A Reddit discussion proposed the idea of using Nuzlocke (a self-imposed high-difficulty Pokémon challenge mode) game data to train a reinforcement learning environment.