Value In Reinforcement Learning