BALANCING OPTIMAL LARGE DEVIATIONS IN RANKING AND SELECTION