Optimization and Improvement for the Game 2048 Based on the MCTS Algorithm

The game 2048 is a popular and fashionable mobile telephone game, which is characterized by the strong uncertainty. The Monte Carlo Tree Search is always used to be applied for the game search algorithm in the field of the computer games. The basic algorithm is usually divided into four steps: path selection, node expansion, simulation experiment and node update. In the step of path selection, the paper proposes the optimized and improved strategy for selecting nodes based on the UCT algorithm. In the step of simulation experiment, the two effective and heuristic simulation strategies are designed and applied based on the parallel computer games system coded by own self. In the steps of node expansion and node update, the adaptive search strategy is provided to improve the number of search times according to the complexity of the game situations. Through a plenty of experiments, the experimental results show that the provided optimization and improvement algorithm is able to be applied for the game 2048 effectively and practically to reach the senior level finally.