Reinforcement learning achieves human expert–level performance in the large-scale imperfect information board game Stratego.