Testing for a Sweet Spot in Randomized Trials