Evolving Gomoku solver by genetic algorithm

Gomoku, also known as Gobang or five-in-a-row, is a popular two-player strategical board game. Given a squared 15×15 board, two players compete to first obtain an unbroken row of five pieces horizontally, vertically or diagonally. Classic methods for solving such games are based on game-tree theory, for example the minimax tree. These methods have a clear disadvantage: the depth of search becomes a bottleneck all the time. In this paper we propose a genetic algorithm for solving the Gomoku game. We investigated the general framework for applying genetic algorithm to strategical games and designed the fitness function from various game-related aspects. Empirical experimental results showed that the proposed genetic solver can search in greater depth than traditional game-tree-based solvers, resulting in better and more enjoyable solutions, and does so more efficiently.