Dynamic Multi-Pair Trading Strategy in Cryptocurrency Markets with Deep Reinforcement Learning
Researchers have developed a novel deep reinforcement learning (DRL) strategy for pair trading in volatile cryptocurrency markets. This system utilizes a hierarchical "Filter-then-Rank" methodology and a "Fixed Risk, Adaptive Mean" execution model, incorporating a Proximal Policy Optimization (PPO) agent with an LSTM layer. Evaluated on Binance USD-M Futures data, the DRL policy significantly outperformed traditional heuristic baselines, demonstrating statistically significant risk-adjusted outperformance at the 10 percent level. AI
IMPACT Introduces a novel framework for applying DRL to cryptocurrency pair trading, potentially improving risk management and performance in volatile markets.