Predictive Systems: The Game Rock-Paper-Scissors as an Example

Beating humans in a penny-matching game by leveraging cognitive hierarchy theory and Bayesian learning

“…Developing artificial intelligence (AI) to beat humans in strategic games has been drawing attention/interest of researchers for decades [1]- [10]. Although AI algorithms targeted at specific games may not directly contribute to solving practical engineering problems other than those in the gambling industry, the theories developed alongside can be used to attack many problems of similar natures and of greater significance [1].…”

“…For instance, this holds for the cases where perfectly rational decision strategies are wellknown, including the game of matching pennies (or "odds and evens") [11] and the game of rock-paper-scissors [12]. For such games, recognizing the decision pattern of the opponent human player often plays a vital role in developing winning strategies for the AI player [6]- [10].…”

“…Their AI algorithms pursue the identification of these patterns and assume that the human player will follow the same patterns the next time the same situation arises. Some later works, including [8]- [10] for the game of rock-paper-scissors, essentially follow the same principle to design their AI algorithms.…”

“…2) Our algorithm exploits the level-k framework [13], [14] of cognitive hierarchy theory, which has not been considered in previously developed AI algorithms for matching-penny or rock-paper-scissors games [6]- [10].…”