Using search strategies at inference-time can provide massive performance boost on numerous complex reinforcement learning tasks, within only a couple seconds of execution time.