Backtesting Intraday Mean Reversion Between SPY and IWM

Summary

The article describes an intraday pairs strategy for SPY and IWM, using one-minute bars from April 2007 through February 2014. It aligns timestamps, estimates a rolling regression hedge ratio, forms a price spread, and converts that spread to a z-score. The strategy enters a long or short spread when the z-score passes an absolute threshold of 2 and exits when it returns within an absolute threshold of 1. The author also proposes varying the regression lookback and shows a sensitivity chart with a reported local performance maximum near 110 bars.

The charts show volatility during the 2009 financial crisis and weaker performance in the final year as SPY trended strongly. The author explicitly flags lookahead bias from calculating spread mean and standard deviation over the full sample, and says a rolling calculation is needed. Results omit fees, bid-ask spread, and slippage, while fractional ETF positions are unrealistic. The article warns that these issues could make the strategy perform poorly in live trading.

Key ideas

• The strategy trades a spread between SPY and IWM using a rolling regression hedge ratio.
• It enters when the spread z-score moves beyond an absolute threshold of 2 and exits inside a threshold of 1.
• The article examines sensitivity to the regression lookback and reports a local maximum near 110 bars.
• The presented z-score calculation uses future data, creating lookahead bias that should be removed with rolling statistics.
• The reported backtest omits transaction costs and assumes fractional ETF positions.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.