Meta Monte-Carlo Tree Search for Automatic Opening Book Generation

Modern game playing programs use opening books in order to perform better. Generating opening books automatically in combination with an αβ program has been well studied. A challenge is to generate automatically an opening book for the new Monte-Carlo Tree Search (MCTS) algorithms. In this article, we tackle this issue by combining two level of MCTS. The resulting algorithm is called Meta-MCTS. Instead of applying a simulation strategy, it uses an MCTS program to play a simulated game. We describe two Meta-MCTS algorithms: the first one, Quasi Best-First, favors exploitation. The second one, Beta Distribution Sampling, favors exploration. Our approach is generic and can be used for general game playing. Itwill be particularly usefulwhen there is off-line time avail- able. In order to evaluate the performances of these algorithms, we generated and tested 99 Go opening books.