Exploring AI Play Patterns in Xiangqi

Abstract: In this study, we aim to explore the characteristics of different AI algorithms using the classical Chinese board game xiangqi as the testbed. We did this by utilizing four different build-in AI agents in Ludii: Random, Flat MC, UCT and Minimax, and having them playing against each other in all possible match ups and exploring their play patterns. The result shows that in terms of the most basic play patterns, i.e., the win/loss rate, we can arrive at a skill ranking as follows: Minimax > UCT > Flat MC > Random. When it comes to more specific metrics such as piece frequency and taking out, Minimax also stands out from all the other three AI agents. We believe this study is a good starting point for looking into deeper how these AI agents differ from each other. The study also suggests that just the type of algorithm and how it is implemented can already possibly affect the observed play patterns and for further research exploring mapping between AI and play types it is important to be familiar with the default playing pattern of these AI algorithms before turning them into specific player types.