Ranking Cointegrated Stock Baskets by Weight Stability and Mean-Reversion Half-Life
Summary
This installment extends a scoring framework for stock baskets used in statistical arbitrage. In addition to liquidity and cointegration strength, it adds stability of the Johansen cointegrating weights and the spread's mean-reversion half-life. The half-life is estimated by fitting a discrete autoregressive relationship to changes in a stationary spread; a nonnegative coefficient is treated as non-reverting. Its value, measured in bars, helps screen baskets whose expected holding times exceed the trader's intended horizon.
The article explains how long half-lives can tie up capital and increase exposure to costs and market events, while emphasizing that mean reversion is not assured. It illustrates the range of observed estimates, including an extremely long one, and reports that a subsequent backtest suggests basket selection improves when both added criteria are included. However, the text does not provide detailed comparative performance figures, and it says the half-life should be evaluated out of sample. Thresholds must be adapted to the strategy timeframe and risk tolerance.
Key ideas
- The basket scoring framework adds weight stability and mean-reversion half-life to liquidity and cointegration strength.
- Half-life estimates how many bars a stationary spread takes to close half of its deviation from its mean.
- Non-reverting estimates and holding periods beyond the chosen risk horizon can be screened out.
- Long half-lives can increase capital lockup, carrying costs, and exposure to market events.
- Mean reversion is not guaranteed, and the half-life estimate should be checked out of sample.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.