Testing Crypto Pair Spreads for Stationarity and Mean Reversion
Summary
This article explains how to assess whether a cryptocurrency pair spread is suitable for mean-reversion trading. It presents the Augmented Dickey-Fuller (ADF) test as a check for a unit root, with a low p-value used to reject non-stationarity, and the Hurst exponent as an additional indicator of persistence or mean-reverting behavior. It recommends looking for evidence from both tests before treating a spread as anchored to a stable average.
The article also discusses logging prices to reduce scale differences, constraining implausible hedge ratios, and using rolling z-scores to identify large spread deviations as possible entry signals. It offers illustrative thresholds but no empirical results or formal validation of the combined process. The discussion simplifies statistical interpretation: passing a stationarity test or observing a low Hurst estimate does not guarantee future mean reversion or profitable trades. The article notes that transaction costs, hedge ratios, and stop rules require further strategy development.
Key ideas
- The ADF test evaluates the null hypothesis that a spread is non-stationary.
- A Hurst exponent below 0.5 is presented as evidence consistent with mean-reverting behavior.
- Log prices can help reduce scale differences when estimating relationships between assets.
- Hedge ratio limits may help avoid impractical or spurious pair constructions.
- Rolling z-scores provide a systematic way to flag spread deviations, but do not establish profitability.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.