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