Consider a generalization of fictitious play in which agents' choices
are perturbed by incomplete information about what the other side has
done, variability in their payoffs, and unexplained trembles. These pe
rturbed best reply dynamics define a nonstationary Markov process on a
n infinite state space. It is shown, using results from stochastic app
roximation theory, that for 2 X 2 games it converges almost surely to
a point that lies close to a stable Nash equilibrium, whether pure or
mixed. This generalizes a result of Fudenberg and Kreps, who demonstra
te convergence when the game has a unique mixed equilibrium. (C) 1995
Academic Press, Inc.