The Problem of "Win-More" — LessWrong