Tracking the Race Between Deep Reinforcement Learning and Imitation Learning -- Extended Version