Playing Amazons Endgames 1

Abstract

The game of the Amazons is a fairly young member of the class of territory games. Since few humans play the game, it is difficult to estimate the level of current programs. However, it is believed that humans could play much stronger than today’s programs, given sufficient training and incentives. With the more general goal of improving the playing level of Amazons programs in mind, we focus here on the playing of endgame positions. Our study compares two solvers, DFPN and WPNS, and three game-playing algorithms, Minimax with Alpha/Beta, Monte-Carlo Tree Search, and Temperature Discovery Search. We show that even though the computing process is quite expensive, traditional PNS-based solvers are best suited for the task of finding the best moves in a subgame of Amazons. No specific improvement is needed to classical game-playing engines to play well in the subdomain of Amazons endgames. Moreover, we show that Monte-Carlo Tree Search (MCTS), the new Amazons standard plays the Amazons endgames pretty well, despite their frequently occurring weaknesses in handling precise tasks such as solving.