Reinforcement Learning for Dynamically Scaled Cryptocurrency Pair Trading

Summary

This study investigates whether reinforcement learning can improve cryptocurrency pair trading, a statistical arbitrage approach that trades price differences between correlated assets. The authors build trading environments and train agents to choose when and how to trade cryptocurrency pairs. They also develop reward shaping and observation and action spaces for the learning task, with the aim of adapting trading decisions dynamically.

Experiments use one-minute BTC-GBP and BTC-EUR data. The reported annualized profit for traditional pair trading is 8.33%, compared with a range from 9.94% to 31.53% for the tested reinforcement learning agents, depending on the learner. These results indicate stronger historical performance in the reported experiments, but they do not by themselves establish live profitability. The document does not specify transaction costs, risk-adjusted comparisons, or out-of-sample validation details, so the results should be read within the limits of the described data and evaluation.

Key ideas

• The study applies reinforcement learning to pair trading between correlated cryptocurrencies.
• Agents are trained to choose trade timing and sizing within constructed environments.
• The approach uses custom reward shaping and observation and action spaces.
• Experiments cover BTC-GBP and BTC-EUR at one-minute intervals.
• The reported RL profit range exceeds the traditional strategy result, but the abstract leaves live robustness and costs unclear.

Tags

# Reinforcement Learning Pair Trading: A Dynamic Scaling approach


# Reinforcement Learning Pair Trading: A Dynamic Scaling approach









Cryptocurrency is a cryptography-based digital asset with extremely volatile prices. Around USD 70 billion worth of cryptocurrency is traded daily on exchanges. Trading cryptocurrency is difficult due to the inherent volatility of the crypto market. This study investigates whether Reinforcement Learning (RL) can enhance decision-making in cryptocurrency algorithmic trading compared to traditional methods. In order to address this question, we combined reinforcement learning with a statistical arbitrage trading technique, pair trading, which exploits the price difference between statistically correlated assets. We constructed RL environments and trained RL agents to determine when and how to trade pairs of cryptocurrencies. We developed new reward shaping and observation/action spaces for reinforcement learning. We performed experiments with the developed reinforcement learner on pairs of BTC-GBP and BTC-EUR data separated by 1 min intervals (n=263,520). The traditional non-RL pair trading technique achieved an annualized profit of 8.33%, while the proposed RL-based pair trading technique achieved annualized profits from 9.94% to 31.53%, depending upon the RL learner. Our results show that RL can significantly outperform manual and traditional pair trading techniques when applied to volatile markets such as~cryptocurrencies.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.