Evaluation of Monte Carlo tree search and the application to Go

Abstract

Abstract-Recent improvements to Monte Carlo tree search have produced strong computer Go programs. This paper presents a method of measuring the accuracy of Monte Carlo tree search in game programming. We use the win percentage of positions in a large database of game records as a benchmark and compare the win probability obtained by simulations with the benchmark. By applying our method to Monte Carlo tree search in Go, we found differences between search methods and their parameters, and the effect of the properties of positions such as the move numbers and the existence of stones in threats. This paper also introduces numerical metrics to evaluate the performance of search methods. Our experiments in Go, as well as Chess, Othello, and Shogi revealed that the metrics were quite close to our empirical understanding of the performance of various search methods and their parameters.

A Survey of Monte Carlo Tree Search Methods

“…Takeuchi et al [210], [211] use the relationship between the win probability obtained from playouts with actual games to calculate evaluation curves, which allow the comparison of different search methods, search parameters, and search performance at different stages of the game. These measurements promise to improve performance in Go and other MCTS applications.…”

“…Takeuchi et al [210], [211] compare the win probabilities obtained for various search methods, including UCT, to those observed in actual games, to evaluate the effectiveness of each search method for Othello. Othello remains an open challenge for future MCTS research.…”

“…The relatively poor performance of UCT for Chess compared to other games may also be due to the occurrence of trap states (3.5) [162]. Takeuchi et al [210], [211] compare the win probabilities obtained for various search methods, including UCT, to those observed in actual games, to investigate the effectiveness of each search method for Chess.…”

The Application Status and Development Trend of Cloud Computing Technology in Chess

“…The research findings indicated that while there were some similarities, AlphaZero did not replicate human history entirely, exhibiting distinct characteristics in opening choices and move diversity. [4,5].…”