AI trained with self-play is unstoppable (Reinforcement Learning)

I used reinforcement learning and self-play to train an AI for a melee combat game Code: https://github.com/HarrisSteven/RL-Game