Reinforcement learning can address complex trading strategies, but it remains strongly influenced by the specifics of each task, and sampling effectiveness remains a significant challenge. In this paper, we have compared different deep reinforcement learning algorithms used to master the task of trading. We have specifically examined Deep Q-Networks (DQN), Advantage Actor-Critic (A2C), and Proximal Policy Optimization (PPO) methods. In order to ensure a fair evaluation, we configured these algorithms with optimized hyperparameters according to the best practices recommended in the literature. Our analysis reveals that PPO is superior to the DQN and A2C algorithms, particularly in terms of total profit and risk management.