Learning middle-game patterns in chess: A case study

Miroslav Kubat; Jan Žižka

doi:10.1007/3-540-45049-1_52

Learning middle-game patterns in chess: A case study

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Despite the undisputed strength of today’s chess-playing programs, the fact that they have to evaluate millions, or even billions, of different positions per move is unsatisfactory. The amount of “computation” carried out by human players is smaller by orders of magnitudes because they employ specific patterns that help them narrow the search tree. Similar approachs hould in principle be feasible also in computer programs. To draw attenion to this issue, we report our experiments with a program that learns to classify chessboard positions that permit the well-known bishop sacrifice at h7. We discuss some problems pertaining to the collection of training examples, their representation, and pre-classification. Classification accuracies achieved with a decision-tree based classifier are encouraging.

Original language	English (US)
Title of host publication	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Publisher	Springer Verlag
Pages	426-433
Number of pages	8
Volume	1821
ISBN (Print)	3540676899, 9783540450498
DOIs	https://doi.org/10.1007/3-540-45049-1_52
State	Published - 2000
Externally published	Yes
Event	13th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 2000 - New Orleans, United States Duration: Jun 19 2000 → Jun 22 2000

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	1821
ISSN (Print)	03029743
ISSN (Electronic)	16113349

Other

Other	13th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 2000
Country	United States
City	New Orleans
Period	6/19/00 → 6/22/00

Keywords

Computer chess
Concept learning
Pattern recognition

ASJC Scopus subject areas

Computer Science(all)
Theoretical Computer Science

Access to Document

10.1007/3-540-45049-1_52

Fingerprint Dive into the research topics of 'Learning middle-game patterns in chess: A case study'. Together they form a unique fingerprint.

Cite this

Kubat, M., & Žižka, J. (2000). Learning middle-game patterns in chess: A case study. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 1821, pp. 426-433). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1821). Springer Verlag. https://doi.org/10.1007/3-540-45049-1_52

Learning middle-game patterns in chess : A case study. / Kubat, Miroslav; Žižka, Jan.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1821 Springer Verlag, 2000. p. 426-433 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 1821).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Kubat, M & Žižka, J 2000, Learning middle-game patterns in chess: A case study. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 1821, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 1821, Springer Verlag, pp. 426-433, 13th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 2000, New Orleans, United States, 6/19/00. https://doi.org/10.1007/3-540-45049-1_52

Kubat M, Žižka J. Learning middle-game patterns in chess: A case study. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1821. Springer Verlag. 2000. p. 426-433. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/3-540-45049-1_52

Kubat, Miroslav ; Žižka, Jan. / Learning middle-game patterns in chess : A case study. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 1821 Springer Verlag, 2000. pp. 426-433 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{effcfda4bf0c4927bd92c202f03584b8,

title = "Learning middle-game patterns in chess: A case study",

abstract = "Despite the undisputed strength of today{\textquoteright}s chess-playing programs, the fact that they have to evaluate millions, or even billions, of different positions per move is unsatisfactory. The amount of “computation” carried out by human players is smaller by orders of magnitudes because they employ specific patterns that help them narrow the search tree. Similar approachs hould in principle be feasible also in computer programs. To draw attenion to this issue, we report our experiments with a program that learns to classify chessboard positions that permit the well-known bishop sacrifice at h7. We discuss some problems pertaining to the collection of training examples, their representation, and pre-classification. Classification accuracies achieved with a decision-tree based classifier are encouraging.",

keywords = "Computer chess, Concept learning, Pattern recognition",

author = "Miroslav Kubat and Jan {\v Z}i{\v z}ka",

year = "2000",

doi = "10.1007/3-540-45049-1_52",

language = "English (US)",

isbn = "3540676899",

volume = "1821",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "426--433",

booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

note = "13th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 2000 ; Conference date: 19-06-2000 Through 22-06-2000",

}

TY - GEN

T1 - Learning middle-game patterns in chess

T2 - 13th International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems, IEA/AIE 2000

AU - Kubat, Miroslav

AU - Žižka, Jan

PY - 2000

Y1 - 2000

N2 - Despite the undisputed strength of today’s chess-playing programs, the fact that they have to evaluate millions, or even billions, of different positions per move is unsatisfactory. The amount of “computation” carried out by human players is smaller by orders of magnitudes because they employ specific patterns that help them narrow the search tree. Similar approachs hould in principle be feasible also in computer programs. To draw attenion to this issue, we report our experiments with a program that learns to classify chessboard positions that permit the well-known bishop sacrifice at h7. We discuss some problems pertaining to the collection of training examples, their representation, and pre-classification. Classification accuracies achieved with a decision-tree based classifier are encouraging.

AB - Despite the undisputed strength of today’s chess-playing programs, the fact that they have to evaluate millions, or even billions, of different positions per move is unsatisfactory. The amount of “computation” carried out by human players is smaller by orders of magnitudes because they employ specific patterns that help them narrow the search tree. Similar approachs hould in principle be feasible also in computer programs. To draw attenion to this issue, we report our experiments with a program that learns to classify chessboard positions that permit the well-known bishop sacrifice at h7. We discuss some problems pertaining to the collection of training examples, their representation, and pre-classification. Classification accuracies achieved with a decision-tree based classifier are encouraging.

KW - Computer chess

KW - Concept learning

KW - Pattern recognition

UR - http://www.scopus.com/inward/record.url?scp=84957899807&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84957899807&partnerID=8YFLogxK

U2 - 10.1007/3-540-45049-1_52

DO - 10.1007/3-540-45049-1_52

M3 - Conference contribution

AN - SCOPUS:84957899807

SN - 3540676899

SN - 9783540450498

VL - 1821

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 426

EP - 433

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

PB - Springer Verlag

Y2 - 19 June 2000 through 22 June 2000

ER -