 Constant-risk limit-of-Bayes-risk criterion

If a rule has constant risk $r$ and a sequence of priors has Bayes risks tending to $r$, then the rule is minimax. Every competing rule has worst-case risk at least every one of those Bayes risks.