Optimizing Market Making Using Deep Reinforcement Learning

In this paper, deep reinforcement learning paradigm is applied to the problem of high frequency algorithmic stock trading for market making purposes. While the application of deep reinforcement learning to the HFT is not entirely novel in the academic realm, its thorough exploration and analysis are still ongoing, as evidenced by the current re- search efforts in this area. Advanced policy gradient-based algorithm was selected for conducted experiments as agent interacting within the environment, consisting of LOB, TFI and OFI snapshots and other indicators. For the context of this paper, the proposed framework is applied and studied on the stock market. The goal of this paper is to push further the level of generalization and performance of DRL stock trading by optimizing it within created framework. The goal of this paper is to further advance the general- ization and performance of DRL in stock trading by optimizing it within the framework established based on existing research for market making purposes. Drawing on insights from prior studies, this work aims to enhance the functionality of a DRL-based market maker, leveraging ideas and methodologies proposed in the literature to achieve improved performance and adaptability in real-world trading scenarios. Keywords: deep reinforcement learning, market making, high-frequency trading, algorithmic trading