Kalman Filter Pairs Trading in Chinese Futures
Summary
This project outlines a mean-reversion pairs strategy for Chinese futures. It screens candidate contract pairs for cointegration with an Augmented Dickey-Fuller test, estimates a changing hedge ratio with a Kalman filter, and measures spread behavior with the Hurst exponent and an estimated half-life. A spread z-score based on rolling statistics supplies entry and exit signals, and equal capital is assigned to each pair in the portfolio.
The article reports in-sample and out-of-sample backtest statistics for daily data, including returns, Sharpe ratios, and drawdowns. It also notes that a mean-reverting spread may take too long to trade, and that the tests omit transaction costs and slippage. Other limitations include reliance on main-contract data, choices about training length and filter recalibration, and unresolved differences in the stated test periods and reported performance figures. The results are historical and do not establish live-trading performance.
Key ideas
- Candidate futures pairs are screened for cointegration before trading their spread.
- A Kalman filter updates the hedge ratio over time without a fixed rolling window.
- Spread z-scores trigger entries on large deviations and exits near the mean.
- The Hurst exponent and half-life help assess mean-reversion behavior and its speed.
- Reported backtests use daily data but omit fees and slippage.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.