LEARNING A MIXED STRATEGY EQUILIBRIUM IN THE LABORATORY

This paper examines a laboratory game with no pure strategy equilibrium, and one mixed strategy equilibrium. A simple adaptive model describes players' strategy adjustments reasonably well, as long as predictable adjustments are not easily exploited by others. Consistent with theory, these adjustments do not lead players to equilibrium. When predictable adjustments are made more easily exploited, players deviate from adaptive adjustments in order to avoid exploitation and exploit others. These deviations cause players to converge more strongly to equilibrium. These results confirm both the predictive power of adaptive models in some settings, and the instability of mixed strategy equilibria.