maybe solving chess is a reasonable challenge something that’s interesting: alphago uses reinforcement learning i think? could be wrong. if everybody is using PPO to make world-class game playing engines then maybe you could still be cool by beating them with decision trees