2026-09-20T04:13:28Zhttps://api.drum.lib.umd.edu/server/oai/request

oai:drum.lib.umd.edu:1903/204492018-01-26T04:16:02Zcom_1903_2224com_1903_12com_1903_2col_1903_2756col_1903_3

Fu, Michael Lin, Jonathan Fun Digital Repository at the University of Maryland University of Maryland (College Park, Md.) Systems Engineering 2018-01-25T06:36:18Z 2018-01-25T06:36:18Z 2017 https://doi.org/10.13016/M2VD6P64Q http://hdl.handle.net/1903/20449 Monte Carlo Tree Search (MCTS) has been successfully applied to a variety of games. Its best-first algorithm enables implementations without evaluation functions. Combined with Upper Confidence bounds applied to Trees (UCT), MCTS has an advantage over traditional depth-limited minimax search with alpha-beta pruning in games with high branching factors such as Go. However, minimax search with alpha-beta pruning still surpasses MCTS in domains like Chess. Studies show that MCTS does not detect shallow traps, where opponents can win within a few moves, as well as minimax search. Thus, minimax search performs better than MCTS in games like Chess, which can end instantly (king is captured). A combination of MCTS and minimax algorithm is proposed in this thesis to see the effectiveness of detecting shallow traps in Go problems. en MONTE CARLO TREE SEARCH AND MINIMAX COMBINATION – APPLICATION OF SOLVING PROBLEMS IN THE GAME OF GO Thesis Engineering Monte Carlo Tree Search open.access