Game Theory Optimization Based on Python: Evidence from Board Game “Go-Moku”

Abstract

Game theory has been widely used in multiple fields such as economics, computer science, and political science to study the rational behavior of multiple decision makers in different scenarios. In recent years, with the rapid development of artificial intelligence, AI trained using game theory has shown outstanding performance in board games such as Go, chess and backgammon. This study explores the application of game theory optimization in the board game "Go-Moku" using a learning algorithm combining Monte Carlo tree search algorithm and reinforcement learning. The paper discusses recent developments in game theory in Go-Moku and explains the Monte Carlo tree search algorithm in detail. The performance of the algorithm is evaluated through experimental results of the application of Alpha Zero in the Go-Moku domain, demonstrating its effectiveness in improving the gaming capabilities of artificial intelligence. According to the analysis, after 1,100 training sessions, the algorithm combining reinforcement learning and MCTS had a 9:1 win rate compared to a pure MCTS approach with 2,000 self-simulations per step. And playing against the 800-times pure MCTS method, it reached a 10:0 win rate for the first time in the 300th game. In addition, the paper discusses potential applications of game-theoretic optimization in other dynamic games. Overall, these results shed light on further exploration of game theory in the area of complete information games and reinforcement learning.