Distinguishing logistic curves — LessWrong