LOADING
Mean reversion is the theory that asset prices and returns tend to move back toward their historical average over time. When a stock's price deviates significantly from a measure of fair value, whether based on fundamentals, historical ranges, or statistical models, mean reversion strategies bet on a return to normalcy. This contrasts with momentum strategies, which bet that price trends will continue. The two approaches are complementary because they operate on different time horizons: momentum tends to work over 3-12 months, while mean reversion operates over very short-term (days) or very long-term (3-5 years) horizons.
Statistical arbitrage (stat arb) is the most sophisticated form of mean reversion trading. Stat arb strategies identify pairs or baskets of related securities whose price relationship has temporarily diverged from its historical norm. The classic pairs trade buys the underperforming stock and shorts the outperforming stock within a correlated pair, profiting when the spread converges. Modern stat arb extends this concept to large portfolios of hundreds of securities using principal component analysis and factor models to identify and trade mispricings.
Bollinger Bands provide a visual framework for mean reversion. They consist of a moving average plus and minus two standard deviations. When prices touch the upper band, the stock is considered overbought; when they touch the lower band, it is oversold. However, Bollinger Band signals alone are unreliable because trending markets can "walk the bands" (prices remaining at the upper or lower band for extended periods). Effective mean reversion strategies typically combine price deviation signals with volume analysis, fundamental anchors, or regime filters.
The primary risk in mean reversion strategies is that the deviation from the mean is not temporary but structural. A stock trading two standard deviations below its historical P/E may not be cheap; it may have experienced a permanent deterioration in its business. This risk is why mean reversion strategies require careful fundamental screening to exclude stocks where the "reversion" target is no longer relevant. Stop-losses are also essential because the most extreme deviations often continue further before reverting.
Cointegration provides a more robust statistical foundation for mean reversion than simple correlation. Two cointegrated time series may diverge in the short term but are bound together in the long term by an economic relationship. For example, the stock prices of two oil companies with similar reserves and cost structures are likely cointegrated even if they temporarily diverge due to company-specific news. The Engle-Granger and Johansen tests are standard methods for testing cointegration. A cointegrated pair will have a stationary spread, meaning the spread reliably reverts to its long-run mean, providing a sound basis for trading.