Lt. Gilbert S. Daniels and the myth of averages — LessWrong