What Learning Models Tell Us to Expect in Three-by-Three Bimatrix Games

Abstract

This paper extends several existing learning models to investigate their fixed points (their long run predictions of play). The fixed points of the model are not necessarily at the Nash equilibria of the payoff matrices but are a function of both the Nash equilibria and the parameters of the model. The stability of these fixed points also depends on both the characteristics of payoff matrix used and the parameters of the model. These new findings indicate that behaviour previously thought to be inconsistent with theory may not necessarily be so.