Where mean reversion is most reliable
Mean reversion is most robust in three settings. Statistical relationships between related assets (pairs trading, spread trading): if two highly correlated assets diverge in price for technical rather than fundamental reasons, the spread tends to revert. Volatility: implied volatility (the VIX) and realised volatility are strongly mean-reverting — high volatility periods reliably decay toward historical averages, making variance risk premia tradeable. Credit spreads: in the absence of fundamental deterioration, credit spreads exhibit mean reversion as technical supply/demand imbalances (forced selling, index rebalancing) create temporary dislocations. Equity index-level P/E ratios also show long-run mean reversion — but on 5–20 year horizons, far too slow for practical trading.
The Ornstein-Uhlenbeck process
Mean-reverting processes are formally modelled by the Ornstein-Uhlenbeck (OU) equation: dXₜ = θ(μ − Xₜ)dt + σdWₜ, where θ is the speed of mean reversion, μ is the long-run mean, and σ is volatility. The OU process is the continuous-time equivalent of an AR(1) time series with positive mean reversion coefficient. Estimating θ (the half-life of the reversion — how long it takes for half of a deviation from the mean to dissipate) is the key to determining whether a mean-reverting spread is tradeable: a half-life of days to weeks is actionable, a half-life of years is theoretically interesting but practically useless for active trading.
The random walk alternative and regime shifts
The most dangerous trap in mean reversion trading is confusing a genuine stationary process with a random walk that happens to look mean-reverting over a limited sample. Statistical tests (Augmented Dickey-Fuller, KPSS) can detect stationarity, but these tests have limited power on short samples. More insidiously, a spread that was stationary for years can become non-stationary permanently — a "regime shift" — when the fundamental relationship between the two assets changes. The classic failure: pairs traders who shorted Volkswagen against Porsche in 2008, believing the spread would revert, were devastated when Porsche’s surprise announcement of a short squeeze created a non-mean-reverting discontinuity. Mean reversion strategies require constant monitoring of whether the structural relationship underpinning the trade still holds.
“The risk in mean reversion trading is not that you are wrong about the long-run mean. It is that you run out of capital before the market runs out of irrationality.”
What this means for you
Mean reversion trading at the professional level is dominated by quant hedge funds using statistical arbitrage strategies — highly diversified portfolios of mean-reverting pairs that generate consistent small returns per trade. Volatility mean reversion is exploited through systematic volatility selling (selling options above fair value when VIX spikes). For non-professional investors, the most actionable mean reversion insight is valuation-based: cyclically-adjusted P/E ratios (CAPE/Shiller P/E) exhibit very long-run mean reversion, suggesting that extremely high equity valuations eventually mean-revert through either lower returns or actual price declines. But "eventually" can mean a decade — mean reversion at the macro level requires patience measured in years, not trades measured in days.