General reinforcement learning algo masters chess, shogi, Go through self-play