Mean reversion is the tendency of a number to drift back toward its own long-run average after straying far from it. Stretch a rubber band and it pulls back — mean-reverting series behave the same way, and the further they stretch, the harder the pull.
Some quantities wander without any home to return to; others have a gravitational centre. A coin-flip random walk has no memory — where it goes next has nothing to do with how far it has already strayed. A mean-reverting series does have memory of exactly that kind: the further it is from its average, the stronger the expected pull back toward it.
Volatility measures are the classic example. When the VIX spikes during a panic, it does not usually stay there — fear is expensive to sustain, and the index historically drifts back toward calmer levels. The same logic applies to the Fear & Greed index and, more arguably, to valuation multiples.
Example (illustrative): suppose a measure has a long-run average of 20 and today it reads 35. A mean-reverting model does not predict "it will be 20 tomorrow." It predicts that the expected NEXT move is downward, and sized in proportion to the 15-point gap — perhaps a pull of a couple of points, with plenty of noise on top that could easily push it higher first. The prediction is about the direction of the drift, not the destination on any given day. If the reading were 22 instead, the pull would be much weaker, because the rubber band is barely stretched.
Several of the model variants we run are mean-reverting by construction. The Ornstein-Uhlenbeck process is the standard mathematical way to write "pulls back toward a long-run mean, with noise," and it underlies our OU-based variants for VIX and Fear & Greed. That is a deliberate modelling CHOICE, not a law of nature: it fits series with a stable centre well, and fits things that trend permanently upward — like a growing company's share price — badly. Which is why an OU model is a reasonable tool for a volatility index and a poor one for a stock on its own.
Mean reversion gives you no timing. "This is stretched and should snap back" can be true and still ruin you, because a stretched market can stretch further for months. It also cannot distinguish a temporary dislocation from a permanent re-rating — that is the structural break problem, and it is genuinely hard. Treat mean reversion as a statement about the balance of probabilities, never as a promise, and always pair it with an invalidation level that tells you when you were simply wrong. None of this is investment advice.
Individual stock prices largely do not — a good business can keep compounding upward for decades with no ceiling to revert to. Volatility measures, sentiment indices, and some valuation ratios show much stronger mean-reverting behaviour, which is why models treat them differently.
Essentially yes — it assumes price will revert toward a prior level. It works when the dip is noise and fails badly when the dip is information, and nothing about the strategy itself tells you which one you are looking at.
They are opposites. Momentum says a move is likely to continue; mean reversion says it is likely to reverse. Both can be true at once over different time horizons, which is part of what makes markets hard.
Browse all S&P 500 tickers to see this metric applied to individual companies.
Educational research only — not investment advice.