RVAB: Rational Varied-Depth Search in Siguo Game

Game playing is one of the classic problems of artificial intelligence. The Siguo game is an emerging field of research in the area of game-playing programs. It provides a new test bed for artificial intelligence with imperfect information. To improve search efficiency for Siguo with more branches and the uncertain payoff in the game tree, this paper presents a modified Alpha-Beta Search algorithm, which is called rational varied-depth Alpha-Beta (RVAB). The RVAB The basic ideas of RVAB algorithm is : if player get much information about opponent during playing game, they can do more deep thinking about strategies of game. If player can only get few, very uncertain information about opponent during playing game, they don’t think more deep about their strategy because it is worthless for player to speculate the strategies of game under very uncertain. Experiments show that RVAB achieves the goals of the improvability of visited nodes efficiency, although it costs a little more memory.