Deep learning and computer chess (part 2)

• Main Author: Ngoh, Guang Wei
• Other Authors: He Ying
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/144972
• Institution: Nanyang Technological University

Monte Carlo Tree Search (MCTS) is a probabilistic algorithm that has gained traction in recent years. MCTS uses lightweight random simulations to selectively grow a game tree and has experienced success in domains with vast search spaces, such as chess. This project explores the usage of the MCTS algorithm in chess engines as well as the various ways MCTS can be improved beyond the base algorithm through the use of a static board state evaluation function. Methods such as early playout termination, implicit minimax backups, MCTS-Solver, as well as a few novel methods were implemented with their results being analyzed and discussed. The implemented methods proved to be a promising step in the right direction in developing a MCTS chess engine that can rival and potentially outperform chess engines that use deterministic algorithms with further development.