How Do Reinforcement Learning Algorithms Optimize Trading Strategies in Financial Markets Compared to Traditional Trading Approaches? A Literature Review

Reinforcement learning (RL) has demonstrated significant potential in optimizing sequential decision-making within financial markets' highly dynamic and uncertain environments, offering distinct advantages over traditional trading approaches. This literature review investigates the use of RL in developing and improving trading strategies by integrating the findings of ten recent studies published between 2018 and 2025, selected for their focus on RL applications in different financial domains. These studies employ a range of RL techniques, such as Q-learning, and Proximal Policy Optimization (PPO), across a variety of financial markets, including stocks, Forex, Bitcoin, and derivatives. The review shows that RL-based strategies, which often use innovations such as multi-agent systems, ensemble learning, and sentiment analysis, demonstrate superiority such as better adaptability to non-dynamic stationary market conditions, enhanced risk-adjusted returns, and capability to learn complex relationships directly from market data, thus outperforming conventional methods and market benchmarks. Challenges hindering the practical application of reinforcement learning in trading include sample efficiency, training stability, market complexity, and the necessity for accurate market assumptions. These are areas requiring further examination and enhancement.