Sampling Best Response Dynamics and Deterministic Equilibrium Selection