Mean Reversion
Mean reversion is the idea that when a variable (like an asset price, return, spread, or implied probability) moves unusually far away from its typical level, it has a tendency to drift back toward its average over time.
In markets, the “mean” is usually defined using a historical window (for example, a moving average). “Reversion” describes the pull back toward that baseline after a large deviation. Mean reversion is not a guarantee—strong trends, regime changes, or new information can keep prices away from prior averages for long periods.
Why It Matters
Mean reversion is a core concept in risk management, statistical trading, and market interpretation. It helps analysts:
- frame “overextended” moves versus moves supported by durable information
- design rules for entries/exits (for example, when deviations become extreme)
- stress-test strategies that assume prices (or probabilities) won’t drift indefinitely
For prediction markets, mean reversion can be useful when studying how odds react after an information shock—some moves overshoot and later correct, while others reflect genuine updates.
What causes mean reversion?
Mean reversion can come from several mechanisms:
- Liquidity and order flow effects: short-lived imbalances can push prices away from fair value and then fade.
- Behavioral effects: overreaction, anchoring, and herding can create temporary extremes that later correct.
- Constraints and arbitrage: when mispricings are detectable and tradable, participants may push prices back toward a reference level.
However, if the underlying fundamentals change (a “new normal”), the historical mean itself may shift.
How is mean reversion different from trend following?
- Mean reversion assumes deviations are temporary: prices/probabilities that move far from a baseline are more likely to move back toward it.
- Trend following assumes persistence: moves tend to continue in the same direction for some time.
Which behavior dominates depends on the market, time horizon, and regime (for example, calm vs. high-volatility periods). It’s common to see mean reversion at short horizons and trend-like behavior at longer horizons—or vice versa.
How do analysts test for mean reversion?
Common approaches include:
- Deviations from a moving average: measure distance from the mean using percent moves or a z-score.
- Stationarity tests: check whether a time series behaves like it fluctuates around a stable level (for example, using unit-root style tests).
- Half-life estimates: quantify how quickly a deviation tends to decay.
- Backtesting: simulate rules (entry, exit, sizing, costs) to see if observed reversion survives realistic frictions.
A key risk is “data-mining”: a strategy may look mean-reverting in one sample but fail out-of-sample.
What does mean reversion look like in prediction markets?
In prediction markets, the variable of interest is often an implied probability (or contract price). Mean reversion might show up as:
- a sharp spike in implied probability after a rumor, followed by a partial pullback when the rumor is refuted
- temporary dislocations between related markets that converge as information spreads
If you analyze prediction-market odds via the FinFeedAPI – Prediction Market API, you can study mean reversion by comparing real-time odds to rolling baselines and monitoring overreaction/correction patterns.
Real-World Example
A market’s implied probability jumps from 45% to 65% within an hour after breaking news. Over the next day, follow-up reporting reduces uncertainty and traders reassess. The probability drifts back to ~55%—a partial mean reversion—suggesting the initial move may have overshot the longer-run consensus.