This paper provides a systematic review of Deep Reinforcement Learning (DRL) applications in low-frequency quantitative algorithmic stock trading. The authors analyze the current state of research in this emerging field, examining how DRL agents—which have demonstrated remarkable success in complex games like Chess and Go—are being adapted to financial markets. The review focuses specifically on automated trading systems that treat historical price series as imperfect information environments where the goal is to maximize returns while minimizing risk.

The analysis reveals that while most studies show statistically significant performance improvements over traditional baseline strategies, the research remains largely theoretical with limited practical application. The majority of experiments are conducted in unrealistic simulated environments without real-time trading implementation, and few achieve meaningful profitability levels. The paper identifies critical gaps in the literature, including the absence of comparative studies between different DRL architectures and human traders, as well as insufficient testing in live market conditions. The authors conclude that DRL holds substantial promise for quantitative trading but remains in early developmental stages, requiring more rigorous real-world validation and methodological refinement before it can reliably compete with professional traders.