Averaged learning of human players