GO-MOKU SOLVED BY NEW SEARCH TECHNIQUES

Citation
Lv. Allis et al., GO-MOKU SOLVED BY NEW SEARCH TECHNIQUES, Computational intelligence, 12(1), 1996, pp. 7-23
Citations number
16
Categorie Soggetti
Computer Sciences, Special Topics","Computer Science Artificial Intelligence
Journal title
ISSN journal
08247935
Volume
12
Issue
1
Year of publication
1996
Pages
7 - 23
Database
ISI
SICI code
0824-7935(1996)12:1<7:GSBNST>2.0.ZU;2-R
Abstract
Many decades ago, Japanese professional Go-Moku players stated that Go -Moku (five-in-a-row on a horizontally placed 15 x 15 board) is a won game for the player to move first. So far, this claim has never been s ubstantiated by (a tree of) variations or by a computer program. Meanw hile, many variants of Go-Moku with slightly different rules have been developed. This paper shows that for two common variants, the game-th eoretical value has been established. Moreover, the Go-Moku program Vi ctoria is described. It uses two new search techniques: threat-space s earch and proof-number search. One of the results is that Victoria is bound to win against any (optimal) counterplay if it moves first. Furt hermore, it achieves good results as a defender against nonoptimally p laying opponents. In this contribution we focus on threat-space search and its advantages compared to conventional search algorithms.