How Can Reinforcement Learning Achieve Expert-Level [Chip] Placement?