Early Playout Termination in MCTS
Many researchers view mini-max and MCTS-based searches as competing and incompatible approaches. For example, it is generally agreed that chess and checkers require a mini-max approach while Go and Havannah require MCTS. However, a hybrid technique is possible that has features of both mini-max and MCTS. It works by stopping the random MCTS playouts early and using an evaluation function to determine the winner of the playout. We call this algorithm MCTS-EPT (MCTS with early playout termination) and study it using MCTS-EPT programs we have written for Amazons, Havannah, and Breakthrough.
- 2.Coulom, R.: Efficient selectivity and backup operators in monte-carlo tree search. In: 5th International Conference on Computers and Games, CG 2006, Turin, Italy, pp. 72–84 (2006)Google Scholar
- 3.Gelly, S., Silver, D.: Combining online and offline knowledge in UCT. In: Ghahramani, Z. (ed.) Proceedings of the 24th International Conference on Machine Learning (ICML 2007), pp. 273–280. ACM, New York (2007)Google Scholar
- 5.Kloetzer, J., Iida, H., Bouzy, B.: The monte-carlo approach in amazons. In: Computer Games Workshop, Amsterdam, The Netherlands, pp. 113–124 (2007)Google Scholar
- 7.Lorentz, R., Horey, T.: Programming breakthrough. In: van den Herik, H.J., Iida, H., Plaat, A. (eds.) CG 2013. LNCS, vol. 8427, pp. 49–59. Springer, Heidelberg (2013) Google Scholar
- 12.Havannah#The Havannah Challenge. https://chessprogramming.wikispaces.com/