Distance-Based Pair Selection and Mean-Reversion Trading
Summary
The distance approach forms pairs by rescaling each asset’s training-period prices to a common range, calculating the sum of squared differences between each pair’s normalized series, and selecting the closest matches. In the cited original study, the formation window is 12 months and the procedure selects 20 pairs. During a separate trading period, the strategy uses the training data’s scaling values and spread standard deviations. When a pair’s spread moves beyond two standard deviations, it takes opposing long and short positions; it closes them when the spread returns through zero or the trading period ends.
The method does not test for cointegration, so historically similar price paths may reflect spurious dependence. The document cites research reporting that up to 32% of pairs chosen this way fail to converge. It describes possible selection refinements: restricting candidates to the same industry, preferring formation-period spreads with more zero crossings, or selecting pairs with higher spread variability. These alternatives are discussed as research-inspired criteria, not as guaranteed improvements, and the page provides no new performance results.
Key ideas
- The approach selects pairs by ranking squared distances between range-normalized training price series.
- Trading signals use the training-period spread standard deviation, with entries beyond two standard deviations and exits at a zero crossing or period end.
- Pair formation does not test for cointegration, leaving the method exposed to spurious relationships and divergence risk.
- Potential selection refinements include industry matching, formation-period zero crossings, and spread variability.
- Signals are defined as long, short, or flat positions for each pair.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.