Review for NeurIPS paper: Near-Optimal Reinforcement Learning with Self-Play