Stochastic imitative game dynamics with committed agents

We consider models of stochastic evolution in two-strategy games in which agents employ imitative decision rules. We introduce committed agents: for each strategy, we suppose that there is at least one agent who plays that strategy without fail. We show that unlike the standard imitative model, the model with committed agents generates unambiguous infinite horizon predictions: the asymptotics of the stationary distribution do not depend on the order in which the mutation rate and population size are taken to their limits.