Treffer: Applying reinforcement learning in Bitcoin trading to select technical strategies based on Deep Q-Network.

This paper proposes a Deep Q-Network-based reinforcement learning framework to optimize the selection of technical trading strategies for Bitcoin. Instead of directly predicting prices, the agent chooses among five predefined strategies—RSI, SMA Crossover, Bollinger Bands, Momentum-20d, and VWAP Reversion—allowing actions to remain interpretable while adapting dynamically to market conditions. Market states are constructed from technical indicators, normalized, and reduced in dimensionality using Principal Component Analysis to ensure robustness and efficiency. Empirical tests on Bitcoin data from 2022 to mid-2025 show remarkable performance: starting from $1,000,000, the agent achieved more than 120-fold growth in Net Asset Value, significantly outperforming Buy-and-Hold and single-strategy benchmarks. The model flexibly emphasized momentum during uptrends, limited mean-reversion during breakouts, and managed drawdowns when temporarily misaligned. Out-of-sample results further confirmed generalization. Overall, the framework demonstrates a practical, explainable, and scalable approach to algorithmic trading in digital asset markets. Impact statement: This study proposes a reinforcement learning framework that selects among predefined technical strategies for Bitcoin trading, improving interpretability and reducing model instability. Using Deep Q-Networks with PCA-based features, the agent achieves over 120-fold NAV growth and shows strong adaptability across different market regimes. The findings demonstrate a practical and scalable approach to algorithmic trading in highly volatile digital asset markets. [ABSTRACT FROM AUTHOR]

Copyright of Cogent Economics & Finance is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)