Evolution, Uncertainty, and the Asymptotic Efficiency of Policy