Mining Expert Play to Guide Monte Carlo Search in the Opening Moves of Go

AAAI Conferences

We propose a method to guide a Monte Carlo search in the initial moves of the game of Go. Our method matches the current state of a Go board against clusters of board configurations that are derived from a large number of games played by experts. The main advantage of this method is that it does not require an exact match of the current board, and hence is effective for a longer sequence of moves compared to traditional opening books. We apply this method to two different open-source Go-playing programs. Our experiments show that this method, through its filtering or biasing the choice of a next move to a small subset of possible moves, improves play effectively in the initial moves of a game.


BTT-Go: An Agent for Go that Uses a Transposition Table to Reduce the Simulations and the Supervision in the Monte-Carlo Tree Search

AAAI Conferences

This paper presents BTT-Go: an agent for Go whose ar- chitecture is based on the well-known agent Fuego, that is, its search process for the best move is based on sim- ulations of games performed by means of Monte- Carlo Tree Search (MCTS). In Fuego, these simulations are guided by supervised heuristics called prior knowledge and play-out policy. In this context, the goal behind the BTT-Go proposal is to reduce the supervised character of Fuego, granting it more autonomy. To cope with this task, the BTT-Go counts on a Transposition Table (TT) whose role is not to waste the history of the nodes that have already been explored throughout the game. By this way, the agent proposed here reduces the super- vised character of Fuego by replacing, whenever pos- sible, the prior knowledge and the play-out policy with the information retrieved from the TT. Several evalua- tive tournaments involving BTT-Go and Fuego confirm that the former obtains satisfactory results in its purpose of attenuating the supervision in Fuego without losing its competitiveness, even in 19x19 game-boards.


Neural Networks Learning the Concept of Influence in Go

AAAI Conferences

This paper describes an intelligent agent that uses a MLP (Multi-Layer Perceptron) Neural Network (NN) in order to evaluate a game state in the game of Go based, exclusively, in an influence analysis. The NN learns the concept of Influence, which is domain specific to the game of Go. The learned function is used to evaluate board states in order to predict which player will win the match. The results show that, in later stages, the NN can achieve an accuracy of up to 89.3% when predicting the winner of the game. As future work the authors propose the incorporation of several improvements to the NN and also its integration intelligent player agents for the game of go, such as Fuego and GnuGo.


Go-Ahead: Improving Prior Knowledge Heuristics by Using Information Retrieved From Play Out Simulations.

AAAI Conferences

The proposal behind this paper is the introduction of a new agent denominated Go-Ahead: this is an automatic Go player that uses a new technique in order to improve the accuracy of the pre estimated values of the moves that are candidates to be introduced into the classical Monte Carlo tree search (MCTS) algorithm which is used by many of the current top agents for Go. Go-Ahead is built upon the framework of one of these agents: the well known open source automatic player Fuego, in which these pre estimated values are obtained by means of a heuristic called prior knowledge. Go-Ahead copes with the task of refining the calculations of these values through a new technique that performs a balanced combination between the prior knowledge heuristic and some relevant information retrieved from the numerous play out simulation phases that are repeatedly executed throughout the Monte Carlo search. With such a strategy, Go-Ahead provides the contribution of enhancing the MCTS process of choosing appropriate moves. Further, this new approach attenuates the supervision level inherent to this process due to the following fact: it allows for the lessening of the impact of the prior knowledge heuris- tics through strengthening the impact of play out information. The results obtained in tournaments against Fuego confirm the benefits and the contributions provided by this approach.