Modification of UCT with Patterns in Monte-Carlo Go

TLDR

UCT, an extension of UCB1 for multi‑armed bandits, has been applied to minimax tree search. The authors modify UCT for Go by incorporating pattern‑based random simulations, pruning large‑board search, and parallelization, markedly improving MoGo’s performance. MoGo, the first UCT‑based Go program, achieves top‑level performance on 9×9 and 13×13 boards.

Abstract

Algorithm UCB1 for multi-armed bandit problem has already been extended to Algorithm UCT (Upper bound Confidence for Tree) which works for minimax tree search. We have developed a Monte-Carlo Go program, MoGo, which is the first computer Go program using UCT. We explain our modification of UCT for Go application and also the intelligent random simulation with patterns which has improved significantly the performance of MoGo. UCT combined with pruning techniques for large Go board is discussed, as well as parallelization of UCT. MoGo is now a top level Go program on $9\times9$ and $13\times13$ Go boards.