Backtesting an Intraday SPY-IWM Mean-Reversion Pairs Strategy
Summary
The document describes an intraday pairs strategy that trades SPY against IWM. It estimates a changing hedge ratio with rolling linear regression, constructs a price spread, and standardizes that spread as a z-score. The strategy enters a long spread when the score falls past a negative threshold and a short spread when it rises past a positive threshold, then exits when the score returns closer to zero. It uses aligned one-minute data and includes a sensitivity analysis that varies the regression lookback. The stated rationale is that the two US equity ETFs reflect related markets and their spread may revert after temporary divergences.
The example reports a lookback-dependent performance peak and describes volatile strategy behavior during the financial crisis, with weaker performance in a strongly trending period. These observations are not reliable evidence of an implementable edge: the article explicitly identifies lookahead bias from calculating spread normalization over the full sample. It also omits commissions, bid-ask spread, slippage, and realistic whole-unit sizing, while using a simplified portfolio model. The author notes that these omissions could materially degrade results and presents the exercise as an initial research example rather than a production-ready backtest.
Key ideas
- The strategy estimates a rolling hedge ratio between SPY and IWM and trades deviations in their price spread.
- Z-score thresholds define spread entries and exits under a mean-reversion assumption.
- The article varies the regression lookback to examine how performance responds to that parameter.
- Full-sample spread normalization introduces lookahead bias and should be replaced with historical rolling estimates.
- Unmodeled trading costs and fractional ETF sizing limit the practical meaning of the reported backtest.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.