Application of Monte Carlo Tree Search algorithm in Go playing

Abstract

The reputation of Monte Carlo Tree Search (MCTS) algorithm is built upon its ability to handle a large number of options and make optimal decisions and detailed plans in situations with limited information. It was initially applied in the game of Go, a traditional and complex combinatorial strategy game, which severely challenged the efficiency of traditional brute-force algorithms that relied on simulating every node and branch in the search tree. The success of AlphaGo, developed by Google, represents a significant advancement in artificial intelligence technology, utilizing deep learning networks and MCTS techniques. In this article, we analyze the limitations of traditional brute-force approaches in Go playing and compare them with the application of MCTS, which overcomes these limitations through the iterative process of four steps: selection, expansion, simulation, and back-propagation. The discussion covers the application of MCTS in the game of Go. Besides, we evaluate the advantages and shortcomings of the MCTS algorithm. The article concludes with a summary and prospects for future research.