Study of strategy selection based on Phantom Go

Recent years, Monte Carlo and UCT (Upper bound Confidence for Tree) have been widely used in algorithms of computer games, and considerably improved computer Go programs. Phantom Go is a game with incomplete information and different with Go. The study of Phantom Go is just beginning. Most algorithms of Phantom Go applying Monte Carlo or UCT directly, but results are not quite satisfactory. In this paper, we analysis the characteristic of Phantom go, apply several strategies in different stages and try to improve the competition of program. Our experimental findings indicate that strategies selection method can be applied to Phantom Go effectively.