Mean Reversion Strategies 101
Key Insights
- Pairs trading and z-score entry rules: when prices revert, when they trend, and how to avoid falling knives.
The core idea
If two cointegrated assets deviate from their long-run relationship, bet on convergence: buy the laggard, short the leader, and exit when the spread normalises. The strategy monetises a statistical relationship rather than a directional view — the market can go up, down, or sideways, and the trade still profits if the spread closes. That market-neutrality is what makes pairs trading the archetypal relative-value strategy.
Signal construction
Compute the spread z-score: (spread − mean) / standard deviation over a lookback window. Common entries: z < −2 (long the spread) and z > +2 (short it). Rebalance the z-score window to the half-life of the spread's mean reversion (via a cointegration/OU fit). The half-life is the single most important parameter: a window far shorter than the half-life turns noise into signals, a window far longer makes entries slow and exits later than optimal.
Risk controls
- Trade only cointegrated pairs — correlation without cointegration fails on convergence tests.
- Cap exposure per pair and add a stop on non-convergence.
- Beware regime breaks: the pair may re-mean to a NEW level after structural change.
- Backtest on out-of-sample windows; pairs decay as arbitrage capital enters.
The stop-on-non-convergence rule is non-negotiable: a pair that was cointegrated can de-cointegrate, and the spread trade then becomes an unhedged directional bet wearing a hedge's clothing. Re-test cointegration on a rolling basis and retire pairs whose relationship has structurally broken.
Portfolio construction
Run multiple pairs with uncorrelated spread innovations — single-pair concentration is the biggest killer of mean-reversion books. Size each pair by inverse spread volatility, and monitor the book's net exposure: if all spreads widen together (common in risk-off), the "market-neutral" book suddenly correlates with the market.
References
Article Metadata
Bloom Taxonomy Questions
What z-score thresholds are commonly used for spread entry rules in pairs trading?
Why is cointegration required rather than mere correlation for a pairs-trading signal?
Estimate a pairs-trading rule for two cointegrated instruments, setting the z-score window from the half-life of spread mean reversion.
Further Reading
SEC
US Securities and Exchange Commission — filings, rules, enforcement
MSCI
MSCI research — factor investing, ESG, market analytics
arXiv q-fin
arXiv Quantitative Finance — mathematical finance papers, market models, portfolio theory
Citadel Securities
Semi-annual market structure reports — OTC, options, equity microstructure
BIS
Bank for International Settlements — monetary and financial stability
Bloomberg Insights
Bloomberg Intelligence — market research, sector analysis
Feynman Concept Cards
Master each concept: read the ELI5, explore analogies, work examples, and teach it back.
Mean Reversion Strategies is a concept in trading strategies. In simple terms, Mean Reversion Strategies covers trading strategies for Markets. This markets concept addresses key topics in the trading strategies for markets domain. Also known as: statistical arbitrage, pairs tra
Analogy
Example
Find Gaps
Explain Mean Reversion Strategies as if teaching a colleague who is new to trading strategies. Cover: what it is, how it works, and why it matters.
Create
Create a diagram that demonstrates Mean Reversion Strategies in a real-world trading strategies scenario. Walk through your design decisions.
Show solution
A diagram for Mean Reversion Strategies should include: 1. The core components of mean reversion 2. How they interact 3. Expected outcomes or outputs
Statistical Arbitrage is a concept in strategies. In simple terms, Statistical Arbitrage covers trading and investment strategies for Markets. This markets concept addresses key topics in the trading and investment strategies for markets domain. Also known as: stat a
Analogy
Example
Find Gaps
Explain Statistical Arbitrage as if teaching a colleague who is new to strategies. Cover: what it is, how it works, and why it matters.
Create
Create a calc that demonstrates Statistical Arbitrage in a real-world strategies scenario. Walk through your design decisions.
Show solution
A calc for Statistical Arbitrage should include: 1. The core components of statistical arbitrage 2. How they interact 3. Expected outcomes or outputs