Grouping nodes for MCTS in imperfect information games

Abstract

Monte Carlo Tree Search (MCTS) with grouping nodes technique improves search efficiency. There is the more playouts in MCTS , the more efficient the tree search is. In this study, we improvie the efficiency of MCTS by grouping in imperfect information games. Imperfect information games has a large number of move nodes in search tree due to unknown information. We are planging to show whether grouping is also effective in imperfect information game. The "Dai Hinmin" that is alst a multiplayer imperfect information game is the subject of veriflcation experiments.