Statistical Arbitrage with Cointegrated Spreads and Mean Reversion
Summary
These notes introduce statistical arbitrage through pairs of assets whose price spread is believed to be stable over time. The example tests for cointegration, estimates the spread’s mean and standard deviation, and sets entry thresholds around that mean. When the spread moves beyond a threshold, the trader buys the relatively cheap asset and shorts the expensive one, then closes the positions as the spread returns toward its usual range. The notes also describe market-neutral, cross-market, cross-asset, and ETF arbitrage, and mention extensions using lead-lag effects, momentum, flows, volatility models, principal components, and machine learning.
The discussion emphasizes that apparent convergence is uncertain: relationships can change, spreads can keep widening, and strategies can suffer extended periods of poor results. Historical patterns do not guarantee future behavior. It also highlights trading costs as essential to realistic backtests and argues that the economic logic behind a spread should be studied alongside its statistical properties. The text offers conceptual examples and historical context, but no independently documented test results for the described rules.
Key ideas
- Cointegration can help identify asset pairs whose price spread may revert toward a stable range.
- A basic pairs trade buys the relatively cheap asset and shorts the expensive one when the spread crosses a threshold.
- Statistical arbitrage can also compare related assets across markets or between an asset and its underlying basket.
- Trading costs should be included when estimating strategy performance in backtests.
- Spread relationships can break down, so mean reversion and historical performance are not assured.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.