Computer shogi

This paper describes the current state of the art in computer shogi. Shogi (Japanese chess) promises to be a good vehicle for future research into game-playing programs that are based on tree-searching paradigms. This paper shows where chess and shogi are similar, and details the important areas that make shogi programming of particular interest. A crucial difference is the game-tree complexity, which is significantly higher in shogi than in chess. Three important differences are the “drop” rule, the diverging character of the game, and the slow build-up of forces. They make it difficult to have effective opening and endgame procedures. After a short summary of the rules of shogi and an outline of the main areas of current work in computer shogi, we provide an overview of the history of computer shogi, in which computershogi activities both in human tournaments and in exhibition events are given. We conjecture that by the year 2010 a computer will be comparable in strength to the best human players. The most important techniques used in computer shogi are described. We focus on issues such as opening play, selective search, quiescence search, solving tactical exchanges without tree searching, position evaluation and endgame play. At the end the key challenges in computer shogi are enumerated, and finally, concluding remarks are given.  2001 Elsevier Science B.V. All rights reserved.