Parallel Market Environments for FinRL Contests

Abstract: Financial reinforcement learning (FinRL) has emerged as a promising paradigm for sequential decision-making in financial engineering. However, applying RL in real-world trading tasks remains challenging due to the non-stationarity of financial data, low signal-to-noise ratios, and various market frictions. Although numerous FinRL methods have been developed for tasks such as trading and portfolio management, the lack of standardized task definitions, datasets, environments, and baselines has hindered consistent evaluation and reproducibility. To bridge this gap, we organized three FinRL Contests from 2023 to 2025, covering a diverse range of financial tasks such as stock trading, order execution, cryptocurrency trading, and the use of LLM-generated signals. These contests attracted 200 participants from over 100 institutions across 22 countries. To promote reproduction, we provided open-source starter kits featuring GPU-optimized parallel market environments and comprehensive documentation. In this paper, we summarize these benchmarking efforts, detailing task formulations, data curation pipelines, environment implementations, evaluation protocols, participant performance, and key organizational insights.