Text this: People Can Adaptively Exploit Model-free and Model-based Reinforcement Learning in Competitive Games.