This note applies large deviation-based optimality theory to evaluate treatment rules for treatment assignment problems. We find nearly optimal treatment rules whose asymptotic maximum large deviation risks can be arbitrary close to the corresponding minimax bounds.