Statistical Arbitrage Through Mean Reversion and Pairs Trading
Summary
The document presents statistical arbitrage as a family of systematic strategies that seek to trade relative mispricing, often using mean reversion in historically related instruments. It describes pairs trading as one approach: identify assets whose prices or spread have moved together, take a long position in the relatively weaker asset and a short position in the stronger one, then look for convergence. Broader variants include market-neutral portfolios, cross-market and cross-asset arbitrage, and ETF arbitrage.
For a pairs workflow, it recommends selecting candidates, examining closing prices and the spread, calculating a spread z-score, testing stationarity with the Augmented Dickey-Fuller test, and generating signals only if the relationship appears stationary. An example reports a test statistic below the stated five-percent critical value, but this alone does not establish future profitability. The article flags changing relationships, external shocks, execution and liquidity risk, turnover, slippage, and transaction costs; historical correlation or cointegration can fail, and costs must be included in backtests.
Key ideas
- Statistical arbitrage uses quantitative methods to trade relative price deviations among related assets.
- Pairs trading typically buys the relatively weaker asset and shorts the stronger asset while anticipating convergence.
- A proposed workflow checks the spread and z-score, then tests spread stationarity with the Augmented Dickey-Fuller test.
- Stationarity evidence in historical data does not guarantee that a relationship will persist or produce profits.
- High turnover, transaction costs, slippage, liquidity constraints, and structural changes can undermine results.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.