Exploring Pessimism and Optimism Dynamics in Deep Reinforcement Learning