Predicting how People Play Games: a Simple Dynamic Model of Choice.

We use the model developed in Sarin and Vahid (1999, GEB) to explain the experiments reported in Erev and Roth (1998, AER). The model supposes that players maximize subject to their "beliefs" which are non-probabilistic and scalar-valued. They are intended to describe the payoffs the players subjectively assess they will obtain from a strategy. In an earlier paper (Sarin and Vahid (1997) we showed that the model predicted behaviour in repeated coordination games remarkably well, and better than equilibrium theory of reinforcement learning models. In this paper we show that the same one-parameter model can also explain behaviour in games with a unique mixed strategy Nash equilibrium better than alternative models. Hence, we obtain further support for the simple dynamic model.