“The challenge of large hidden state spaces in complex games is primarily a problem for search-based AI techniques, not for model-free reinforcement learning approaches.”