Multi-agent modeling of interaction-based card games

Many card games have their strategy defined at high levels of play not by statistical probabilities of cards being drawn, but rather by the interactions of the players themselves. These interactions, often based on prediction of the opposing players ' likely decisions and/or holdings, become the determining factors in developing a successful strategy. This play has often been seen as psychological, falling under such labels as bluffing and illogical play. The creation of a system that is truly free enough to develop its own strategy without influence from the limited expertise of the engineer is detailed within, being of crucial importance. Through use of intelligent agents that learn to play a game purely through empirical observation, it is shown that agents can in fact be created that are capable of reproducing such behaviour, directly disputing the notion of such actions being either illogical or psychological in nature, but rather resulting from incorporating the opponents into the modelled system.