 ZKLFK UHFRQVWUXFWV FKHVV SRVLWLRQV E\ VLPXODWLQJ KXPDQ H\H PRYHPHQWV UHFRUGHG LQ WKHLU VWXG\ 6LPRQ DQG *LOPDUWLQfV UHVHDUFK VKRZV WKDW KXPDQV WHQG WR JURXS RU FKXQN WRJHWKHU SLHFHV ZKLFK VKDUH JHRPHWULF SUR[LPLW\ :H ZLOO XVH JHRPHWULF SUR[LPLW\ DV WKH SULPDU\ ILOWHU IRU VHOHFWLQJ FKHVV SLHFHV WR EH IRUPHG LQWR FKXQNV 3UR[LPLW\ LV RQH RI WKH *HVWDOW SULQFLSOHV RI RUJDQL]DWLRQ $QGHUVRQ f DORQJ ZLWK VLPLODULW\ JRRG FRQWLQXDWLRQ DQG V\PPHWU\ :H ZLOO DOVR XWLOL]H WKH *HVWDOW SULQFLSOHV RI VLPLODULW\ DQG JRRG FRQWLQXDWLRQ WR FRQVWUXFW FKXQNV &KDVH DQG 6LPRQ f IRXQG WKDW SUR[LPLW\ DQG SLHFH FRORU RU VLPLODULW\ ZHUH WKH WZR VWURQJHVW IDFWRUV DIIHFWLQJ FKXQN SDWWHUQ UHFDOO E\ WKHLU FKHVV PDVWHU VXEMHFWV 7KH VLJQLILFDQW XVH RI SDWWHUQV RI JDPH SLHFHV E\ FKHVV PDVWHUV KDV OHG XV WR PDNH WKH IROORZLQJ WKUHH SDUW K\SRWKHVLV

PAGE 49

f &KHVV PDVWHUV DFTXLUH VHYHUDO WKRXVDQG SDWWHUQV RI SLHFHV WKDW DUH XVHG WR HYDOXDWH D JDPH SRVLWLRQ f &KHVV PDVWHUV ZLOO FRQWLQXH WR XVH D VSHFLILF VWUDWHJ\ ZKLOH ZLQQLQJ JDPHV DQG WRXUQDPHQWV f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f DQG WKH FKRLFH WDVN /DLUG HW DO ff ZKLFK LQGLFDWH WKDW KXPDQ VXEMHFWV ZLOO FRQWLQXH WR XVH D OHDUQHG VWUDWHJ\ IRU VROYLQJ D FODVV RI SUREOHPV HYHQ ZKHQ D PRUH HIILFLHQW VWUDWHJ\ ZRXOG DOVR VROYH D VSHFLILF SUREOHP :KLOH WKH WKLUG SDUW RI RXU K\SRWKHVLV LV LQWXLWLYHO\ DSSHDOLQJ GLUHFW VXSSRUW FRPHV IURP WKH UHVHDUFK RI +RUJDQ HW DO f DQG 6DDULOXRPD f ZKLFK GHPRQVWUDWHV WKH XVH RI FKXQNLQJ DQG DQDORJ\ WR VLPLODU SRVLWLRQV E\ FKHVV PDVWHUV LQ VROYLQJ FKHVV SUREOHPV 7KH FKXQNLQJ RI VHYHUDO SLHFHV LQWR D NQRZQ SDWWHUQ SURGXFHV D FRJQLWLYH HFRQRP\ ZKLOH HYDOXDWLQJ FRPSOH[ ERDUG SRVLWLRQV

PAGE 50

7KH VHDUFK FRPSOH[LW\ RI JDPH WUHHV LV UHGXFHG DV D UHVXOW RI WKH DSSOLFDWLRQ RI RXU K\SRWKHVLV 7KH VWUDWHJLF DQG WDFWLFDO GHFLVLRQV RI DQ DGYHUVDU\ FDQ EH DFFXUDWHO\ SUHGLFWHG EHFDXVH WKH DGYHUVDU\ ZLOO WHQG WR UHFUHDWH IDPLOLDU SDWWHUQV WR UHGXFH WKH FRJQLWLYH FRPSOH[LW\ RI HYDOXDWLQJ WKH FXUUHQW JDPH SRVLWLRQ $ SUDFWLFDO H[DPSOH RI WKH XVH RI IDPLOLDU SDWWHUQV LV GHPRQVWUDWHG LQ WKH %XGDSHVW WRXUQDPHQW JDPH EHWZHHQ 3LQWHU DQG %URQVWHLQ 5R\FURIW f 7KLV JDPH UHVXOWHG LQ D .1.%% HQGJDPH &KHVV OLWHUDWXUH KDG SUHYLRXVO\ FODLPHG WKDW D SRVLWLRQ NQRZQ DV WKH .OLQJ DQG +RURZLW] SRVLWLRQ .E1Ef ZLOO UHVXOW LQ D GUDZ IRU WKH .1.%% HQGJDPH 3LQWHUfV NQRZOHGJH RI WKH .OLQJ DQG +RURZLW] SRVLWLRQ OHG KLP WR FRQVLVWHQWO\ PDQHXYHU WR FUHDWH WKLV SRVLWLRQ 3LQWHU VXFFHHGHG LQ IRUPLQJ WKH .OLQJ DQG +RURZLW] SRVLWLRQ WKUHH GLIIHUHQW WLPHV LQ WKUHH GLIIHUHQW FRUQHUV RI WKH ERDUG +RZHYHU 3LQWHUfV KDVWH WR FUHDWH WKH FKXQN FDXVHG KLP WR PDNH LQIHULRU PRYHV QHDUO\ WKLUW\ SHUFHQW RI WKH WLPH DFFRUGLQJ WR WKH SRVLWLRQ %%1 GDWDEDVH 5R\FURIW f ZKLFK H[KDXVWLYHO\ VROYHV WKH .1.%% HQGJDPH LQ DQ RSWLPDO QXPEHU RI PRYHV &RPSXWHU FKHVV SURJUDPV QHHG WR VLPXODWH WKH FRJQLWLYH HFRQRP\ RI FKHVV PDVWHUV WR FRQWLQXH WR LPSURYH LQ SHUIRUPDQFH /HQDW HW DO f VWDWH WKDW LQWHOOLJHQW V\VWHPV PXVW SHUIRUP H[SHFWDWLRQ ILOWHULQJ XVLQJ SUHGLFWLRQV WR ILOWHU XQVXUSULVLQJ GDWD 6XFK D ILOWHULQJ SURFHVV LV SURYLGHG LQ DGYHUVDULDO GRPDLQV E\ WKH DELOLW\ WR SUHGLFW DGYHUVDU\ PRYHV LQ VSHFLILF VLWXDWLRQV &RPSXWHU &KHVV DQG $UWLILFLDO ,QWHOOLJHQFH 7KH ILUVW FKHVV FRPSXWHU ZDV WKH DXWRPDWRQ FRQVWUXFWHG E\ 7RUUHV \ 4XHYHGR FLUFD %HOO f 7RUUHV \ 4XHYHGRfV FKHVV PDFKLQH ZDV FDSDEOH RI

PAGE 51

DFFXUDWHO\ SOD\LQJ FHUWDLQ YDULDWLRQV RI WKH .5. .LQJ DQG 5RRN YHUVXV .LQJ HQGJDPH 6LQFH &ODXGH 6KDQQRQfV f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fV W\SH $ FKHVV SURJUDP &+(66; )UH\ f ZHUH ; LV D YHUVLRQ QXPEHU UHLJQHG GXULQJ WKH V DV WKH 1RUWK $PHULFDQ FRPSXWHU FKHVV SURJUDP FKDPSLRQ DQG ZDV WKH :RUOG &RPSXWHU &KHVV SURJUDP FKDPSLRQ IURP XQWLO .RSHF t 1HZERUQ f 7KH WKUHH W\SH $ SURJUDPV %(//( +,7(&+ DQG '((3 7+28*+7 KDYH HDFK VHW WKH KLJK ZDWHU PDUN IRU FKHVV SURJUDP SOD\LQJ VWUHQJWK DW 86&) UDWLQJV RI DQG UHVSHFWLYHO\ 7KHVH WKUHH SURJUDPV KDYH DOVR EHHQ WKH 1RUWK $PHULFDQ FRPSXWHU FKHVV SURJUDP FKDPSLRQV IURP WR WKH SUHVHQW 1HZERUQ t .RSHF f 7KH FKDUW LQ )LJXUH VKRZV WKH SURJUHVV LQ SOD\LQJ VWUHQJWK RI FKHVV SURJUDPV ZLWK UHVSHFW WR VHDUFK VSHHGV 7KH ERWWRP D[LV LV ORJDULWKPLF ZKLFK PHDQV WKDW WKH OLQHDU SURJUHVVLRQ LQ VWUHQJWK RI FKHVV SURJUDPV GXULQJ WKH V ZDV DFFRPSOLVKHG E\ DQ H[SRQHQWLDO LQFUHDVH LQ VHDUFK VSHHGV 5HFHQW SURJUHVV GHSLFWHG E\ WKH GDVKHG

PAGE 53

FXUYH UHIOHFWV WKH K\SHUH[SRQHQWLDO LQFUHDVH LQ VHDUFK VSHHGV UHTXLUHG WR PDLQWDLQ WKH OLQHDU JURZWK LQ SOD\LQJ DELOLW\ %(//( +,7(&+ DQG '((3 7+28*+7 HDFK XVH VSHFLDOL]HG K\EULG DUFKLWHFWXUHV WR DFKLHYH WKH GUDPDWLF LQFUHDVHV LQ VHDUFK VSHHG GHSLFWHG LQ )LJXUH (EHOLQJ f +,7(&+ XVHV D 9/6, DUFKLWHFWXUH ZLWK ILQH JUDLQHG SDUDOOHOLVP WR SHUIRUP SDWWHUQ PDWFKLQJ ZKLFK LV FRJQLWLYHO\ VLPLODU WR WKH SDWWHUQ UHFRJQLWLRQ SHUIRUPHG E\ KXPDQ FKHVV PDVWHUV '((3 7+28*+7 1HZERUQ t .RSHF f XVHV VSHFLDO 9/6, FLUFXLWU\ WR DFKLHYH D VHDUFK UDWH RI SRVLWLRQ QRGHV SHU VHFRQG $OWKRXJK W\SH $ VWUDWHJ\ FKHVV SURJUDPV DUH WKH EHVW SHUIRUPLQJ FKHVV SURJUDPV WKHLU SHUIRUPDQFH DJDLQVW JUDQGPDVWHU UDWHG KXPDQ SOD\HUV KDV EHHQ DE\VPDO ,Q DQ H[KLELWLRQ URXQG URELQ WRXUQDPHQW SOD\HG LQ 2FWREHU DW %RVWRQ IRXU KXPDQ JUDQGPDVWHUV FUXVKHG IRXU RI WKH WRS FRPSXWHU FKHVV SURJUDPV LQFOXGLQJ +,7(&+ DQG '((3 7+28*+7 +DPLOWRQ f ZLWK D VFRUH RI WR :RUOG &KHVV &KDPSLRQ *DU\ .DVSDURY GHIHDWHG '((3 7+28*+7 LQ D WZR JDPH PDWFK SOD\HG LQ 1HZ
PAGE 54

SUREOHPV XVHG LQ %UDWNR DQG 0LFKLHfV UHVHDUFK +RZHYHU $/ ZKLFK XVHV D W\SH % KHXULVWLF PHWKRG ZLWK GRPDLQ NQRZOHGJH ZDV DEOH WR VROYH DOO RI WKH HQGJDPH SUREOHPV 7\SH % VWUDWHJLHV YDU\ LQ WKHLU DSSURDFKHV WR SOD\LQJ FKHVV 7KH ILUVW KHXULVWLF PHWKRG XVHV SODQV RU JRDOV WKDW KDYH EHHQ FRQVWUXFWHG IURP GRPDLQ VSHFLILF H[SHUWLVH WR VROYH FKHVV SUREOHPV $/ LV D JRDO EDVHG FKHVV SURJUDP ZKLFK VROYHV .5.1 HQGJDPHV /(%/ 7DGHSDOOL f XVHV JRDOV GXULQJ .LQJ3DZQ HQGJDPHV WR UHGXFH VHDUFK FRPSOH[LW\ :LONLQV DQG f DOVR XVHV GRPDLQ VSHFLILF NQRZOHGJH WR SURGXFH SODQV ZKLFK KLV SURJUDP 3$5$',6( WULHV WR DFKLHYH *RDO EDVHG SODQQLQJ DOORZV 3$5$',6( WR DFFXUDWHO\ SOD\ TXLHVFHQW PLGGOH JDPH SRVLWLRQV 7KH XVH RI JRDOV WR DFKLHYH KLJKHU SOD\LQJ SHUIRUPDQFH LQ FKHVV SURJUDPV VLPXODWHV WKH FRJQLWLYH PHWKRG RI PHDQVHQGV DQDO\VLV XVHG E\ KXPDQV LQ FHUWDLQ SUREOHP VROYLQJ WDVNV $QGHUVRQ f $QRWKHU KHXULVWLF DSSURDFK ZKLFK LV VLPLODU WR WKH FRJQLWLYH PHWKRGV DQDO\]HG E\ &KDVH DQG 6LPRQ GHVFULEHG LQ VHFWLRQ VLPXODWHV WKH KXPDQ FRJQLWLYH SURFHVV RI FKXQNLQJ 7KLV KHXULVWLF DSSURDFK WULHV WR LGHQWLI\ FRPPRQ FRQILJXUDWLRQV RU FKXQNV DQG WKHQ XVHV WKHVH FKXQNV WR LGHQWLI\ DQDORJRXV VROXWLRQV WR FKHVV SUREOHPV 4XLQODQfV ,' SURJUDP LQ 0LFKDOVNL HW DO f WULHV WR LQGXFWLYHO\ FODVVLI\ HTXLYDOHQW ERDUG FRQILJXUDWLRQV LQ WKH .5.1 HQGJDPH 7KH ,' LQGXFHG SDWWHUQV DUH XVHG WR GHWHUPLQH LI WKH .LQJ5RRN VLGH KDV D ZLQQLQJ SRVLWLRQ ZLWKLQ D VSHFLILHG QXPEHU RI PRYHV &XUUHQWO\ WKH SURJUDP KDV VROYHG WKH .5.1 HQGJDPH IRU DOO SRVLWLRQV ZKLFK FDQ EH ZRQ LQ WKUHH SO\ &DPSEHOOfV f &+81.(5 DOVR

PAGE 55

WDNHV DGYDQWDJH RI ERDUG FRQILJXUDWLRQV WR SURGXFH FKXQNV ZKLFK DUH XVHG WR VROYH .LQJ3DZQ HQGJDPHV (DFK RI WKH KHXULVWLF PHWKRGV GHVFULEHG DERYH KDYH EHHQ LPSOHPHQWHG RQ HQGJDPH RU PLGGOH JDPH SRUWLRQV RI FKHVV SUREOHPV DQG GHDO SULPDULO\ ZLWK WDFWLFDO VROXWLRQV WR FKHVV SUREOHPV 7KH NQRZOHGJH DYDLODEOH WR FKHVV SURJUDPV FDQ EH VLJQLILFDQWO\ DXJPHQWHG E\ H[WHQGLQJ WKH UHVHDUFK RI 4XLQODQ DQG &DPSEHOO LQ WKH IROORZLQJ ZD\V f &ROOHFW FKXQNV IURP FRPSOHWH FKHVV JDPHV LQVWHDG RI MXVW WKH HQGJDPH VHJPHQW f 8VH WKH FKXQNV RI SLHFHV WR SUHGLFW DQ DGYHUVDU\fV WDFWLFDO DQG VWUDWHJLF PRYHPHQW GHFLVLRQV -XVWLILFDWLRQ IRU &KHVV 'RPDLQ $SSOLFDWLRQ 6KDSLUR f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

PAGE 56

LV QRW D IHDVLEOH VROXWLRQ 7\SLFDO FKHVV JDPH WUHHV FRQVLVW RI QRGHV %HUOLQHU t *RHWVFK f ZKLFK SURKLELWV H[KDXVWLYH VHDUFK VROXWLRQV JLYHQ FXUUHQW WHFKQRORJLFDO FDSDELOLWLHV %HFDXVH FKHVV LV D ZHOO XQGHUVWRRG GRPDLQ DQG WKHUH LV DQ DFFHSWHG QXPHULFDO UDWLQJ VFDOH SHUIRUPDQFH LPSURYHPHQWV FRQWULEXWHG E\ WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ FDQ EH HYDOXDWHG

PAGE 57

&+$37(5 $'9(56$5< 02'(/,1* 0(7+2'2/2*< ,Q WKLV FKDSWHU ZH H[DPLQH LQ GHWDLO HDFK RI WKH IXQFWLRQV UHTXLUHG WR FUHDWH DQ DGYHUVDU\ PRGHO IURP WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ 7KH IRXQGDWLRQ IRU RXU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ LV WKH WKUHH SDUW K\SRWKHVLV f 5HOHYDQW SDWWHUQV DQG SODQV DUH OHDUQHG E\ WKH DGYHUVDU\ IURP SUDFWLFDO H[SHULHQFH LQ WKH GRPDLQ f 7KH VWUDWHJLF DQG WDFWLFDO DFWLRQV RI DQ DGYHUVDU\ DUH UHSHDWHG DV ORQJ DV WKH DFWLRQV SURGXFH D SRVLWLYH UHVXOW f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

PAGE 58

.QRZOHGJH $FTXLVLWLRQ RI &KXQNV )RU FKHVV WKH SDWWHUQV LQYROYH ERWK WKH JHRPHWULF SDWWHUQV RI SLHFHV DQG WKH WH[WXDO SDWWHUQV UHODWHG WR WKH DFWXDO PRYHV WKDW DUH H[HFXWHG 7KH JHRPHWULF SDWWHUQV FRUUHVSRQG WR FKXQNV WKDW DUH VWRUHG LQ YLVXDO PHPRU\ ZKLOH WKH WH[WXDO SDWWHUQV FRUUHVSRQG WR FKXQNV WKDW DUH VWRUHG LQ WKH DGYHUVDU\f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f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

PAGE 59

VDWLVI\ WKH SULQFLSOH RI VLPLODULW\ 5HTXLULQJ FKXQNV WR EH RI D VLQJOH FRORU QDPHO\ WKH SLHFHV SOD\HG E\ WKH DGYHUVDU\ UHVWULFWV WKH FRQWHQWV RI WKH FKXQN WR EH XQGHU GLUHFW FRQWURO RI WKH DGYHUVDU\ $Q DGYHUVDU\ FDQQRW FRQWURO WKH PRYH FKRLFHV RI DQ RSSRQHQW H[FHSW IRU LQWHUORFNHG SDZQ FKDLQV DQG WKHUHIRUH FDQQRW XVH D FKXQN ZLWK PL[HG FRORUV RI SLHFHV IRU ORQJ UDQJH VWUDWHJLF SODQQLQJ 95HVHDUFK SHUIRUPHG E\ &KXUFK DQG &KXUFK +ROGLQJ f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

PAGE 60

)LJXUH )URP WKHVH FKXQNV ZH FDQ VHH WKDW D VL[WHHQ VTXDUH DUHD LV VXIILFLHQW WR KROG DOO RI WKH FKXQNV IRXQG :H LQFUHDVHG WKH PD[LPXP VL]H RI GHILQHG FKXQNV WR ILYHE\ILYH DQG VL[E\VL[ WR PHDVXUH WKH DIIHFW RI D ODUJHU FKXQN VL]H 2XU UHVHDUFK LQGLFDWHG WKDW RQO\ RQH DGGLWLRQDO FKXQN ZDV FDSWXUHG GXULQJ DQ DQDO\VLV RI WZHQW\ JDPHV SOD\HG E\ %RWYLQQLN VHH 6HFWLRQ )LJXUH f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f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

PAGE 61

D % D D D D D $ % F D 5 5 D D D D D 5 %  t D 5 D e D D D e $ D $ D D 5 D D % $ D $ D % D D D $ e 5 D D $ % D e e 5 5 D D D )LJXUH &KXQNV DFTXLUHG IRU %RWYLQQLN

PAGE 62

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

PAGE 63

DQ LPDJH WR EH SURFHVVHG LQWR FKXQNV %\ YLHZLQJ WKH JDPH ERDUG DV DQ LPDJH ZH FRQVLGHU YLD DQDORJ\ SUHYLRXVO\ GHILQHG FRPSXWHU YLVLRQ WHFKQLTXHV IRU SURFHVVLQJ LPDJHV DQG FROOHFWLQJ LQIRUPDWLRQ 7KH FRQYROXWLRQ RSHUDWRU LV XVHG LQ WKH LPDJH DOJHEUD 5LWWHU HW DO f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fM? 7KH ILUVW VXEVFULSW GHQRWHV WKH FROXPQ DQG WKH VHFRQG VXEVFULSW GHQRWHV WKH URZ 2XU LQWHUQDO UHSUHVHQWDWLRQ LV VLPLODU WR WKH RQH GHVFULEHG E\ 6KDQQRQ f ZLWK QHJDWLYH YDOXHV DVVLJQHG WR QRQDGYHUVDU\ SLHFHV DQG WKH QXPEHUV RQH WKURXJK VL[ UHSUHVHQWLQJ WKH SDZQV NQLJKWV ELVKRSV URRNV TXHHQV DQG NLQJV UHVSHFWLYHO\ 8QRFFXSLHG VTXDUHV DUH GHQRWHG E\ WKH YDOXH ]HUR $ ERDUG SRVLWLRQ DQG LWV FRUUHVSRQGLQJ LQWHUQDO UHSUHVHQWDWLRQ DUH GHPRQVWUDWHG LQ )LJXUH 7KH SLHFH YDOXH DW HDFK ERDUG ORFDWLRQ LV GHQRWHG E\ WKH DOJHEUDLF V\PERO

PAGE 64

M L )LJXUH ,QWHUQDO UHSUHVHQWDWLRQ RI D ERDUG SRVLWLRQ 7KH WKUHH GLIIHUHQW WHPSODWHV XVHG LQ RXU LPSOHPHQWDWLRQ RI WKH FRQYROXWLRQ DOJRULWKP DUH VKRZQ LQ )LJXUH (DFK WHPSODWH FDQ EH WKRXJKW RI DV D WKUHHE\ WKUHH DUUD\ ZLWK WKH FHQWHU ORFDWLRQ GHQRWHG E\ WKH VXEVFULSW f 2WKHU WHPSODWH VXEVFULSWV UDQJH IURP QHJDWLYH RQH WR RQH 7KH JHQHUDO WHPSODWH LQ )LJXUH LV XVHG IRU ELVKRSV TXHHQV DQG WKH NLQJ ZLWK WKH RWKHU SLHFHV EHLQJ FRQYROYHG E\ WKH DSSURSULDWHO\ QDPHG WHPSODWH 7KH SVHXGRFRGH LQ )LJXUH GHPRQVWUDWHV KRZ WKH FRQYROXWLRQ RSHUDWLRQ LV SHUIRUPHG :KHQ DQ DGYHUVDU\fV SLHFH LV GHWHFWHG WKH DSSURSULDWH WHPSODWH LV ORDGHG LQWR D FRS\ RI WKH FXUUHQW WHPSODWH73 7KH YDOXHV GHSLFWHG LQ WKH WHPSODWHV ZHUH D SUR[LPDO SLHFH LV ORFDWHG DUH VXPPHG DQG VWRUHG LQ WKH UHWXUQ DUUD\ < XVLQJ DQ LQFOXVLYH RU RSHUDWRUp :H KDYH GHILQHG WKH SUHGLFDWH IXQFWLRQ 21 %2$5' WR YHULI\ WKDW WKH DUUD\ VXEVFULSWV fL 0f DQG fM 1f FRUUHVSRQG WR D YDOLG ERDUG ORFDWLRQ 7KH UHWXUQ DUUD\ < LV DQ HLJKWE\HLJKW FRS\ RI WKH LQWHUQDO ERDUG UHSUHVHQWDWLRQ ZLWK DOO YDOXHV LQLWLDOO\ VHW WR ]HUR 7KLV

PAGE 65

*HQHUDO 3DZQ )LJXUH &RQYROXWLRQ WHPSODWHV D 5RRN .QLJKW &DVH ;M M RI 73 WKH JHQHUDO WHPSODWH 6HH )LJXUH f 73 WKH SDZQ WHPSODWH 73 WKH URRNNQLJKW WHPSODWH RWKHUZLVH 5HWXUQ 1RQDGYHUVDU\ SLHFH QR QHHG WR FRQYROYH (QG &DVH )RU 1 WR GR )RU 0 WR GR ,I 21B%2$5'L 1 M 0f DQG ;L1f M0f 7KHQ %HJLQ <\ <\ p 73Q P
PAGE 66

WKHQ WKH YDOXH DW ;[L ZRXOG EH VHW WR IRXU DQG WKH YDOXH RI ;X ZRXOG EH FKDQJHG IURP IRXU WR ]HUR &XUUHQWO\ ZH XVH D PRGLILHG DOJHEUDLF QRWDWLRQ IRU UHFRJQL]LQJ WKH PRYHV SOD\HG GXULQJ WKH JDPH 2WKHU QRWDWLRQV FDQ EH HDVLO\ WUDQVODWHG WR DOJHEUDLF QRWDWLRQ IRU XVH E\ WKH ,$0 SURJUDP 7KH FKXQNV ZH GHVLUH WR FDSWXUH DUH WKH RQHV WKDW KDYH MXVW EHHQ IRUPHG E\ WKH DGYHUVDU\ ZKLFK LPSURYHV RXU FDSDELOLW\ IRU SUHGLFWLQJ GHILQLWH DGYHUVDU\ DFWLRQV 7KHUHIRUH ZH FROOHFW FKXQNV IROORZLQJ WKH DGYHUVDU\f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

PAGE 67

)RU WR GR )RU WR GR ,I JDWKHULQJ VWDQGDUG JHRPHWULF FKXQNV $1' ; M 7KHQ
PAGE 68

WLPHV RI D FKXQNfV LQFHSWLRQ DQG VXEVHTXHQW GLVVROXWLRQ DUH VWRUHG ZLWK HDFK FKXQN :H DOVR XVH WKH WLPH WDJV DVVRFLDWHG ZLWK HDFK FKXQN WR HOLPLQDWH VWRULQJ D FKXQN ZKLFK LV GLVVROYHG E\ WKH DGYHUVDU\ DQG WKHQ UHFUHDWHG DW D ODWHU WLPH GXULQJ WKH VDPH JDPH 3VHXGRFRGH WR SHUIRUP WKH FKXQN FRPSDULVRQV ZKLFK SURGXFH WKH VWRUDJH VDYLQJV MXVW PHQWLRQHG LV VKRZQ LQ )LJXUH 7KH YDULDEOH f& UHSUHVHQWV WKH GRXEOH OLQNHG OLVW RI FKXQN UHFRUGV ZLWK WKH VXEVFULSW 7 GHQRWLQJ WKH SRVLWLRQ LQ WKH OLVW DQG WKH YDULDEOH f'f UHSUHVHQWV WKH PRVW UHFHQW FKXQN WKDW KDV EHHQ LGHQWLILHG IRU WKH FXUUHQW JDPH )RU WR WKH WRWDO QXPEHU RI FKXQNV FUHDWHG IRU WKLV JDPH GR ,I & 7KHQ %HJLQ ,I 'VWDUWBWLPH &VWDUWBWLPH 7KHQ 4VWDUWBWLPH 'VWDUWBWLPH ,I 'HQGBWLPH &>HQGBWLPH 7KHQ &MHQGBWLPH m 'HQGBWLPH ([LW WKH ORRS (QG %HJLQ )LJXUH 3VHXGRFRGH WR SHUIRUP FKXQN FRDOHVFLQJ $Q H[DPSOH RI WKH FKXQNV LGHQWLILHG E\ WKH FRQYROXWLRQ RSHUDWRU DQG WKH FRQYROXWLRQ YDOXHV UHWXUQHG LV VKRZQ LQ )LJXUH ZKHUH WKH FKXQNV KDYH EHHQ DFTXLUHG IRU WKH ZKLWH SOD\HU 7KH SRVLWLRQ VKRZQ LQ )LJXUH LV WDNHQ IURP WKH ILIWK JDPH RI WKH :RUOG &KDPSLRQVKLS PDWFK EHWZHHQ %RWYLQQLN DQG %URQVWHLQ DIWHU ZKLWHfV HLJKWHHQWK PRYH 7KH WKUHH FKXQNV DFTXLUHG E\ WKH FRQYROXWLRQ

PAGE 69

‘ M PL / LLLLLLL 9 99 n9 LOO + 5 ,,, A P L D Z rrL___LL r,rr,rrr,r, rrnrrrr ,r,r ,rrrr nLnVO 6f6 rc M OMLMM MOMLMOL r mf` -( __rLMc OLMMMOMO P ,,,) L J ,,, L_ 6c 66LMrr ‘ L 0 5 LQ = /LG / %ODQNV )LJXUH &KXQNV RQ D FKHVV ERDUG DQG WKHLU FRQYROXWLRQ YDOXHV

PAGE 70

DOJRULWKP DUH RXWOLQHG LQ WKH ILJXUH 7KH SDZQ VWUXFWXUH LQ WKH PLGGOHOHIW RI WKH ERDUG FRXOG EHFRPH D VWURQJ IDFWRU LQ WKH HQGJDPH DQG SRVVLEO\ FUHDWH D SDVVHG SDZQ 7KLV NQRZOHGJH LV DYDLODEOH IURP WKH JHRPHWULF UHODWLRQVKLS RI WKH SLHFHV LQ WKH FKXQN 7KH RWKHU WZR FKXQNV DUH D VWDQGDUG RIIHQVLYH FRQILJXUDWLRQ LH WKH TXHHQ DQG URRNf DQG D VWDQGDUG GHIHQVLYH FRQILJXUDWLRQ LH WKH ILDQFKHWWRHG ELVKRS DQG FDVWOHG NLQJf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fV PRXWKV DUH DOVR DFTXLUHG E\ WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\fV FRQYROXWLRQ DOJRULWKP $FTXLULQJ 7H[WXDO &KXQNV 7KH EDFNJURXQG UHVHDUFK SUHVHQWHG LQ &KDSWHU LQGLFDWHV WKDW FKHVV PDVWHUV PDNH XVH RI WKHLU YHUEDO PHPRU\ WR UHFDOO VSHFLILF RSHQLQJ VHTXHQFHV RI SOD\ 2XU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ DFTXLUHV WKLV IDFHW RI H[SHUW FKHVV SHUIRUPDQFH LQ DGGLWLRQ WR WKH JHRPHWULF SLHFH FKXQNV &KHVV PDVWHUV FDQ UHFDOO YHUEDWLP SDUWLFXODU RSHQLQJ VHTXHQFHV RI SOD\ ZKLFK LQGLFDWHV WKH XVH RI URWH OHDUQLQJ :H HPXODWH WKH FRJQLWLYH SURFHVV RI URWH OHDUQLQJ

PAGE 71

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f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

PAGE 72

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fV PRYH FKRLFHV 7KHUHIRUH LI DQ LGHQWLFDO PRYH LV PDGH E\ DQ DGYHUVDU\ DV ERWK D ZKLWH SLHFH DQG EODFN SLHFH SOD\HU WKHQ WKH WZR PRYHV DUH FRQVLGHUHG DV GLIIHUHQW PRYHV DQG DUH VWRUHG VHSDUDWHO\ 2WKHU W\SHV RI PRYHV DUH XVHIXO IRU SUHGLFWLQJ WKH SOD\LQJ VW\OH RI DQ DGYHUVDU\ :H VDYH WKH QXPEHU RI SDZQ PRYHV WKDW RFFXU GXULQJ WKH ILUVW WHQ PRYHV RI HDFK JDPH DQG WKH QXPEHU RI DWWDFNLQJ PRYHV PDGH GXULQJ WKH ILUVW WHQ DQG WZHQW\ PRYHV RI HDFK JDPH WR DLG LQ LQIHUULQJ WKH SOD\LQJ VW\OH RI DQ DGYHUVDU\ $Q\ RWKHU PRYH ZKLFK FDQ EH LGHQWLILHG WH[WXDOO\ LV DOVR D FDQGLGDWH IRU WKH WH[WXDO PRYH NQRZOHGJH RI WKH DGYHUVDU\ PRGHO $ IORZFKDUW IRU WKH SURFHVV RI FROOHFWLQJ WH[WXDO EDVHG NQRZOHGJH LV VKRZQ LQ )LJXUH 7KH IUHTXHQF\ DQG GLUHFWLRQ RI FDVWOLQJ PRYHV DUH DOVR DFTXLUHG WR DVVLVW LQ SUHGLFWLRQ RI SRVVLEOH FDVWOH PRYHV GXULQJ DQ DGYHUVDU\fV IXWXUH JDPHV

PAGE 73

7R GHPRQVWUDWH WKH IORZFKDUW DVVXPH WKDW WKH PRYH FSDZQ WR TXHHQfV ELVKRS IRXULV PDGH E\ WKH DGYHUVDU\ RQ KLV IRXUWK JDPH WXUQ 6LQFH WKH PRYH RFFXUV GXULQJ WKH ILUVW ILYH PRYHV RI WKH JDPH WKH PRYH ZLOO EH VDYHG LQWR WKH WH[WXDO FKXQN NQRZOHGJH EDVH ,I WKH PRYH DOUHDG\ H[LVWV LQ WKH WH[WXDO FKXQN NQRZOHGJH EDVH WKHQ WKH QXPEHU RI WLPHV WKLV PRYH KDV EHHQ REVHUYHG LV LQFUHPHQWHG DQG WKH VWDWLVWLFV UHJDUGLQJ WKH WLPH RI WKH PRYH DUH XSGDWHG WR UHIOHFW WKH QHZ NQRZOHGJH %HFDXVH D SDZQWKH TXHHQfV ELVKRS SDZQLV EHLQJ PRYHG ZH LQFUHPHQW WKH FRXQW RI SDZQ PRYHV PDGH GXULQJ WKH ILUVW WHQ PRYHV RI WKH JDPH :H VDYH WKH WRWDO QXPEHU RI JDPHV DQDO\]HG WKH QXPEHU RI JDPHV WKH DGYHUVDU\ SOD\HG DV ZKLWH WKH QXPEHU RI JDPHV ZRQ E\ WKH DGYHUVDU\ WKH QXPEHU

PAGE 74

RI JDPHV ZRQ DV ZKLWH DQG WKH DYHUDJH OHQJWK RI DOO DQDO\]HG JDPHV 7KLV LQIRUPDWLRQ LV XVHG E\ WKH DGYHUVDU\ PRGHO WR GHWHUPLQH WKH IUHTXHQF\ RI D VSHFLILF PRYH DQG WKH OLNHOLKRRG WKDW WKH PRYH ZLOO EH UHSHDWHG E\ WKH DGYHUVDU\ 2WKHU DGYHUVDULDO GRPDLQV FDQ EHQHILW IURP WKH WH[WXDO NQRZOHGJH DFTXLVLWLRQ SHUIRUPHG E\ WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ )RU H[DPSOH WKH DFTXLVLWLRQ RI WH[WXDO SDWWHUQV ZLOO FDSWXUH VWDQGDUG RSHUDWLQJ SURFHGXUHV DQG SURWRFROV VXFK DV KRZ WR DFFHSW DQ HQHP\f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

PAGE 75

LQGXFWLYH SUHPLVH ZH UHTXLUH WKDW D PLQLPXP RI WZR DGYHUVDU\ JDPHV EH VWXGLHG SULRU WR SHUIRUPLQJ WKH LQGXFWLRQ 7KH LQGXFWLRQ DOJRULWKP UHTXLUHV FKXQNV WR EH LGHQWLFDO EHIRUH EHLQJ JURXSHG WRJHWKHU 2XU LQGXFWLRQ DOJRULWKP LV QRLVH LQWROHUDQW VLQFH HYHQ VPDOO YDULDWLRQV LQ D FKXQN FDQ KDYH D ODUJH HIIHFW RQ WKH VWUDWHJLF HYDOXDWLRQ YDOXH RI D FKXQN $Q H[DPSOH RI WKUHH FKXQNV ZKLFK RQO\ GLIIHU E\ WKH ORFDWLRQ RU YDOXH RI D VLQJOH SLHFH DQG ZKLFK KDYH VLJQLILFDQWO\ GLIIHUHQW VWUDWHJLF PHDQLQJV LV VKRZQ LQ )LJXUH 7KH ILUVW FKXQN VKRZV D VWURQJ SDZQ IRUPDWLRQ ZKLOH WKH VHFRQG FKXQN VKRZV WZR SDZQV EHLQJ DWWDFNHG E\ DQ RSSRQHQWf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

PAGE 76

FKXQNV :KHQ WKH DGYHUVDU\ LV SOD\LQJ WKH EODFN SLHFHV ZH UHIOHFW WKH SLHFHV WKURXJK WKH PLGERDUG KRUL]RQWDO D[LV WR JURXS FKXQNV WKDW RFFXU ZKLOH WKH DGYHUVDU\ LV SOD\LQJ HLWKHU FRORU RI SLHFHV &KXQNV FRQVLGHUHG WR EH LGHQWLFDO DV VKRZQ LQ )LJXUH DUH JURXSHG WRJHWKHU DQG VWRUHG LQ WKH NQRZOHGJH EDVH WKDW IRUPV WKH DGYHUVDU\ PRGHO 6rWr@LL6M mcM U377Yn7 L O r)9} m0 ‘ +, ‘ e IO LOO e IOrr )LJXUH )RXU LGHQWLFDO FKXQNV :H XVH D IUDPH EDVHG V\VWHP WR VWRUH WKH FKXQNV IRXQG E\ WKH LQGXFWLRQ DOJRULWKP LQ WKH DGYHUVDU\ PRGHOfV NQRZOHGJH EDVH 7KH VORWV LQ HDFK IUDPH DUH WKH DFWXDO FKXQN SLHFHV DQG UHODWLYH SRVLWLRQV RI WKH LQGXFHG FKXQNV WKH QXPEHU RI WLPHV WKLV FKXQN KDV EHHQ REVHUYHG WKH HDUOLHVW PRYH QXPEHU ZKHQ WKH FKXQN ZDV FUHDWHG WKH ODWHVW PRYH QXPEHU ZKHQ WKH FKXQN ZDV GLVVROYHG WKH RXWFRPH ZLQ RU ORVVf RI

PAGE 77

WKH JDPHV LQ ZKLFK WKH FKXQN RFFXUUHG DQG WKH FRORU RI WKH DGYHUVDU\fV SLHFHV $OWKRXJK ZH KDYH QRW LPSOHPHQWHG WKLV VORW FKXQNV ZKLFK DUH REVHUYHG IUHTXHQWO\ FDQ KDYH D EHVW PRYH VORW DGGHG ZKLFK ZRXOG PRGHO WKH VFULSWV XVHG E\ FKHVV PDVWHUV IRU FHUWDLQ FRQILJXUDWLRQV 7KH DGYHUVDU\ PRGHOfV NQRZOHGJH EDVH RI SDWWHUQV NQRZQ WR DQ DGYHUVDU\ FDQ EH FRQVWUXFWHG LQ SDUWV $W OHDVW WZR JDPHV QHHG WR EH SURFHVVHG E\ WKH FRQYROXWLRQ RSHUDWRU WR DFTXLUH SRWHQWLDO FKXQNV $IWHU WKH SRWHQWLDO FKXQNV KDYH EHHQ DFTXLUHG WKH LQGXFWLRQ OHDUQLQJ PHFKDQLVP SURFHVVHV WKH SRWHQWLDO FKXQNV RI HDFK JDPH E\ ILUVW FKHFNLQJ WR VHH LI WKH FKXQN DOUHDG\ H[LVWV LQ WKH DGYHUVDU\ PRGHOf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fV NQRZOHGJH EDVHV RI VSHFLILF

PAGE 78

DGYHUVDULHV DUH XVHG WR LQIHU WKH SOD\LQJ VW\OH RI DQ DGYHUVDU\ DQG WR SUHGLFW WKH WDFWLFDO DQG VWUDWHJLF PRYHPHQW FKRLFHV RI DQ DGYHUVDU\ $FTXLULQJ DQ $GYHUVDU\fV 3OD\LQJ 6W\OH .QRZLQJ WKH JHQHUDO SOD\LQJ VW\OH RI DQ DGYHUVDU\ HQDEOHV FKHVV SURJUDPV WR VHOHFW PRYHV WKDW ZLOO SODFH WKH DGYHUVDU\ LQ D GLVDGYDQWDJHRXV VWUDWHJLF SRVLWLRQ 7KH JHQHUDO SOD\LQJ VW\OH RI DQ DGYHUVDU\ LV LQIHUUHG IURP WKH FROOHFWLRQ RI SDWWHUQV FRQWDLQHG LQ WKH DGYHUVDU\ PRGHO NQRZOHGJH EDVH :H XVH D UXOHEDVHG LQIHUHQFH HQJLQH WR KHXULVWLFDOO\ GHWHUPLQH WKH SOD\LQJ VW\OH RI DQ DGYHUVDU\ &XUUHQWO\ ZH LQIHU WZR GLIIHUHQW VW\OHV RI SOD\ 7KH LQIHUHQFH HQJLQH LQGLFDWHV WKDW DQ DGYHUVDU\ SUHIHUV RSHQ SRVLWLRQV FORVHG SRVLWLRQV RU KDV QR REYLRXV SUHIHUHQFH 7KH FORVHG SRVLWLRQ VW\OH RI SOD\ LV IXUWKHU XVHG WR LPSO\ VWURQJ SRVLWLRQDO SOD\ E\ WKH DGYHUVDU\ /LNHZLVH DQ RSHQ SRVLWLRQ SOD\LQJ VW\OH LPSOLHV D VWURQJ WDFWLFDO SOD\HU 7KH UXOH EDVH KHXULVWLFV DGGUHVV WKH IROORZLQJ GRPDLQ VSHFLILF IDFWRUV ZKLFK LQGLFDWH DQ DGYHUVDU\fV SOD\LQJ VW\OH SUHIHUHQFH f 7KH SUHVHQFH DQG UHODWLYH VL]H RI SDZQ VWUXFWXUHV f 7KH QXPEHU RI SDZQ PRYHV PDGH GXULQJ WKH RSHQLQJ VHTXHQFH f 7KH SUHVHQFH RI SDZQ FKDLQV RU LQWHUORFNHG SDZQ VWUXFWXUHV f 7KH QXPEHU RI RSHQLQJ PRYHV WKDW DUH DWWDFNLQJPRYH D SLHFH DFURVV WKH PLGERDUG KRUL]RQWDO D[LV (DFK UXOH LV HYDOXDWHG WR GHWHUPLQH LI D VSHFLILF SOD\LQJ VW\OH LV GHPRQVWUDWHG 7KH

PAGE 79

UXOHV DUH QRW FRPSHWLWLYH EXW LQVWHDG WKH\ OHQG VXSSRUW WR WKH DGYHUVDU\ PRGHOfV EHOLHI WKDW D SDUWLFXODU SOD\LQJ VW\OH LV SUHIHUUHG E\ WKH DGYHUVDU\ E\ DGMXVWLQJ WKH OLNHOLKRRG YDOXH DWWDFKHG WR HDFK SOD\LQJ VW\OH 7KH SOD\LQJ VW\OH ZLWK WKH JUHDWHVW OLNHOLKRRG LV WKHQ LQIHUUHG DV WKH JHQHUDO SOD\LQJ VW\OH RI WKH DGYHUVDU\ 3DXQ 0RYHV 3DZQ 6WUXFWXUHV RI r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

PAGE 80

7KH KHXULVWLF UXOHV XVHG WR LQIHU DQ DGYHUVDU\fV SOD\LQJ VW\OH DUH QRW FRPSHWLWLYH QHZ KHXULVWLF UXOHV FDQ EH DGGHG DW DQ\ ORFDWLRQ RI WKH UXOH EDVH $V QHZ GRPDLQ VSHFLILF NQRZOHGJH EHFRPHV DYDLODEOH WKDW ZLOO LQIHU DGGLWLRQDO SOD\LQJ VW\OHV RU VXSSRUW WKH FXUUHQW SOD\LQJ VW\OH LQIHUHQFHV WKH FXUUHQW UXOH EDVH FDQ EH HDVLO\ DXJPHQWHG 7KH GHWHUPLQDWLRQ RI DQ DGYHUVDU\fV SOD\LQJ VW\OH SUHIHUHQFH KDV EHHQ DGGHG WR WKH JHQHUDOL]HG DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ WR LQFUHDVH WKH DGYHUVDU\ PRGHOfV SHUIRUPDQFH LQ WKH FKHVV GRPDLQ 7KH SOD\LQJ VW\OH LQIHUHQFH HQJLQH PDNHV XVH RI GRPDLQ VSHFLILF NQRZOHGJH LQ DGGLWLRQ WR WKH WH[WXDO DQG JHRPHWULF SDWWHUQV DQG IRU WKLV UHDVRQ LV QRW FRQVLGHUHG DV D PDLQ SDUW RI WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ 7KH SDWWHUQ DFTXLVLWLRQ LQGXFWLRQ OHDUQLQJ PHFKDQLVP DQG DGYHUVDU\ DFWLRQ SUHGLFWLRQ WRRO HDFK XVH RQO\ PLQLPDO DPRXQWV RI GRPDLQ VSHFLILF NQRZOHGJH VR WKH DGYHUVDU\ PRGHO LPSOHPHQWDWLRQ FDQ EH XVHG LQ RWKHU DGYHUVDULDO GRPDLQV ZLWK OLWWOH RU QR PRGLILFDWLRQV $SSOLFDWLRQ RI WKH $GYHUVDU\ 0RGHO 7KH DGYHUVDU\ PRGHO FRQVLVWV RI D NQRZOHGJH EDVH RI WH[WXDO DQG JHRPHWULF RU YLVXDO SDWWHUQV NQRZQ WR WKH DGYHUVDU\ 7KH WKLUG SDUW RI RXU K\SRWKHVLV SUHVHQWHG DW WKH EHJLQQLQJ RI WKLV FKDSWHU LQGLFDWHV KRZ ZH DUH JRLQJ WR XVH WKLV NQRZOHGJH $Q DGYHUVDU\ ZLOO DWWHPSW WR UHGXFH WKH FRPSOH[LW\ RI HYDOXDWLQJ WKH FXUUHQW GRPDLQ VLWXDWLRQ E\ PDQHXYHULQJ WKH ERDUG SRVLWLRQ WR FRQWDLQ RQH RU PRUH RI WKH SDWWHUQ FKXQNV ZLWK ZKLFK KH LV IDPLOLDU

PAGE 81

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fV JDPH FXUUHQWO\ LQ SURJUHVV LV VHDUFKHG SULRU WR HDFK PRYH WR EH PDGH E\ WKH DGYHUVDU\ XQWLO D SLHFH EHORQJLQJ WR WKH DGYHUVDU\ LV ORFDWHG 1H[W DOO RI WKH FKXQNV LQ WKH DGYHUVDU\ PRGHOfV NQRZOHGJH EDVH ZKLFK FDQ ILW LQWR WKH FXUUHQW ERDUG ORFDWLRQ DUH FRPSDUHG DJDLQVW WKH FXUUHQW ERDUG SDWWHUQ ,QFRQVLVWHQFLHV EHWZHHQ WKH FKXQNV LQ WKH NQRZOHGJH EDVH DQG WKH FXUUHQW ERDUG FRQILJXUDWLRQ DUH FRXQWHG $Q LQFRQVLVWHQF\ H[LVWV LI D ERDUG ORFDWLRQ FRUUHVSRQGLQJ WR WKH ORFDWLRQ RI D SLHFH

PAGE 82

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f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

PAGE 83

IURP WKH EDVH SUREDELOLW\ RI HDFK VXJJHVWHG PRYH 7KH PRYH ZLWK WKH KLJKHVW SUREDELOLW\ LV WKHQ XVHG WR SUHGLFW WKH DGYHUVDU\fV QH[W PRYH ,I WZR RU PRUH PRYHV KDYH QHDUO\ HTXDO SUREDELOLWLHV WKHQ WKH LQIHUHQFH HQJLQH PDNHV PXOWLSOH PRYH SUHGLFWLRQV &RPSXWHU FKHVV SURJUDPV FDQ XVH WKH SUREDELOLW\ YDOXH DVVRFLDWHG ZLWK HDFK PRYH DV D PHDVXUH RI EHOLHI :KHQ WKH DGYHUVDU\ PRGHO LQIHUV D SUREDELOLW\ IRU D PRYH WKDW LV OHVV WKDQ VRPH VSHFLILHG YDOXH WKH FKHVV SURJUDP FDQ FKRRVH WR LJQRUH WKH DGYHUVDU\ PRGHOfV SUHGLFWLRQ 7KH KHXULVWLFV XVHG LQ WKH SUHGLFWLRQ LQIHUHQFH HQJLQH DWWHPSW WR IROORZ VHYHUDO SV\FKRORJLFDO SULQFLSOHV RI FRJQLWLYH RUJDQL]DWLRQ DQG HFRQRP\ 7KH JHQHUDO GHVFULSWLRQ RI WKH SXUSRVH RI RXU KHXULVWLFV IROORZV f /DUJH FKXQNV DUH IDYRUHG RYHU VPDOO FKXQNV &KXQN VL]H LV GHSHQGHQW RQ WKH QXPEHU RI SLHFHV FRQWDLQHG LQ D FKXQN f &KXQNV FRQWDLQLQJ PDMRU SLHFHV DUH IDYRUHG RYHU FKXQNV FRQWDLQLQJ RQO\ SDZQV f )DYRU PRYHV ZKLFK KDYH EHHQ VXJJHVWHG E\ PRUH WKDQ RQH FKXQN f 5HGXFH WKH SUREDELOLW\ RI FKXQNV WKDW KDYH RQO\ EHHQ REVHUYHG LQ ORVW JDPHV f ,QFUHDVH WKH SUREDELOLW\ RI FKXQNV ZKRVH PRYH VXJJHVWLRQ FDXVHV D JDLQ LQ PDWHULDO DGYDQWDJH 7KLV KHXULVWLF LV RQO\ XVHG LQ TXLHVFHQW SRVLWLRQV GXH WR WKH YRODWLOH QDWXUH RI PDWHULDO DGYDQWDJH LQ QRQTXLHVFHQW SRVLWLRQVf f &KXQNV ZKLFK KDYH RFFXUUHG ZKLOH XVLQJ WKH VDPH FRORU SLHFHV DV WKH FXUUHQW JDPH DUH IDYRUHG RYHU FKXQNV WKDW KDYH RQO\ EHHQ REVHUYHG IRU WKH RSSRVLWH FRORU

PAGE 84

f $GMXVW WKH SUREDELOLW\ WR DFFRXQW IRU WHPSRUDO GLVFUHSDQFLHV ,I WKH FXUUHQW JDPH PRYH QXPEHU LV RXWVLGH RI WKH PRYH UDQJH GHILQHG E\ WKH WZR VORWV IRU WKH FKXQNfV WLPH RI LQFHSWLRQ DQG GLVVROXWLRQ WKHQ UHGXFH WKH SUREDELOLW\ SURSRUWLRQDO WR WKH GLVWDQFH EHWZHHQ WKH FXUUHQW PRYH DQG WKH FKXQNfV WHPSRUDO UDQJH f (OLPLQDWH VPDOOHU FKXQNV WKDW DWWHPSW WR ERUURZ SLHFHV IURP DQ H[LVWLQJ ODUJHU FKXQN f 5HGXFH WKH SUREDELOLW\ RI FKXQNV FRQWDLQLQJ RQO\ WZR SLHFHV GXULQJ WKH PLGGOH JDPH VHJPHQW f ,I WKH DGYHUVDU\ KDV MXVW GLVVROYHG D FKXQN WKHQ GR QRW UHFUHDWH WKH FKXQN LPPHGLDWHO\ f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

PAGE 85

7KH UROH RI WKH DGYHUVDU\ PRGHO DV D FRDFK QHHGV WR EH UHPHPEHUHG 3UHGLFWLRQV RI SUREDEOH DGYHUVDU\ DFWLRQV DUH RQO\ PDGH ZKHQ WKH FXUUHQW JDPH VLWXDWLRQ UHVHPEOHV GRPDLQ VLWXDWLRQV ZKLFK KDYH DOUHDG\ EHHQ OHDUQHG LQGXFWLYHO\ $IWHU DQDO\]LQJ QLQH DQG WKHQ WZHOYH JDPHV RI %RWYLQQLN WKH UHVXOWDQW DGYHUVDU\ PRGHOV ZHUH XVHG WR SUHGLFW WKH DFWLRQV RI %RWYLQQLN LQ D QHZ JDPH ZKLFK KDG QRW EHHQ SUHYLRXVO\ DQDO\]HG )RU HDFK RI PRGHOVf SUHGLFWLRQV RYHU IRUW\ SHUFHQW RI WKH QHZ JDPHf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

PAGE 86

0RYHV ZKLFK KDYH DOUHDG\ EHHQ PDGH LQ WKH FXUUHQW JDPH FDXVH WKH LGHQWLFDO PRYH LQ WKH NQRZOHGJH EDVH WR EH UHMHFWHG $GGLWLRQDOO\ ZKHQ WKH DGYHUVDU\ LV SOD\LQJ WKH EODFN SLHFHV ZH LPSOHPHQWHG UHVSRQVH FRGH LQ WKH SUHGLFWLRQ SURFHVV ZKLFK KHXULVWLFDOO\ IDYRUV WH[WXDO PRYH SDWWHUQV WKDW KDYH SUHYLRXVO\ IROORZHG WKH FXUUHQW PRYH MXVW PDGH E\ WKH ZKLWH SOD\HU 7KH HIIHFW RI WKH UHVSRQVH FRGH LV H[DPLQHG LQ &KDSWHU

PAGE 87

&+$37(5 ,$0 ,Q WKLV FKDSWHU ZH SUHVHQW D GHWDLOHG H[DPLQDWLRQ RI RXU LPSOHPHQWDWLRQ RI RXU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ IRU WKH GRPDLQ RI FKHVV 2XU SURJUDP LV QDPHG ,$0 DQ DFURQ\P IRU ,QGXFWLYH $GYHUVDU\ 0RGHOHU :H ILUVW SUHVHQW VHYHUDO GHWDLOHG H[DPSOHV WKDW GHPRQVWUDWH ,$0fV IXQFWLRQDOLW\ 1H[W ZH UHYLHZ WKH HIIHFWLYHQHVV RI ,$0 E\ H[DPLQLQJ WKH SUHGLFWLYH FDSDELOLWLHV RI RXU DGYHUVDU\ PRGHO ZKLOH SHUIRUPLQJ DJDLQVW DQ DFWXDO RSSRQHQW 2YHUYLHZ RI ,$0fV ([HFXWLRQ ,$0 LV GHILQHG LQ WZR VWDJHV 7KH ILUVW VWDJH SHUIRUPV WKH NQRZOHGJH DFTXLVLWLRQ RI FKXQNV DQG WKH OHDUQLQJ E\ LQGXFWLRQ SKDVHV RI WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ ZKLFK HVWDEOLVK WKH NQRZOHGJH EDVH RI JHRPHWULF DQG WH[WXDO FKXQNV FRQWDLQHG LQ WKH DGYHUVDU\ PRGHO 7KH VHFRQG VWDJH DSSOLHV WKH NQRZOHGJH LQ WKH NQRZOHGJH EDVH WR D JDPH LQ SURJUHVV WR SUHGLFW SUREDEOH DGYHUVDU\ PRYHV DQG WR LGHQWLI\ WKH DGYHUVDU\fV SOD\LQJ VW\OH ,$0 KDV VHYHUDO JOREDO GDWD VWUXFWXUHV ZKLFK IDFLOLWDWH WKH DOJRULWKPLF GHVLJQ RI WKH SURJUDP &KXQNV ERWK WH[WXDO DQG JHRPHWULF DQG WKH VXJJHVWHG PRYHV RI WKH DGYHUVDU\ PRGHO DUH VWRUHG LQ G\QDPLF GDWD VWUXFWXUHV 7KH IRUP RI WKHVH VWUXFWXUHV DQG WKH SRLQWHUV WR WKHP DUH DPRQJ WKH JOREDO GDWD VWUXFWXUHV $GGLWLRQDOO\

PAGE 88

DSSOLFDWLRQ FRQVWDQWV VXFK DV WKH VL]H RI JHRPHWULF FKXQNV DQG SDZQ VWUXFWXUH FKXQNV WKH QXPEHU RI RSHQLQJ PRYHV WR DFTXLUH WKH LQWHUQDO UHSUHVHQWDWLRQ RI WKH FKHVV ERDUG DQG YDULRXV FRQVWDQWV XVHG E\ WKH KHXULVWLFV RI WKH WZR LQIHUHQFH HQJLQHV DUH VWRUHG DV JOREDO YDULDEOHV 7KH XVH RI WKHVH JOREDO VWUXFWXUHV DOORZV XV WR HDVLO\ FKDQJH WKH DPRXQW VL]H RU W\SH RI NQRZOHGJH WKDW LV WR EH DFTXLUHG DQG SURYLGHV D PHDQV IRU ILQH WXQLQJ WKH KHXULVWLFV XVHG LQ SUHGLFWLQJ DQ DGYHUVDU\fV PRYH FKRLFHV .QRZOHGJH $FTXLVLWLRQ DQG ,QGXFWLYH /HDUQLQJ 6WDJH ,QSXW WR ,$0 IRU WKH NQRZOHGJH DFTXLVLWLRQ DQG OHDUQLQJ VWDJH FRQVLVWV RI WZR RU WKUHH JDPH UHFRUGV RI WKH DGYHUVDU\f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

PAGE 89

%RWYLQQLN YV .HUHV 0RVFRZ 0D\ f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

PAGE 90

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f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fV SOD\LQJ VW\OH SDZQ VWUXFWXUH FKXQNV DUH QRW FRDOHVFHG $IWHU WKH LQGLYLGXDO JHRPHWULF FKXQNV KDYH EHHQ FROOHFWHG IRU HDFK JDPH WKH LQGXFWLRQ DOJRULWKP SURFHVVHV WKH FKXQNV WR ILQG SDWWHUQV ZKLFK DUH GXSOLFDWHG DFURVV JDPH ERXQGDULHV 7KH LQGXFWLRQ DOJRULWKP IRXQG WZHQW\HLJKW XQLTXH FKXQNV VKRZQ LQ )LJXUH IURP WKH SRWHQWLDO FKXQNV LGHQWLILHG IRU %RWYLQQLN LQ WKH +DJXH0RVFRZ WRXUQDPHQW (DFK RI WKH WZHQW\HLJKW FKXQNV LQGXFHG ZDV IRXQG LQ

PAGE 91

DW OHDVW WZR JDPHV )RU H[DPSOH WKH SDUWLDO ERDUG UHSUHVHQWDWLRQV IRU WKH WKUHH JDPHV ZKLFK SURGXFHG RQH RI WKH WZHQW\HLJKW FKXQNV DUH VKRZQ LQ )LJXUH ’ OLOL LLOL m LOO LOO PP ‘ c, LOO ZZU PP LLOOL ?Pc c, L OLOL OL6LL PP L PP P ‘ OLOL 0LO SL ‘‘ L £N LLMLLL = LLMLLL L LLLLLL LOMLOLM LOO LLLLLL LLLLLLLLL LM+ ,,, MMMMMMMLLLLLMLL ,LMLLL N MMMMLLMMMML I" LMOMLMLLL_L Q % V MMMMMMMLMLMML ), %FUWYLPLLN YV 5HVKHYVNLM .HUHV YV %RWYLUPLN %RWYLPLLN YV (XXH )LJXUH &KXQN RFFXUUHQFH LQ WKUHH GLIIHUHQW JDPHV

PAGE 92

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

PAGE 93

)RU WR QXPEHU RI FKXQNV IRXQG IRU WKH FXUUHQW JDPHV GR %HJLQ )ODJBO )$/6( LQGLFDWHV ZKHQ D FKXQN KDV EHHQ UHSHDWHG )ODJB m )$/6( LQGLFDWHV ZKHQ D FKXQN LV GXSOLFDWHG LQ WKH .% )RU WR QXPEHU RI FKXQNV IRXQG IRU WKH FXUUHQW JDPHV GR )LUVW FKHFN WKLV FKXQN DJDLQVW WKH FKXQNV WKDW ZHUH MXVW IRXQG ,I &M 4 7KHQ %HJLQ ,QFUHPHQW QXPEHU RI WLPHV 4 KDV EHHQ REVHUYHG ,I &VWDUWBWLPH 4VWDUWWLPH 7KHQ &VWDUWBWLPH m &MVWDUWBWLPH ,I 4HQGWLPH &MHQGBWLPH 7KHQ 4HQG WLPH 4HQG WLPH 5(029(4f )ODJ O 758( (QG %HJLQ )RU WR QXPEHU RI FKXQNV FXUUHQWO\ LQ WKH FKXQN .QRZOHGJH %DVH GR 1H[W FKHFN WKLV FKXQN DJDLQVW DOO FKXQNV DOUHDG\ LQ WKH .QRZOHGJH %DVH ,I 4 4 7KHQ %HJLQ 8SGDWH .% NQRZOHGJH ,QFUHPHQW QXPEHU RI WLPHV 4 KDV EHHQ REVHUYHG ,I 4VWDUW WLPH 4VWDUWWLPH 7KHQ 4VWDUW WLPH m 4VWDUWBWLPH ,I 4HQG WLPH 4HQG WLPH 7KHQ 4HQGBWLPH m 4HQGBWLPH )ODJB 758( (QG %HJLQ ,I )ODJ O 758( $1' )ODJB )$/6( 7KHQ %HJLQ 1(:JHRPHWULFBIUDPH )f )4 :5,7() *HRPHWULF &KXQN .QRZOHGJH %DVHf (QG %HJLQ (QG %HJLQ )LJXUH 3VHXGRFRGH WR SHUIRUP LQGXFWLRQ RQ JHRPHWULF FKXQNV

PAGE 94

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

PAGE 95

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fV SUHIHUUHG VW\OH 'XULQJ WKH RSHQLQJ SRUWLRQ RI WKH VLPXODWHG JDPH ,$0 UHOLHV H[FOXVLYHO\ RQ WKH WH[WXDO PRYH NQRZOHGJH EDVH ZKLFK FRQWDLQV WKH VWDWLVWLFDO DQDO\VLV RI RSHQLQJ SDWWHUQV GLVSOD\HG E\ WKH DGYHUVDU\ 3ULRU WR WKH DGYHUVDU\fV PRYH ,$0 SUHGLFWV WKH DGYHUVDU\fV PRYH IURP WKH PRYHV FXUUHQWO\ LQ WKH NQRZOHGJH EDVH )ROORZLQJ WKH DFWXDO PRYH PDGH E\ WKH DGYHUVDU\ WKH FRUUHVSRQGLQJ PRYH LQ WKH NQRZOHGJH EDVH LV HOLPLQDWHG IURP IXUWKHU FRQVLGHUDWLRQ )RU DQ DGYHUVDU\ SOD\LQJ WKH ZKLWH SLHFHV ,$0 KDV KDG DFFXUDWH SUHGLFWLRQV UDQJLQJ IURP IRUW\ SHUFHQW WR RQH KXQGUHG SHUFHQW ZLWK D PHDQ SUHGLFWLRQ DFFXUDF\ RI VHYHQW\ILYH SHUFHQW ,$0fV RSHQLQJ VHTXHQFH SUHGLFWLRQV DUH OHVV DFFXUDWH IRU DQ DGYHUVDU\ SOD\LQJ WKH EODFN SLHFHV 7KLV LV EHFDXVH WKH DGYHUVDU\ LV SOD\LQJ UHVSRQVLYHO\ WR KLV RSSRQHQWfV PRYHV $FFXUDF\ IRU RSHQLQJ PRYH SUHGLFWLRQV DJDLQVW D EODFN SLHFH SOD\LQJ DGYHUVDU\ UDQJHG IURP ]HUR WR WZHQW\ SHUFHQW

PAGE 96

%\ PRGLI\LQJ WKH WH[WXDO FKXQN OHDUQLQJ DOJRULWKP WR UHPHPEHU WKH ZKLWH SOD\HUfV PRYH ZKLFK SUHFHGHG DQ DGYHUVDU\fV EODFN PRYH ZH ZHUH DEOH WR VLJQLILFDQWO\ LQFUHDVH ,$0fV RSHQLQJ VHTXHQFH SUHGLFWLRQ DFFXUDF\ WR UDQJH EHWZHHQ IRUW\ DQG RQH KXQGUHG SHUFHQW ZLWK D PHDQ RI VL[W\ SHUFHQW 7KLV PRGLILFDWLRQ SURGXFHV UHVXOWV ZKLFK DUH VLPLODU WR ,$0f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fV RSSRQHQWf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

PAGE 97

2QFH DQ DGYHUVDU\f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fV WKLUWHHQWK PRYH IURP WKH JDPH GLVSOD\HG LQ )LJXUH 7KH FRUUHVSRQGLQJ PRYHV WKDW ZRXOG SURGXFH WKH FKXQNV DUH DOVR GLVSOD\HG LQ WKH ILJXUH 7KH FROOHFWLRQ RI PRYHV VXJJHVWHG E\ WKH DQDORJRXV FKXQNV LV WKHQ SURFHVVHG E\ DQ LQIHUHQFH HQJLQH WR KHXULVWLFDOO\ GHWHUPLQH WKH PRVW SUREDEOH PRYH 7KH LQIHUHQFH HQJLQH SUHGLFWHG WKH FDVWOLQJ PRYH IRU WKH )LJXUH ERDUG SRVLWLRQ 7KLV SUHGLFWLRQ UHSUHVHQWV WKH UHVXOW RI DSSO\LQJ WKH YDULRXV SV\FKRORJLFDOO\ EDVHG KHXULVWLFV FRQFHUQLQJ FKXQN VL]H DQG UHOHYDQFH WR WKH FXUUHQW GRPDLQ 7KH TXHHQ EHLQJ HQ SULVH LV GRPDLQ VSHFLILF LQIRUPDWLRQ DQG LV WKHUHIRUH QRW FRQVLGHUHG E\ WKH PRYH SUHGLFWLRQ LQIHUHQFH HQJLQH +RZHYHU WKH DFWXDO PRYH PDGH E\ %RWYLQQLN 4F RU EF LV RQH RI WKH PRYHV VXJJHVWHG E\ DQ DQDORJRXV FKXQN DQG UHYLHZHG E\ WKH LQIHUHQFH HQJLQH $Q H[SHULPHQW WR DXJPHQW WKH DGYHUVDU\ SDWWHUQ NQRZOHGJH EDVH ZLWK D PLQLPDO DPRXQW RI GRPDLQ VSHFLILF NQRZOHGJH LV SUHVHQWHG LQ VHFWLRQ

PAGE 98

7KH SURFHVV RI VHDUFKLQJ WKH FXUUHQW ERDUG FRQILJXUDWLRQ WR LGHQWLI\ DQDORJRXV SRVLWLRQV LV UHSHDWHG EHIRUH HDFK PRYH WR EH PDGH E\ WKH DGYHUVDU\ ([LVWLQJ FKHVV SURJUDPV FDQ XVH ,$0fV SUHGLFWLYH FDSDELOLWLHV E\ KDYLQJ ,$0 VHDUFK WKH SRVVLEOH ERDUG FRQILJXUDWLRQ DW D VSHFLILF QRGH RI WKH JDPH WUHH DQG ILQGLQJ DQDORJRXV FKXQN VLWXDWLRQV LQ WKH SURSRVHG ERDUG FRQILJXUDWLRQ ,$0 ZLOO WKHQ SUHGLFW WKH PRVW SUREDEOH DGYHUVDU\ PRYH WKDW FRUUHVSRQGV WR RXU GRPDLQ VLPSOLILFDWLRQ K\SRWKHVLV RI &KDSWHU ,QWHJUDWLQJ ,$0fV .QRZOHGJH LQWR D &KHVV 3URJUDP 2XU LQGXFWLRQ EDVHG DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ SURGXFHV NQRZOHGJH ZKLFK FDQ EH LQFRUSRUDWHG LQWR FXUUHQW FKHVV SURJUDPV E\ WZR PHWKRGV .QRZOHGJH

PAGE 99

DERXW WKH VSHFLILF RSHQLQJ VHTXHQFHV SUHIHUUHG E\ DQ DGYHUVDU\ FDQ EH XVHG WR PDQXDOO\ RSWLPL]H D FKHVV SURJUDPf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

PAGE 100

,$0fV 3HUIRUPDQFH 5HVXOWV ,Q WKLV VHFWLRQ ZH GHVFULEH VHYHUDO H[SHULPHQWV WKDW KDYH EHHQ SHUIRUPHG ZLWK ,$0 WR WHVW WKH RYHUDOO FDSDELOLWLHV RI WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ (DFK H[SHULPHQW LV GHVFULEHG SUHFHGLQJ WKH DQDO\VLV RI WKH UHVXOWV 7KH JDPHV RI WKH DGYHUVDU\ %RWYLQQLN DUH DOO WDNHQ IURP KLV :RUOG &KDPSLRQVKLS PDWFKHV IURP WKH +DJXH0RVFRZ WRXUQDPHQW WKURXJK KLV PDWFK DJDLQVW 7DO 7KH WZR JDPHV ZKLFK DUH XVHG WR VLPXODWH %RWYLQQLNfV SHUIRUPDQFH LQ D IXWXUH PDWFK FRPH IURP %RWYLQQLNf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fV NQRZOHGJH EDVH IRU HDFK RI WKH WRXUQDPHQWV RI %RWYLQQLNfV FDUHHU 7KH

PAGE 101

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f WKH QXPEHU RI SUHGLFWLRQV ZKLFK H[DFWO\ PDWFKHG WKH DGYHUVDU\fV HQVXLQJ PRYH &f DQG WKH

PAGE 102

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f 7KH ILQDO WKUHH FROXPQV UHYHDO WKH VWDWLVWLFDO PHDVXUHV IRU WKH QXPEHU RI DQDORJRXV SRVLWLRQV IRXQG b*f WKH DFFXUDF\ RI WKH SUHGLFWLRQV WKDW ZHUH PDGH b&f DQG WKH SHUFHQWDJH RI SUHGLFWLRQV LGHQWLI\LQJ WKH FRUUHFW SLHFH WR EH PRYHG b&3f 7KH SHUFHQWDJH RI FRUUHFW SUHGLFWLRQV IRU WKH HQWLUH JDPH FDQ EH REWDLQHG E\ PXOWLSO\LQJ WKH DSSURSULDWH FROXPQ b& RU b&3f E\ WKH SHUFHQWDJH RI DQDORJRXV SRVLWLRQV :H ZLOO XVH WKH SHUFHQWDJH RI FRUUHFW SUHGLFWLRQV DQG FRUUHFW SLHFH LGHQWLILFDWLRQV IRU WKH HQWLUH JDPH LQ RXU DQDO\VLV RI ,$0fV SHUIRUPDQFH

PAGE 103

7DEOH $0 3HUIRUPDQFH 0HDVXUHV IRU 7(67 &K 0 3 & &3:/ b* b& b& 3 7KH JDPHV DQDO\]HG E\ ,$0 LQ 7DEOH DUH IRU WKH DGYHUVDU\ %RWYLQQLN DQG DUH IURP WKH IROORZLQJ WRXUQDPHQWV f JDPHV DUH IURP WKH +DJXH0RVFRZ 7RXUQDPHQW f QHZ JDPHV DUH IURP WKH %URQVWHLQ 0DWFK f QHZ JDPHV DUH IURP WKH 6P\VORY 0DWFK f QHZ JDPHV DUH IURP WKH 6P\VORY 0DWFK f QHZ JDPHV DUH IURP WKH 6P\VORY 0DWFK DQG WKH DQG 7DO 0DWFKHV 7KH VL[WK DQG QLQWK FROXPQV RI 7DEOH DUH VLJQLILFDQW LQ GHPRQVWUDWLQJ WKDW WKH DGYHUVDU\ PRGHO LV IRFXVLQJ RQ WKH FRUUHFW DUHD RI WKH JDPH ERDUG RQO\ XVLQJ NQRZOHGJH RI WKH SDWWHUQ FKXQNV NQRZQ E\ DQ DGYHUVDU\ ,$0fV DQDO\VLV RI WKH ILUVW WZHOYH JDPHV SOD\HG E\ %RWYLQQLN SURGXFHV FRUUHFW SUHGLFWLRQV RI DGYHUVDU\ PRYHV IRU QLQH SHUFHQW RI WKH WRWDO PRYHV DQG FRUUHFWO\

PAGE 104

LGHQWLILHV WKH SLHFH WR EH PRYHG LQ WZHQW\WKUHH SHUFHQW RI WKH WRWDO PRYHV LQ WKH JDPH 7KH VL]H RI WKH JHRPHWULF FKXQN NQRZOHGJH EDVH QHDUO\ GRXEOHV ZLWK ,$0fV DQDO\VLV RI WKH QH[W WHQ JDPHV SOD\HG E\ %RWYLQQLN :H FDQ VHH IURP WKH VHFRQG URZ RI WKH WDEOH WKDW WKH FRUUHFW SLHFH LV VWLOO LGHQWLILHG IRU WZHQW\WKUHH SHUFHQW RI WKH WRWDO PRYHV +RZHYHU WKH SHUFHQWDJH RI FRUUHFW SUHGLFWLRQV LQFUHDVHV WR DOPRVW HOHYHQ SHUFHQW RI WKH WRWDO JDPH PRYHV 7KH QHZ FKXQNV HQDEOH WKH DGYHUVDU\ PRGHO WR FRQYHUW RQH RI WKH FRUUHFW SLHFH LGHQWLILFDWLRQV LQWR D FRUUHFW SUHGLFWLRQ ,$0f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fV SUHGLFWLRQV LV GLVFDUGHG E\ WKH LQIHUHQFH HQJLQH PDNLQJ WKH SUHGLFWLRQV

PAGE 105

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fV PRYH VXJJHVWLRQ LV UHMHFWHG E\ WKH LQIHUHQFH HQJLQH GXH WR WKH H[LVWLQJ FKXQN

PAGE 106

7KH SVHXGRFRGH LQ )LJXUH UHTXLUHV WKDW WKH ERDUG SRVLWLRQ L Mf FRUUHVSRQG WR WKH XSSHUOHIWPRVW VTXDUH RI WKH GHILQHG FKXQN )LUVW ZH GHWHUPLQH WKH SLHFH WKDW LV UHTXLUHG WR FRPSOHWH WKH FKXQN DQG WKHQ ZH VHDUFK WKH ERDUG UHSUHVHQWDWLRQ WR ORFDWH WKH UHTXLUHG SLHFH ,I WKH SLHFH FDQ EH PRYHG IURP LWV FXUUHQW VTXDUH WR WKH VTXDUH ZKLFK FRPSOHWHV WKH FKXQN ZLWKRXW YLRODWLQJ WKH PRYHPHQW UXOHV RI FKHVV WKHQ WKH FRUUHVSRQGLQJ PRYH LV VXJJHVWHG $Q LQGLYLGXDO ,I QXPEHU RI SLHFHV QHHGHG WR FRPSOHWH WKH FKXQN & 7KHQ VHH LI HLWKHU D NLQJVLGH RU TXHHQVLGH FDVWOH ZLOO FRPSOHWH & (OVH %HJLQ )RU $ WR WKH ZLGWK RI & GR WKH FKXQN WR EH FRPSOHWHG )RU % WR WKH OHQJWK RI & GR ,I &$% $1' ;L$M% &$ % 7KHQ %HJLQ 3 F$% 3 LV WKH SLHFH QHHGHG WR FRPSOHWH FKXQN & D $ E % (QG %HJLQ )RU WR GR )RU WR GR ,I ;X 3 7KHQ ,I SLHFH 3 FDQ EH PRYHG IURP ERDUG ORFDWLRQ ,-f WR ORFDWLRQ L D M Ef 7KHQ VXJJHVW WKLV PRYH DV D SRVVLEOH QH[W PRYH (QG %HJLQ LJXUH 3VHXGRFRGH WR VXJJHVW SRVVLEOH DGYHUVDU\ PRYHV

PAGE 107

FKXQN FDQ SURGXFH PXOWLSOH VXJJHVWLRQV LI PRUH WKDQ RQH RI WKH UHTXLUHG SLHFH WR FRPSOHWH WKH FKXQN H[LVWV RQ WKH ERDUG DQG FDQ EH PRYHG WR WKH SURSHU ORFDWLRQ ,$0fV DFFXUDF\ ZLWK UHVSHFW WR WKH HQWLUH JDPH LV IDLUO\ FRQVLVWHQW ZLWK FRUUHFW SUHGLFWLRQV UDQJLQJ EHWZHHQ QLQH DQG HOHYHQ SHUFHQW RI WKH WRWDO PRYHV RI WKH JDPH $ VOLJKWO\ ZLGHU UDQJH IRXUWHHQ WR WZHQW\WKUHH SHUFHQW RI WKH WRWDO PRYHV LV REWDLQHG IRU SUHGLFWLRQV LGHQWLI\LQJ WKH FRUUHFW SLHFH WR EH PRYHG :KLOH EHLQJ DEOH WR FRUUHFWO\ SUHGLFW DSSUR[LPDWHO\ WHQ SHUFHQW RI DQ DGYHUVDU\fV PRYHV LV DQ H[WUDRUGLQDU\ DFFRPSOLVKPHQW WKH XVHIXOQHVV RI RXU PHWKRGRORJ\ IRU FRDFKLQJ H[LVWLQJ FKHVV SURJUDPV LV GHSHQGHQW RQ WKH UHOLDELOLW\ RI WKH SUHGLFWLRQV RU WKH UDWLR RI WKH FRUUHFW SUHGLFWLRQV WR WKH WRWDO QXPEHU RI SUHGLFWLRQV PDGH IRU D JDPH 5HGXFLQJ LQFRUUHFW SUHGLFWLRQV WKURXJK OLNHOLKRRG 7KH SHUFHQWDJH RI FRUUHFW SUHGLFWLRQV WR WRWDO SUHGLFWLRQV PDGH E\ ,$0 VKRZQ LQ 7DEOH IDOOV EHWZHHQ WZHOYH DQG VHYHQWHHQ SHUFHQW 7KLV PHDQV WKDW D FKHVV SURJUDPfV HYDOXDWLRQ IXQFWLRQ FDQ UHO\ RQ ,$0f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

PAGE 108

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fV NQRZOHGJH ZLWKRXW IHDU RI EOXQGHULQJ LQWR WKH WUDS RI XWLOL]LQJ DQ LQFRUUHFW SUHGLFWLRQ 7KH RQO\ SUREOHP ZKLFK DULVHV IURP UHTXLULQJ SUHGLFWLRQV WR KDYH D KLJK OLNHOLKRRG LV WKDW FRUUHFW SUHGLFWLRQV ZLWK VPDOOHU OLNHOLKRRGV DUH LJQRUHG )URP 7DEOH ZH VHH WKDW WKH SHUFHQWDJH RI JDPH PRYHV ZKLFK DUH FRUUHFWO\ SUHGLFWHG GURSV WR MXVW RYHU ILYH SHUFHQW 7DEOH (IIHFW RI /LNHOLKRRG RQ 3UHGLFWLRQ 5DWLRV *DPHV b b b b b b b b b b b b b b b b b b b b 3HUIRUPDQFH YHUVXV WKH DGYHUVDU\ DV WKH EODFN SOD\HU 7KH VHFRQG SDUW RI RXU LQLWLDO H[SHULPHQW WR GHWHUPLQH WKH EDVH SHUIRUPDQFH OHYHOV RI ,$0 XVHV 7(67 WR HVWDEOLVK ,$0fV FDSDELOLWLHV DJDLQVW DQ DGYHUVDU\ SOD\LQJ WKH EODFN SLHFHV 7(67 LV WKH JDPH IURP %RWYLQQLNfV FKDPSLRQVKLS PDWFK

PAGE 109

DJDLQVW 3HWURVLDQ ZKLFK IROORZHG WKH 7(67 JDPH 7KH OHQJWK RI WKH 7(67 JDPH LV WXUQV 7KH SHUIRUPDQFH UHVXOWV IRU ,$0 GXULQJ WKH VLPXODWHG 7(67 JDPH DUH GLVSOD\HG LQ 7DEOH 7KH PHDQLQJV RI HDFK RI WKH KHDGLQJV LQ 7DEOH DUH WKH VDPH DV IRU 7DEOH %HFDXVH WKH NQRZOHGJH DFTXLVLWLRQ DQG LQGXFWLYH OHDUQLQJ VHJPHQWV KDYH DOUHDG\ EHHQ FRPSOHWHG WKH ILUVW WKUHH FROXPQV RI 7DEOH DUH LGHQWLFDO WR 7DEOH H[FHSW WKH WH[WXDO RSHQLQJ PRYH NQRZOHGJH EDVH QXPEHUV UHSUHVHQW WKH WRWDO FKXQNV DQG WKH FKXQNV IRU WKH EODFN RSHQLQJ PRYHV RI WKH DGYHUVDU\ 7DEOH ,$0 3HUIRUPDQFH 0HDVXUHV IRU 7(67 &K 0 3 & &3:/ b* b& b&3 :H LPPHGLDWHO\ VHH WKDW 7DEOH GHPRQVWUDWHV OHDUQLQJ RU DQ LQFUHDVH LQ SHUIRUPDQFH DV PRUH JDPHV DUH DQDO\]HG E\ ,$0 7KH FRUUHFW SUHGLFWLRQV ULVH IURP ]HUR WR D FRQVLVWHQW WZR SHUFHQW SHUFHQW PXOWLSOLHG E\ SHUFHQWf RI WKH WRWDO PRYHV RI WKH JDPH ,GHQWLILFDWLRQ RI WKH SLHFH WR EH PRYHG LQFUHDVHV JUDGXDOO\ IURP VHYHQ SHUFHQW WR RYHU WHQ SHUFHQW RI WKH JDPH PRYHV :H SUHYLRXVO\ PHQWLRQHG D PRGLILFDWLRQ WR WKH WH[WXDO FKXQN NQRZOHGJH EDVH WR FDSWXUH WKH UHVSRQVLYH QDWXUH RI EODFNfV SOD\ 7KH UHVXOWV RI WKLV PRGLILFDWLRQ DUH SUHVHQWHG LQ 7DEOH 7KH GLIIHUHQFHV EHWZHHQ 7DEOH DQG 7DEOH RWKHU WKDQ WKH

PAGE 110

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b* b& b& 3 (YHQ ZLWK WKH PRGLILFDWLRQ WR DFFRXQW IRU WKH UHVSRQVLYH QDWXUH RI EODFN SOD\ ,$0fV SHUIRUPDQFH DJDLQVW DQ DGYHUVDU\ SOD\LQJ WKH EODFN SLHFHV LV RQO\ ILIW\ SHUFHQW RI WKH SHUIRUPDQFH DJDLQVW DQ DGYHUVDU\ SOD\LQJ ZKLWH 7KH UHDVRQ IRU WKLV LV WKDW WKH TXDOLW\ RI NQRZOHGJH DYDLODEOH DERXW WKH DGYHUVDU\ DV D EODFN SOD\HU LV LQIHULRU WR WKH NQRZOHGJH DYDLODEOH DERXW WKH DGYHUVDU\ DV D ZKLWH SOD\HU 7KH ILUVW VHYHUDO

PAGE 111

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f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

PAGE 112

%RWYLQQLNfV FKHVV PDWFK YHUVXV 7DO LV XVHG WR DQDO\]H JDPHV WKDW DUH UHFHQW WR WKH 7(67 JDPH ,$0f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fV SUHGLFWLRQV 7DEOH (IIHFW RI 5HFHQF\ RQ ,$0 3HUIRUPDQFH IRU 7(67 &K 0 3 & &3:/ b* b& b& 3 7KH VHFRQG SDUW RI WKH WHPSRUDO H[SHULPHQW WULHV WR VLPXODWH WKH HIIHFW RI IRUJHWWLQJ RQ DQ DGYHUVDU\fV SDWWHUQ NQRZOHGJH E\ DJLQJ WKH JHRPHWULF SDWWHUQ FKXQNV 2QFH D NQRZOHGJH EDVH H[LVWV LQ WKH DGYHUVDU\ PRGHO IRU D SDUWLFXODU DGYHUVDU\ WKH JHRPHWULF SDWWHUQ FKXQNV KDYH WKHLU QXPEHU RI REVHUYDQFHV KDOYHG SULRU WR LQFUHPHQWDOO\ DGGLQJ QHZ FKXQNV RU UHLQIRUFHPHQW RI H[LVWLQJ FKXQNV IURP DQDO\]LQJ DGGLWLRQDO JDPHV SOD\HG E\ WKH DGYHUVDU\ ,I D FKXQNfV QXPEHU RI REVHUYDWLRQV LV OHVV WKDQ RQH WKHQ WKH FKXQN LV GHOHWHG IURP WKH NQRZOHGJH EDVH 7KLV KDV WKH HIIHFW RI UHPRYLQJ FKXQNV OHDUQHG IURP HDUOLHUROGHU JDPHV LI WKH\ DUH QRW UHSHDWHG LQ PRUH UHFHQW JDPHV

PAGE 113

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fV 3HUIRUPDQFH &K 5(0 3 & &3:/ b* b& b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

PAGE 114

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fV WKUHH HDUOLHVW WRXUQDPHQWV :H IHHO WKDW WKH UHVXOWV DUH VXIILFLHQW WR GHPRQVWUDWH WKH HIIHFW RI FKDQJLQJ WKH RSHQLQJ PRYH NQRZOHGJH EDVH VL]H ,$0fV SHUIRUPDQFH ZLWK WKH ODUJHU WH[WXDO NQRZOHGJH EDVHV LV EHWWHU IRU WKH FDVH RI DQDO\]LQJ WZHOYH JDPHV ZLWK D WZHQW\ SHUFHQW LQFUHDVH LQ FRUUHFW SUHGLFWLRQV DQG D ILIWHHQ WR WZHQW\WKUHH SHUFHQW LQFUHDVH LQ FRUUHFW SLHFH WR EH PRYHG LGHQWLILFDWLRQ IRU WKH WHQ DQG ILIWHHQ PRYH NQRZOHGJH EDVHV UHVSHFWLYHO\ +RZHYHU

PAGE 115

7DEOH 3HUIRUPDQFH 7(67f ZLWK 2SHQLQJ 0RYHV &K 0 3 & &3:/ b* b& b&3 7DEOH 3HUIRUPDQFH 7(67f ZLWK 2SHQLQJ 0RYHV &K 0 3 & &3:/ b* b& b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

PAGE 116

ILOOLQJ LQ WKH JDSV IRU WKH VL[WK DQG ODWHU PRYHV WKDW DUH SUHVHQW LQ WKH VPDOOHU WZHOYH JDPH DGYHUVDU\ PRGHO 7KH QXPEHU RI WH[WXDO SDWWHUQV VWRUHG LQ WKH DGYHUVDU\ PRGHOfV NQRZOHGJH EDVH LQFUHDVHV VLJQLILFDQWO\ ZLWK WKH FKDQJH WR FDSWXUH DGGLWLRQDO RSHQLQJ PRYHV DV WH[WXDO SDWWHUQV )RU WKH WHQ RSHQLQJ PRYHV WH[WXDO SDWWHUQ PRGLILFDWLRQ WKH WH[WXDO SRUWLRQ RI WKH NQRZOHGJH EDVH LV WULSOHG DQG WKH ILIWHHQ PRYH PRGLILFDWLRQ H[FHHGV TXLQWXSOLQJ WKH QXPEHU RI WH[WXDO FKXQNV 6LQFH QR DSSDUHQW EHQHILW LV SURGXFHG E\ WKH H[WHQVLRQ RI WKH WH[WXDO SDWWHUQV ZH GR QRW UHFRPPHQG WKLV PRGLILFDWLRQ H[FHSW ZKHQ WKH QXPEHU RI JHRPHWULF FKXQNV LV VPDOO HJ OHVV WKDQ ILIW\ FKXQNVf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fV SHUIRUPDQFH PD\ EH ORVW :H UHVHDUFK WKH SHUIRUPDQFH RI WKH DGYHUVDU\ PRGHO ZLWK UHVSHFW WR JHRPHWULF FKXQN VL]H E\ DOWHULQJ WKH FKXQN DFTXLVLWLRQ DOJRULWKP IROORZLQJ WKH FRQYROXWLRQ RSHUDWRU WR DFFHSW FKXQNV RI D WZHQW\ILYH VTXDUH DUHD DQG DJDLQ WR DFFHSW FKXQNV RI D WKLUW\VL[ VTXDUH DUHD 7KH UHVXOWV IRU WKH ILUVW WKUHH DGYHUVDU\ PRGHOV DUH VKRZQ LQ 7DEOH IRU WKH ILYHE\ILYH DUHD FKXQNV DQG 7DEOH IRU WKH

PAGE 117

VL[E\VL[ DUHD FKXQNV 7KH fQFf LQ WKH WH[WXDO FKXQNV FROXPQ LQGLFDWHV fQR FKDQJHf IURP WKH 7DEOH YDOXHV 7DEOH ,$0 3HUIRUPDQFH 7(67f ZLWK )LYHE\)LYH &KXQNV &K 0 3 & &3:/ b* b& b& 3 QF QF QF 7DEOH ,$0 3HUIRUPDQFH 7(67f ZLWK 6L[E\6L[ &KXQNV &K 0 3 & &3:/ b* b& b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

PAGE 118

)LJXUH &KXQN DFTXLUHG ZLWK [ VL]H DQG FRUUHVSRQGLQJ VPDOOHU FKXQN WKH VHYHQ SLHFHV RI WKH FKXQN +RZHYHU WKH ILYHE\ILYH FKXQN UHTXLUHV WKDW DQ DGGLWLRQDO WZR SLHFHV WKH TXHHQ DQG H[WUD SDZQ EH LQ SODFH WR XVH WKLV FKXQN IRU SUHGLFWLQJ WKH LGHQWLFDO PRYHV RI WKH IRXUE\IRXU FKXQN ,$0f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fV LQIHUHQFH HQJLQH VLPXODWHV WKH

PAGE 119

KLHUDUFKLFDO FRQVWUXFWLRQ RI FRPSOH[ FKXQNV IURP VPDOOHU OHVV FRPSOH[ FKXQNV WKH RYHUDOO SHUIRUPDQFH RI ,$0 LV EHWWHU ZKHQ ZH XVH WKH VPDOOHVW FKXQN VL]H RI IRXUE\IRXU %DVHG RQ ,$0f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fV SHUIRUPDQFH IRU WKH WZHOYH JDPH DQG WZHQW\WZR JDPH DGYHUVDU\ PRGHOV DJDLQVW 7(67 LV VKRZQ LQ 7DEOH 7KH SHUIRUPDQFH IRU WKH DGYHUVDU\ PRGHO FUHDWHG IURP WKH DQDO\VLV RI WZHOYH RI %RWYLQQLNfV JDPHV LV LGHQWLFDO WR WKH RULJLQDO UHVXOWV JLYHQ LQ 7DEOH +RZHYHU WKH SHUIRUPDQFH IRU WKH WZHQW\WZR JDPH DGYHUVDU\ PRGHO VKRZV D UHGXFWLRQ LQ WKH QXPEHU RI SUHGLFWLRQV PDGH E\ WKH

PAGE 120

DGYHUVDU\ PRGHO IURP IRUW\VHYHQ WR WKLUW\IRXU 7KLV FRUUHVSRQGV WR DQ LQFUHDVH LQ WKH UHOLDELOLW\ RI WKH SUHGLFWLRQV DQG IXUWKHU LQFUHDVHV WKH XWLOLW\ RI WKH DGYHUVDU\ PRGHOfV SUHGLFWLRQV IRU H[LVWLQJ FKHVV SURJUDPV DV GLVFXVVHG LQ VHFWLRQ 7DEOH (IIHFW RI 'RPDLQ 6SHFLILF .QRZOHGJH &K 0 3 & &3:/ b* b& b& 3 7KH DGGLWLRQ RI WKH VPDOO DPRXQW RI NQRZOHGJH FRQFHUQLQJ HQ SULVH SLHFHV LQFUHDVHV WKH UHOLDELOLW\ RI ,$0fV FRUUHFW PRYH SUHGLFWLRQV E\ ILYH SHUFHQW DQG LV KLJKHU WKDQ DOO RI WKH UHOLDELOLW\ YDOXHV GLVSOD\HG LQ 7DEOH $Q H[SOLFLW H[DPSOH RI ,$0fV SHUIRUPDQFH LPSURYHPHQW LV VHHQ LQ WKH FKXQNVf SUHGLFWLRQV IRU WKH SRVLWLRQ LQ )LJXUH 6LQFH WKH TXHHQ LV HQ SULVH RQO\ WKH IRXUWK FKXQNfV VXJJHVWLRQ ZKLFK SUHGLFWV WKH FRUUHFW PRYH LV SHUPLWWHG E\ WKH QHZ KHXULVWLFV WKXV LQFUHDVLQJ WKH FRUUHFW SUHGLFWLRQV IRU WKDW JDPH E\ RQH ,$0f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

PAGE 121

+DJXH0RVFRZ 7RXUQDPHQW WKHUHIRUH RQO\ WKH JDPHV SUHFHGLQJ WKHVH WHVW JDPHV ZHUH XVHG WR FUHDWH WKH DGYHUVDU\ PRGHO NQRZOHGJH EDVH DQG WKH RWKHU WKUHH DUH DGGLWLRQDO JDPHV IURP WKH 3HWURVLDQ PDWFK ,$0f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fV SHUIRUPDQFH LV FRQVLVWHQW DFURVV WKH WKUHH JDPHV LQ ZKLFK WKH DGYHUVDU\ SOD\HG WKH ZKLWH SLHFHV DQG WKH IRXU JDPHV LQ ZKLFK WKH DGYHUVDU\ SOD\HG WKH EODFN SLHFHV 7KH DELOLW\ RI ,$0 WR UHSHDWHGO\ REWDLQ VLPLODU SHUIRUPDQFH OHYHOV LQFUHDVHV WKH VWDWLVWLFDO VLJQLILFDQFH RI ,$0fV SHUIRUPDQFH UHVXOWV

PAGE 122

&+$37(5 5(68/76 :H KDYH VXFFHVVIXOO\ DFTXLUHG JHRPHWULF DQG WH[WXDO FKXQNV WKDW KDYH EHHQ UHSHDWHG E\ DQ DGYHUVDU\ LQ PRUH WKDQ RQH JDPH 2XU DVVXPSWLRQ WKDW WKHVH FKXQNV IRUP WKH IRXQGDWLRQV RI DQ DGYHUVDU\fV HYDOXDWLRQ FULWHULD IRU WKH GRPDLQ LV YDOLGDWHG E\ WKH DFWXDO SUHGLFWLRQV PDGH E\ WKH DGYHUVDU\ PRGHO $SSUR[LPDWHO\ WHQ SHUFHQW RI DOO JDPH PRYHV ZKHQ WKH DGYHUVDU\ SOD\V WKH ZKLWH SLHFHV DUH FRUUHFWO\ SUHGLFWHG )XUWKHUPRUH IRU RYHU WZHQW\ SHUFHQW RI WKH PRYHV LQ D JDPH ZH FDQ LGHQWLI\ WKH SLHFH ZKLFK LV DERXW WR EH PRYHG 7KH LGHQWLILFDWLRQ RI ZKLFK SLHFH LV WR EH PRYHG LQGLFDWHV WKDW RXU DGYHUVDU\ PRGHO LV IRFXVVLQJ DWWHQWLRQ LQ WKH VDPH ORFDOH DV WKH DGYHUVDU\fV DWWHQWLRQ $GGLWLRQDO VXSSRUW IRU WKH VLPLODULWLHV EHWZHHQ WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\fV SHUFHSWLRQ RI WKH ERDUG DQG WKH DGYHUVDU\fV ERDUG HYDOXDWLRQ LV IRXQG E\ ORRNLQJ DW WKH WZR PRYHV IROORZLQJ WKH DGYHUVDU\ PRGHOfV SUHGLFWLRQV )RU WKH 7(67 JDPH ZLWKLQ WZR PRYHV RI D SUHGLFWLRQ PDGH E\ ,$0 WKH SUHGLFWHG PRYH ZDV H[HFXWHG RQFH DQG WKH SLHFH ZKLFK ,$0 SUHGLFWHG WR EH PRYHG ZDV XVHG IRXU WLPHV $OWKRXJK ,$0fV SUHGLFWLRQV ZHUH D OLWWOH SUHPDWXUH ZH FDQ VHH WKDW WKH DGYHUVDU\ PRGHO LV FRJQL]DQW RI WKH DUHD RI WKH ERDUG ZKHUH DFWLRQ LV DERXW WR WDNH SODFH 7KH SHUIRUPDQFH RI WKH DGYHUVDU\ PRGHO LV H[FHSWLRQDO ZKHQ ZH DFFRXQW IRU WKH TXDOLW\ RI WKH LQIRUPDWLRQ DQDO\]HG 7DEOH VKRZV WKH ZLQ WR ORVV UDWLR IRU WKH

PAGE 123

,OO %RWYLQQLN JDPHV DQDO\]HG E\ ,$0 +LJK TXDOLW\ LQIRUPDWLRQ LV REWDLQHG IURP ZRQ JDPHV DQG ORVW JDPHV SURYLGH XV ZLWK TXHVWLRQDEOHTXDOLW\ NQRZOHGJH $JDLQ ZH VHH WKDW ,$0f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t .RSHF f 7KH VL]H RI DQ RSHQLQJ ERRN DIIHFWV WKH VSHHG RI WKH SURJUDPfV SOD\ ZLWK ODUJHU ERRNV WDNLQJ PRUH WLPH WR SURFHVV 5HVHDUFK LV EHLQJ SHUIRUPHG WR UHGXFH WKH VL]H RI RSHQLQJ ERRNV IRU FKHVV SURJUDPV

PAGE 124

E\ WDNLQJ DGYDQWDJH RI UHGXQGDQFLHV DPRQJ YDULRXV RSHQLQJ OLQHV :KLWH f :KLWHfV PHWKRG DIIRUGV D WKLUW\HLJKW SHUFHQW VDYLQJV LQ VSDFH UHGXFLQJ D E\WH RSHQLQJ ERRN WR E\WHV 7KH FROOHFWLRQ RI WH[WXDO FKXQNV UHIHUULQJ WR RSHQLQJ PRYHV VLPXODWHV D KXPDQ PDVWHUf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f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

PAGE 125

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fV 0,1 OHYHO GHFLVLRQ QRGH 7KH 0$; SOD\HU WKH FRPSXWHU QRZ NQRZV WKDW D GUDZ FDQ EH REWDLQHG E\ VHOHFWLQJ WKH PRYH FRUUHVSRQGLQJ WR WKH OHIW EUDQFK RI WKLV JDPH WUHH /LNHZLVH WKH ULJKW EUDQFK UHWXUQV D YDOXH RI QHJDWLYH RQH RU D VOLJKW FKDQFH RI D ORVV DV WKH 0,1 SOD\HUfV PRYHPHQW FKRLFH 7KH ILUVW QRGH HYDOXDWHG IRU WKH FHQWHU EUDQFK KDV D

PAGE 126

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

PAGE 127

PRGHO RI WKLV DGYHUVDU\ WKH 0,1 SOD\HU VKRZV WKDW WKH DGYHUVDU\ KDV QHYHU DFFHSWHG D JDPELW LQ SUHYLRXV SOD\ 7KHUHIRUH WKH QHJDWLYH WKUHH YDOXHG QRGH VKRXOG QRW EH FRQVLGHUHG DV D YLDEOH PRYH FKRLFH IRU WKH DGYHUVDU\ WR FRQVLGHU 7KH QHZ JDPH WUHH PLQXV WKH JDPELW DFFHSWDQFH QRGH QRZ UHWXUQV D YDOXH RI SRVLWLYH WZR WR WKH % QRGH DV 0,1fV FKRLFH 7KH QHZ EHVW PRYH IRU WKH 0$; SOD\HU XVLQJ NQRZOHGJH RI WKH DGYHUVDU\f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

PAGE 128

&2 ‘ ZP mr, Lrrrr L VMM L f mr ‘ OL 5 L ? P ODS LYLYLYL @n 9 ‘  A cLOO b L6L ;  ,,, VLMr YOYL r M 6 e VL 0LLL6 LL  LIOL c5 OO ILJXUH 3UHGLFWLRQ RI %RWYLQQLNfV QH[W PRYH &KXQN XVHG ‘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

PAGE 129

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fV NQRZOHGJH IRU WKH VDPH QXPEHU RI H[SDQGHG QRGHV

PAGE 131

7KH WLPH RUGHU IRU WKH VWDQGDUG VHDUFK DOJRULWKP LV VWDQGDUG VHDUFKf NO r 1P 1Pf ZKHUH fPf LV WKH GHSWK RI WKH VHDUFK DQG fNOf LV D FRQVWDQW JUHDWHU WKDQ RQH FRUUHVSRQGLQJ WR WKH WLPH UHTXLUHG WR DFWXDOO\ SHUIRUP WKH VHDUFK DQG HYDOXDWLRQ DOJRULWKPV :LWK fSf UHSUHVHQWLQJ WKH SHUFHQWDJH RI FRUUHFW SUHGLFWLRQV PDGH E\ WKH DGYHUVDU\ PRGHO DQG WKH FRQVWDQW fNf FRUUHVSRQGLQJ WR WKH WLPH UHTXLUHG WR XVH WKH DGYHUVDU\ PRGHOfV NQRZOHGJH EDVH DQG SUHGLFWLRQ DOJRULWKP WKH QHZ WLPH RUGHU RI WKH VHDUFK DOJRULWKP LV SUHGLFWLRQ VHDUFKf N NO r 1PfnSff 1PfnSff 7KLV VKRZV WKDW WKH WLPH RUGHU FDQ EH UHGXFHG VLQFH fSf LV OHVV WKDQ RU HTXDO WR RQH E\ DQ H[SRQHQWLDO IDFWRU ZLWK WKH DGGLWLRQ RI D OLQHDU FRQVWDQW WR WKH WLPH HTXDWLRQ (VVHQWLDOO\ D VHDUFK DOJRULWKP FDQ QRZ DIIRUG WR VHDUFK P r Sf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f WR 1 rn ffff U 1f $ IRUW\RQH SO\ GHSWK VHDUFK FDQ EH DFFRPSOLVKHG LQ WKH VDPH WLPH DV WKH SUHYLRXV IRUW\ SO\ VHDUFK 7KH UHDVRQ IRU

PAGE 132

GLYLGLQJ WKH fSf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f1f RI ILYH RQO\ WZR RU WKUHH RI WKH VXJJHVWHG PRYHV ZLOO DFWXDOO\ XVH WKH SLHFH SUHGLFWHG E\ WKH

PAGE 133

DGYHUVDU\ PRGHO $Q H[DPSOH RI WKH LPSURYHG VHDUFK WUHH IURP XVLQJ WKH DGYHUVDU\ PRGHOfV FRUUHFW SLHFH LGHQWLILFDWLRQV LV VKRZQ LQ )LJXUH )LJXUH 6HDUFK WUHH REWDLQHG IURP &RUUHFW 3LHFH LGHQWLILFDWLRQ 7KH WLPH RUGHU HTXDWLRQ XVHV fTf WR UHSUHVHQW WKH SHUFHQWDJH RI DFWXDO PRYHV IRU ZKLFK WKH DGYHUVDU\ PRGHO FDQ SUHGLFW HLWKHU WKH FRUUHFW PRYH RU LGHQWLI\ WKH FRUUHFW SLHFH WR EH PRYH :H PXVW UHPHPEHU IURP &KDSWHU WKDW fTf LV WZLFH DV ODUJH DV fSf 7KH QHZ UHGXFHG EUDQFKLQJ IDFWRU IRU QRGHV ZKHUH SUHGLFWLRQV DUH PDGH LV f5f DQG ZH FDQ JXDUDQWHH WKDW f5f LV OHVV WKDQ f1f 7KH QHZ WLPH RUGHU LV 2LPSURYHG SUHGLFWLRQ VHDUFKf N r 5PfTf NO r 1P nTff 1PrTff RU 5P fTf

PAGE 134

7KH DFWXDO WLPH RUGHU GHSHQGV RQ WKH YDOXH RI fT? %HFDXVH RI WKH QHHG WR DFWXDOO\ SHUIRUP D UHGXFHG VHDUFK DW HDFK RI WKH WUHH QRGHV DVVRFLDWHG ZLWK DGYHUVDU\ PRGHO SUHGLFWLRQV WKH UHVXOWDQW DGGLWLYH IDFWRU LV DOVR DQ H[SRQHQWLDO LQVWHDG RI D OLQHDU FRQVWDQW ,I fTf LV OHVV WKDQ RQH KDOI RU ILIW\ SHUFHQW WKHQ WKH RUGHU LV WKH ILUVW DOWHUQDWLYH 2WKHUZLVH WKH VHFRQG DOWHUQDWLYH LV ODUJHU DQG ZLOO EHFRPH WKH WLPH RUGHU RI WKH VHDUFK DOJRULWKP :LWK WKH WZHQW\ SHUFHQW FRUUHFW SLHFH LGHQWLILFDWLRQ SHUIRUPDQFH GHPRQVWUDWHG E\ ,$0 LQ &KDSWHU WKH LPSURYHG DOJRULWKP UHGXFHV WKH WLPH RUGHU RI WKH VWDQGDUG VHDUFK DOJRULWKP IURP 1f WR 1f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fV PRYH E\ WKH DGYHUVDU\ PRGHO KDV UHVXOWHG LQ DQ H[SRQHQWLDO VDYLQJV LQ WLPH IRU VWDQGDUG JDPH VHDUFK DOJRULWKPV 7LPH VDYLQJV UHDOL]HG IURP WKH XVH RI WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ FRUUHVSRQG WR LQFUHDVHV LQ WKH GHSWK RI VHDUFK WKDW

PAGE 135

LV DFFRPSOLVKHG E\ D VHDUFK DOJRULWKP IRU WKH VDPH DPRXQW RI WLPH 'HHSHU VHDUFKHV LQ WKH JDPH WUHH SURYLGH EHWWHU WDFWLFDO VROXWLRQV WR FKHVV SUREOHPV 7KH WH[WXDO FKXQNV ZKLFK DUH XVHG WR UHGXFH WKH RSHQLQJ ERRN VL]H KDYH D FRUUHODWLYH VWUDWHJLF FRQWULEXWLRQ WR DGYHUVDU\ GRPDLQ SURJUDPV $ FKHVV SURJUDP FDQ XVH WKH NQRZOHGJH LQ WKH WH[WXDO SRUWLRQ RI WKH DGYHUVDU\ PRGHOfV NQRZOHGJH EDVH WR VHOHFW RSHQLQJ VHTXHQFHV RI SOD\ WKDW KDYH QRW EHHQ H[SHULHQFHG E\ WKH DGYHUVDU\ 7KLV SURGXFHV D VWUDWHJLF EHQHILW WR WKH FKHVV SURJUDP E\ SODFLQJ WKH DGYHUVDU\ LQ XQIDPLOLDU WHUULWRU\ ,PSURYHPHQWV LQ SOD\LQJ DELOLW\ RI FXUUHQW DGYHUVDULDO SURJUDPV DUH UHDOL]HG WKURXJK WKH UHYHODWLRQ RI SUHYLRXVO\ XQFRQVLGHUHG VHDUFK SDWKV 7KH DGYHUVDU\ PRGHO HQDEOHV D FKHVV SURJUDP WR GHWHUPLQH WKH PRVW OLNHO\ PRYH WR EH PDGH E\ DQ DGYHUVDU\ EDVHG RQ WKH DGYHUVDU\f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

PAGE 136

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fV PRYHV ZKHQ KLJK TXDOLW\ LQIRUPDWLRQ RU D KLJK SHUFHQWDJH RI ZRQ JDPHV KDV EHHQ DQDO\]HG $GGLWLRQDOO\ ZH FDQ SUHGLFW WKH SLHFH WR EH PRYHG IRU RYHU WZHQW\ SHUFHQW RI WKH DGYHUVDU\fV PRYH FKRLFHV LQ D JDPH 7KHVH UHVXOWV UHYHDO WKDW DGYHUVDULDO GRPDLQ H[SHUWV IUHTXHQWO\ XVH EDVLF FRJQLWLYH WHFKQLTXHV WKDW DUH GRPDLQ LQGHSHQGHQW WR FKRRVH WKHLU PRYHV 7KH FRJQLWLYH WHFKQLTXH WKDW ZH UHO\ RQ LV WKH VLPSOLILFDWLRQ RI FRPSOH[ GRPDLQ VLWXDWLRQV WR VLWXDWLRQV WKDW KDYH EHHQ HQFRXQWHUHG SUHYLRXVO\

PAGE 137

&XUUHQW FKHVV SURJUDP FDSDELOLWLHV UHVXOW IURP D FRPELQDWLRQ RI VHDUFK VSHHG DQG XVH RI GRPDLQ VSHFLILF NQRZOHGJH 2XU DGYHUVDU\ PRGHO LV QRW H[HPSW IURP WKH XVH RI GRPDLQ VSHFLILF VLQFH ZH KDYH VHHQ LQ &KDSWHU WKDW WKH DGGLWLRQ RI GRPDLQ VSHFLILF NQRZOHGJH LQFUHDVHV WKH UHOLDELOLW\ RI WKH DGYHUVDU\ PRGHOfV SUHGLFWLRQV :H KDYH GHPRQVWUDWHG WKDW E\ XVLQJ RXU GRPDLQ LQGHSHQGHQW NQRZOHGJH WKH FRPSOH[LW\ DQG WLPH RUGHU RI D VHDUFK DOJRULWKP FDQ EH UHGXFHG 7KH XVH RI RXU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\fV GRPDLQ LQGHSHQGHQW NQRZOHGJH LQFUHDVHV WKH SHUIRUPDQFH FDSDELOLWLHV RI FXUUHQW FKHVV SURJUDPV :H EHOLHYH WKDW WKH FRQWLQXHG LQFUHDVH LQ SHUIRUPDQFH RI LQWHOOLJHQW FKHVV SURJUDPV UHTXLUHV WKH XVH RI DGGLWLRQDO NQRZOHGJH LQFOXGLQJ GRPDLQ LQGHSHQGHQW NQRZOHGJH VLPLODU WR DQ DGYHUVDU\ PRGHO )XWXUH 5HVHDUFK 'LUHFWLRQV :H KDYH WULHG WR LQGLFDWH WKURXJKRXW WKLV GLVVHUWDWLRQ PHDQLQJIXO H[WHQVLRQV WR WKH DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\f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

PAGE 138

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fV SLHFHV ,QFUHDVH WKH .QRZOHGJH &RQWDLQHG LQ (DFK &KXQN 7KH NQLJKW LV FXUUHQWO\ LQFOXGHG LQ FKXQNV ZLWKRXW UHJDUG WR LWV PRYHPHQW FDSDELOLWLHV 7KH FRQYROXWLRQ DOJRULWKP FDQ EH PRGLILHG WR XVH D ILYHE\ILYH WHPSODWH ZKHQ D NQLJKW LV FRQWDLQHG LQ D FKXQN WR FDSWXUH WKH PRYHPHQW FDSDELOLWLHV RI WKH NQLJKW ,QFUHDVLQJ WKH WHPSODWH VL]H ZLWK VXLWDEOH PRGLILFDWLRQV IRU HDFK LQGLYLGXDO SLHFHf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

PAGE 139

.QLJKW 7HPSODWH %ODQNV KDYH ]HUR YDOXH ,. 3DUW RI WKH [ 5RRN 7HPSODWH )LJXUH .QLJKW WHPSODWH DQG SDUWLDO 5RRN WHPSODWH $GGLWLRQDO NQRZOHGJH FDQ EH REWDLQHG E\ GURSSLQJ RXU VLPLODULW\ FRQVWUDLQW RI FKXQNV FRQWDLQLQJ RQH FRORU RI SLHFHV 3HUPLWWLQJ FKXQNV WR FRQWDLQ ERWK FRORUV RI SLHFHV VLPXOWDQHRXVO\ KDV DOUHDG\ EHHQ XVHG WR DFTXLUH NQRZOHGJH DERXW WKH SDZQ VWUXFWXUHV DQG SDZQ FKDLQV IRXQG LQ DQ DGYHUVDU\f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

PAGE 140

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f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

PAGE 141

2WKHU W\SHV RI JOREDO NQRZOHGJH DUH DOVR DYDLODEOH WR LQFUHDVH WKH DGYHUVDU\ PRGHOfV FDSDELOLWLHV ,Q &KDSWHU ZH PHQWLRQHG WKDW FXOWXUDO DQG HGXFDWLRQDO GLIIHUHQFHV DIIHFW WKH VWUDWHJLF DQG WDFWLFDO SODQQLQJ RI DGYHUVDULHV .RWRY DQG
PAGE 142

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b :HDN PLGGOHJDPH RI b 0RUH HUURUV PDGH IURP GHIHQGLQJ WKDQ DWWDFNLQJ SRVLWLRQV RI b 6XIIHU IURP WLPH SUHVVXUH RI b 0RUH HUURUV GXH WR VWUDWHJ\ WKDQ WDFWLFV RI b 2XU DGYHUVDU\ PRGHOLQJ PHWKRGRORJ\ VXFFHVVIXOO\ SUHGLFWV PDQ\ RI WKH PRYHV WKDW DQ DGYHUVDU\ ZLOO PDNH 7KH SUHGLFWLRQV DUH EDVHG RQ GRPDLQ LQGHSHQGHQW NQRZOHGJH DQG GHPRQVWUDWH WKDW WKH XVH RI GRPDLQ LQGHSHQGHQW NQRZOHGJH DV ZHOO DV GRPDLQ GHSHQGHQW NQRZOHGJH DUH WKH FULWLFDO HOHPHQWV LQ FRQWLQXLQJ WR LPSURYH DGYHUVDULDO GRPDLQ DSSOLFDWLRQ SURJUDP SHUIRUPDQFH 7KH FXUUHQW PRGHO IRFXVHV

PAGE 143

SULPDULO\ RQ WKH FRJQLWLYH DQG SHUFHSWXDO SURFHVVHV XVHG E\ DQ DGYHUVDU\ WR HYDOXDWH WKH GRPDLQ 7KH WH[WXDO DQG YLVXDO FKXQNV ZKLFK IRUP WKH IRXQGDWLRQV RI WKH HYDOXDWLRQ PHFKDQLVPV HPSOR\HG E\ H[SHUWV DUH WKH GRPDLQ LQGHSHQGHQW NQRZOHGJH WKDW HQDEOHV WKH DGYHUVDU\ PRGHOfV SUHGLFWLRQV %\ IXUWKHU LQFUHDVLQJ WKH NQRZOHGJH FRQWDLQHG LQ WKH DGYHUVDU\ PRGHO ZLWK GRPDLQ GHSHQGHQW DQG JOREDO NQRZOHGJH WKH DGYHUVDU\ PRGHO ZLOO EH DEOH WR VXJJHVW VWUDWHJLFDOO\ DGYDQWDJHRXV PRYHV DV ZHOO DV SUHGLFWLQJ OLNHO\ DGYHUVDU\ PRYHV 7KH XVH RI WKH DGYHUVDU\ PRGHO DV D FRDFK ZLOO JUHDWO\ HQKDQFH WKH SHUIRUPDQFH FDSDELOLWLHV RI H[LVWLQJ DGYHUVDULDO GRPDLQ SODQQLQJ SURJUDPV

PAGE 144

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fDf WR fKf DQG HDFK URZ LV GHVLJQDWHG E\ D QXPEHU IURP WR DV VKRZQ EHORZ D E F G H I J K D E F G H I J K D E F G H I J K D E F G H e J K D E F G H I J K D E F G H I J K D E F G H e J K DO EO FO GO HO IO JO KL

PAGE 145

0RYHV DUH GHVFULEHG E\ JLYLQJ RQH RI WKH OHWWHUV 1 % 5 4 RU WR GHVLJQDWH WKH SLHFH WKDW LV EHLQJ PRYHG QR OHWWHU WUDQVODWHV WR D SDZQ PRYHf IROORZHG E\ WKH OHWWHU DQG QXPEHU RI WKH GHVWLQDWLRQ VTXDUH )RU H[DPSOH 1I GHVFULEHV WKH PRYH RI D NQLJKW WR WKH NLQJfVELVKRSWKUHH VTXDUH 7KH RQH URZ LV DOZD\V ZKLWHfV KRPH URZ DQG FRUUHVSRQGLQJO\ WKH HLJKW URZ LV DOZD\V EODFNfV KRPH URZ HQ SULVH $ SLHFH LV HQ SULVH LI LW FDQ EH FDSWXUHG E\ WKH RSSRQHQW RQ WKH QH[W PRYH HYDOXDWLRQ IXQFWLRQ (YDOXDWLRQ IXQFWLRQV DUH XVHG E\ FKHVV JDPHV WR GHWHUPLQH WKH YDOXH RU JRRGQHVV RI D VSHFLILF ERDUG SRVLWLRQ 7\SLFDO HYDOXDWLRQ IXQFWLRQV XVH GRPDLQ GHSHQGHQW WDFWLFDO DQG SRVLWLRQDO NQRZOHGJH VXFK DV PDWHULDO EDODQFH DQG FRQWURO RI WKH FHQWHU VTXDUHV RI WKH JDPH ERDUG LQ GHWHUPLQLQJ D SRVLWLRQf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fV PDWHULDO VXP LV VXEWUDFWHG IURP \RXU PDWHULDO VXP WR REWDLQ WKH PDWHULDO EDODQFH PLQLPD[ $Q DOJRULWKP WKDW GHFLGHV WKH RSWLPDO WUHH SDWK WR IROORZ E\ DOWHUQDWHO\ EDFNLQJ XS WKH PLQLPXP DQG PD[LPXP RI DOO OHDI QRGHV 7KH 0$; SOD\HU WULHV WR REWDLQ WKH PD[LPXP SRVVLEOH YDOXH RI DOO EUDQFKHV

PAGE 146

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f

PAGE 147

5()(5(1&(6 $GHOVRQ9HOVN\ 0 $UOD]DURY 9 / t 'RQVN\ 0 9 f 6RPH 0HWKRGV RI &RQWUROOLQJ WKH 7UHH 6HDUFK LQ &KHVV 3URJUDPV $UWLILFLDO ,QWHOOLJHQFH $GHOVRQ9HOVN\ 0 $UOD]DURY 9 / t 'RQVN\ 0 9 f $OJRULWKPV IRU *DPHV 6SULQJHU9HUODJ 1HZ
PAGE 148

%HUOLQHU + t (EHOLQJ & f 3DWWHUQ .QRZOHGJH DQG 6HDUFK 7KH 6835(0 $UFKLWHFWXUH 7HFKQLFDO 5HSRUW &08&6 &DUQHJLH0HOORQ 8QLYHUVLW\ 3LWWVEXUJK %HUOLQHU + t *RHWVFK f $ 4XDQWLWDWLYH 6WXG\ RI 6HDUFK 0HWKRGV DQG WKH (IIHFW RI &RQVWUDLQW 6DWLVIDFWLRQ 7HFKQLFDO 5HSRUW &08&6 &DUQHJLH0HOORQ 8QLYHUVLW\ 3LWWVEXUJK %OXP / t 0 %OXP 0 f 7RZDUG D 0DWKHPDWLFDO 7KHRU\ RI ,QGXFWLYH ,QIHUHQFH ,QIRUPDWLRQ DQG &RQWURO %ROH / (Gf f &RPSXWDWLRQDO 0RGHOV RI /HDUQLQJ 6SULQJHU9HUODJ 1HZ
PAGE 149

&KDVH : t 6LPRQ + $ f 7KH 0LQGfV (\H LQ &KHVV ,Q 5HDGLQJV LQ &RJQLWLYH 6FLHQFH 0RUJDQ .DXIPDQQ 6DQ 0DWHR &$ &KULVWHQVHQ t .RUI 5 ( f $ 8QLILHG 7KHRU\ RI +HXULVWLF (YDOXDWLRQ )XQFWLRQV DQG ,WV $SSOLFDWLRQ WR /HDUQLQJ ,Q 3URFHHGLQJV $$$, 7KH )LIWK 1DWLRQDO &RQIHUHQFH RQ $UWLILFLDO ,QWHOOLJHQFH 3KLODGHOSKLD &LPEDOD 6 f $UWLILFLDO ,QWHOOLJHQFH DQG 1DWLRQDO 6HFXULW\ /H[LQJWRQ %RRNV /H[LQJWRQ 0$ &URRNDOO *UHHQEODW & 6 &RRWH $ .ODEEHUV :DWVRQ 5 (GVf f 6LPXODWLRQ*DPLQJ LQ WKH /DWH fV 3HUJDPRQ 3UHVV 1HZ
PAGE 150

(OLWKRUQ $ t %DQHUML 5 (GVf f $UWLILFLDO DQG +XPDQ ,QWHOOLJHQFH (OVHYLHU 1HZ
PAGE 151

.RWRY $ t
PAGE 152

0LFKDOVNL 5 6 &DUERQHOO mIH 0LWFKHOO 7 0 (GVf f 0DFKLQH /HDUQLQJ $Q $UWLILFLDO ,QWHOOLJHQFH $SSURDFK 9ROXPH ,, 0RUJDQ .DXIPDQQ 6DQ 0DWHR &$ 0LOOHU $ f 7KH 0DJLFDO 1XPEHU 6HYHQ 3OXV RU 0LQXV 7ZR 6RPH /LPLWV RQ 2XU &DSDFLW\ IRU 3URFHVVLQJ ,QIRUPDWLRQ 7KH 3V\FKRORJLFDO 5HYLHZ 0LWFKHOO 7 0 &DUERQHOO IH 0LFKDOVNL 5 6 (GVf f 0DFKLQH /HDUQLQJ $ *XLGH WR &XUUHQW 5HVHDUFK .OXZHU $FDGHPLF 1RUZHOO 0$ 0XJJOHWRQ 6 f ,QGXFWLYH $FTXLVLWLRQ RI ([SHUW .QRZOHGJH $GGLVRQ:HVOH\ 5HDGLQJ 0$ 1DUHQGUD .f mIH 7KDWKDFKDU 0 f /HDUQLQJ $XWRPDWD $Q ,QWURGXFWLRQ 3UHQWLFH +DOO (QJOHZRRG &OLIIV 11HZERUQ 0 IH .RSHF f 5HVXOWV RI WKH 1LQHWHHQWK $&0 1RUWK $PHULFDQ &RPSXWHU &KHVV &KDPSLRQVKLS &RPPXQLFDWLRQV RI WKH $&0 f 1HZERUQ 0 mIH .RSHF f 7KH 7ZHQWLHWK $QQXDO $&0 1RUWK $PHULFDQ &RPSXWHU &KHVV &KDPSLRQVKLS &RPPXQLFDWLRQV RI WKH $&0 f 1XQQ IH *ULIILWKV 3 f 6HFUHWV RI *UDQGPDVWHU 3ODY 0DFPLOODQ 1HZ
PAGE 153

5LWWHU ; :LOVRQ 1 t 'DYLGVRQ / f ,PDJH $OJHEUD $Q 2YHUYLHZ 7HFKQLFDO 5HSRUW 75 8QLYHUVLW\ RI )ORULGD *DLQHVYLOOH )/ 5REELQV f $PHULFDfV 5HG $UP\ 7KH 1HZ
PAGE 154

6RNRORYVNL\ 9 f 6RYLHW 0LOLWDU\ 6WUDWHJ\ 0DFGRQDOG DQG -DQHfV 6WDQIRUG &$ 6ROWLV $ f 3DZQ 6WUXFWXUH &KHVV 'DYLG 0F.D\ &RPSDQ\ 1HZ
PAGE 155


PAGE 156

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

PAGE 157

, FHUWLI\ WKDW KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ IR-//IL8f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f" \ & :DOWHU 5 &XQQLQJKDUT\ 3URIHVVRU RI 3V\FKRORJ\

PAGE 158

, FHUWLI\ WKDW KDYH UHDG WKLV VWXG\ DQG WKDW LQ P\ RSLQLRQ LW FRQIRUPV WR DFFHSWDEOH VWDQGDUGV RI VFKRODUO\ SUHVHQWDWLRQ DQG LV IXOO\ DGHTXDWH LQ VFRSH DQG TXDOLW\ DV D GLVVHUWDWLRQ IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ ,U£ 6 )LVFKOHU 3URIHVVRU RI 3V\FKRORJ\ 7KLV GLVVHUWDWLRQ ZDV VXEPLWWHG WR WKH *UDGXDWH )DFXOW\ RI WKH &ROOHJH RI (QJLQHHULQJ DQG WR WKH *UDGXDWH 6FKRRO DQG ZDV DFFHSWHG DV SDUWLDO IXOILOOPHQW RI WKH UHTXLUHPHQWV IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ 'HFHPEHU cKrtXFF tf Af§ :LQIUHG 0 3KLOOLSV 'HDQ &ROOHJH RI (QJLQHHULQJ 0DGHO\Q 0 /RFNKDUW 'HDQ *UDGXDWH 6FKRRO

PAGE 159

81,9(56,7< 2) )/25,'$