Mean Reversion

Mean reversion is one of the most important concepts in quantitative trading. The idea is intuitive: when a price, return, or spread deviates significantly from its long-term average, it tends to move back toward that average over time. Extreme values are temporary; normal values are the equilibrium.

This concept appears everywhere in finance and nature. A stock that has fallen 40% in a week — absent fundamental deterioration — is likely oversold and may bounce back. The implied volatility of options tends to spike during panics and then revert to more typical levels. Interest rate spreads widen during crises and narrow during calm periods.

Mean reversion stands in contrast to momentum (the tendency of rising prices to keep rising) and the random walk hypothesis (the idea that price changes are unpredictable). In practice, financial markets exhibit both mean reversion and momentum, depending on the time horizon and the specific instrument. Understanding when each force dominates is a key skill for quant traders.

The standard mathematical model for mean reversion is the Ornstein-Uhlenbeck (OU) process:

dX = θ(μ - X) dt + σ dW

Where:

• X = the value of the process (e.g., a spread between two stocks)
• θ = the speed of mean reversion (higher θ = faster reversion)
• μ = the long-term mean
• σ = the volatility of the process
• dW = a Brownian motion increment (random noise)

The key feature is the drift term θ(μ - X): when X is above μ, the drift is negative (pulling X down toward μ); when X is below μ, the drift is positive (pushing X up). The parameter θ controls how quickly the process reverts.

The half-life of mean reversion — how long it takes for a deviation to decay by half — is:

t_1/2 = ln(2) / θ

A half-life of 5 days means a $1 deviation from the mean will, on average, shrink to $0.50 in 5 days. Shorter half-lives are more attractive for trading because the edge is captured faster.

Before trading a mean-reversion strategy, you must statistically test whether the series actually mean-reverts. The primary tools are:

• Augmented Dickey-Fuller (ADF) test: Tests the null hypothesis that a time series has a unit root (is non-stationary / not mean-reverting). A low p-value (<0.05) provides evidence of stationarity (mean reversion).
• Hurst exponent: A value between 0 and 1. H < 0.5 indicates mean reversion, H = 0.5 indicates a random walk, and H > 0.5 indicates trending behavior (momentum). Calculated using rescaled range analysis or DFA.
• Variance ratio test: Compares the variance of multi-period returns to single-period returns. For a random walk, variance scales linearly with time. If variance grows slower than linearly, the series is mean-reverting.

In practice, raw stock prices are almost never mean-reverting — they tend to follow random walks. However, spreads between related stocks, valuation ratios (like P/E ratios), and volatility levels often do mean-revert, making them suitable for mean-reversion strategies.

Mean reversion is the foundation of several major quant strategy types:

• Pairs trading: The classic mean-reversion strategy. When the spread between two correlated stocks widens abnormally, go long the underperformer and short the outperformer, betting on convergence.
• Statistical arbitrage: A more sophisticated, portfolio-level extension of mean reversion. Factor-based models identify stocks that are "cheap" or "expensive" relative to their factor-predicted values and trade them toward convergence.
• Volatility mean reversion: Implied volatility is one of the most strongly mean-reverting quantities in finance. When the VIX spikes to 40 during a panic, selling volatility (shorting options) is a mean-reversion bet that volatility will decline toward its long-term average (~18).
• Fixed income: Interest rate spreads (e.g., the yield spread between corporate bonds and Treasuries) tend to mean-revert, especially after crisis-driven widening.

The flip side of mean reversion is momentum — the tendency of recent winners to keep winning. Research shows that mean reversion tends to dominate at very short horizons (intraday) and very long horizons (3-5 years), while momentum dominates at intermediate horizons (3-12 months). Sophisticated quant strategies often combine both signals.