An Dynamic Statistical Arbitrage Trading System

Objective - The paper's aim is to explore an efficient statistical arbitrage system. Methods - The paper use moving-window and Regression model to identify the volatility of the relation between two assets. When the relation move beyond normal range which defined by quantile, arbitrage opportunities occur. Result -Residuals from moving-window regression model is very close to normal distribution. Generally the arbitrage system is profitable under different parameters. An instance show the system's total rate of return is 18.9%. Conclusion -The profit curve tend to be flat in recent years. Parameters used in the framework should be changed intelligently, because misprice may be corrected in shorter or longer term than history.