Efficient reinforcement learning for reversi AI

Reversi (or Othello) is a simple and popular board game played on an eight-by-eight board. In the field of reinforcement learning, searching of the game tree of Reversi is widely studied as a classic problem, since it has a small board and thus a state space not too complex to analyze. Monte Carlo tree search (MCTS) is a heuristic search algorithm for decision tree search, which is often applied to the AI methods for board games, such as the application of AlphaGo in the field of Go games. We modify and apply the Monte Carlo tree search strategy to Reversi AI. Applying some engineering optimizations (such as multithreading), we achieve significant results with high time efficiency.