We propose a learning dynamic with agents using samples of past play to estimate the distribution of other players' strategy choices and best responding to this estimate. To account for noisy play, estimated distributions over other players' strategy choices have full support in the other...

In a dynamic model of assignment problems, small deviations suffice to move between stable outcomes. This result is used to obtain no-selection and almost-no-selection results under the stochastic stability concept for uniform and payoff-dependent errors. There is no-selection of partner or...