Why Choose Model-Based Reinforcement Learning