Improvement on Amazons Search Algorithm

Abstract

The Game of the Amazons is a two-player abstract territory game. It has attracted attention in game research because of its simple rules and its complexity of play. A number of Amazons-playing programs have emerged in recent years. Different search algorithms are used in Amazons programs. Minimax is one of the most commonly used Amazons search algorithms. It is usually optimized with augmentations, such as alpha-beta pruning, transposition table, history heuristic, and self-play learning. In this paper, we make a new improvement on the search algorithm by using the game progress to modify search depths dynamically. We explain the improvement and its implementation in detail. To test the effectiveness of the improvement, we conduct comparative experiments that test program search time and let our program play against other Amazons programs without the improvement. The experimental results show that the improvement reduces program search time and improves the performance of our program.