Email Post: Multi-agent reinforcement learning with approximate model learning for competitive games