Proposal of simple and strategic two-person zero-sum finite uncertain imperfect information game and development of AI

Abstract

We would like to advance game research on the psychology of imperfect information games. Rock-paperscissors is famous as an imperfect information game, but it is difficult to add strategy because it is too simple. In this research, we extend the game so that it is possible to change hands later, and by incorporating cards and probabilities into the way of changing hands, we propose rock-paper-scissors with a strategy that is not too complicated. Automatic play by AI is important for game analysis. The AI for this game was built using a rule-based approach, but we will also introduce the AI's thinking for imperfect information games obtained during the development of this game.