A curling agent based on the monte-carlo tree search considering the similarity of the best action among similar states

Abstract

Curling is one of the most strategic winter sports. Recently, many computer scientists have studied curling strategies. The Digital Curling system is a framework used to compare curling strategies. Herein, we present a computer agent based on the Monte-Carlo Tree Search (MCTS) for the Digital Curling framework. We implemented a novel action decision method based on MCTS for Markov decision processes with continuous state space. The experimental results show that our search method is effective for agents with a simple simulation policy and agents with a handmade complex one.