Robust Multi-Asset Framework for Automated Pairs Trading
Pairs trading is a market-neutral statistical arbitrage strategy that exploits temporary divergences in the price relationship between two historically correlated assets. Despite extensive research in backtesting contexts, deploying such strategies in real markets remains limited. This work presents a fully automated multi-asset pairs trading framework designed for real-world operation. The system uses Principal Component Analysis (PCA) and agglomerative clustering for search-space reduction, portfolio optimization based on NSGA-II, statistical validation, and liquidity filtering. Trading decisions rely on Z-score thresholds combined with directional confirmation to improve entry precision. The implemented system autonomously executes and monitors trades in real time. Backtesting on S&P 500 data (2018–2025), achieved a total return of 355.13% (24.87% annualized), a Sharpe Ratio of 0.84, and a Maximum Drawdown of -28.34%, outperforming traditional sector-based methods. In live cryptocurrency trading, it generated a 1.42% net profit in a single day with a profit factor of 11.61, confirming the framework's operational robustness.