Skip to content

Mean Reversion

OVERVIEW

Mean reversion is the tendency of a measurable quantity — price, volatility, spread, or valuation ratio — to return to its long-run average after deviating significantly from it. Tapeboard surfaces mean-reversion signals across equities, spreads, and implied volatility in real time.

Note

Mean reversion is the mathematical opposite of momentum. Most quantitative strategies combine both: momentum in trending regimes, mean reversion in ranging ones.

Extremes are treated as temporary, not self-reinforcing. A mean-reversion position bets on a quantity returning toward its historical norm after breaching a defined threshold.


HOW MEAN REVERSION IS QUANTIFIED

Tapeboard uses four core measures to detect and rank mean-reversion setups:

Measure Formula / Definition Interpretation
Ornstein-Uhlenbeck Process dX_t = θ(μ − X_t)dt + σdW_t Continuous-time model; θ = speed, μ = long-run mean
Z-Score (Current Value − Rolling Mean) / Rolling Std Dev Practical signal threshold; ±2 or ±3 triggers entry
Half-Life −ln(2) / ln(β₁) where β₁ = lag-1 autocorrelation Time for a deviation to revert 50% toward the mean
Hurst Exponent H < 0.5 = mean-reverting; H = 0.5 = random walk; H > 0.5 = trending Regime classification for a given instrument

Tip

Use the Hurst Exponent column in the scanner to pre-filter instruments before applying z-score thresholds. Applying mean-reversion logic to a trending instrument (H > 0.5) produces false signals.


WORKED EXAMPLE

SPY has a 20-day rolling mean of $520 and a 20-day standard deviation of $8. During a sharp intraday selloff, SPY drops to $496.

Z-score = ($496 − $520) / $8 = −3.0

A z-score of −3.0 places SPY three standard deviations below its rolling mean. Historically, SPY closes at or below −3 sigma fewer than 0.3% of trading days.

Parameter Value
Entry $496 (−3.0σ)
Target $512 (−1.0σ)
Stop $492 (−3.5σ)

A pairs trader observing QQQ at −1.1 sigma would go long SPY / short QQQ, targeting normalization of the spread between them.


FEATURES

Signal Types

Signal Timeframe Trigger
Gap Fade Intraday Gap-open beyond ±2σ from prior close or VWAP
Bollinger Band Extreme Swing Price touches outer band with elevated RSI reading
Pairs Spread Statistical arb Cointegrated pair spread reaches z-score threshold
IV Mean Reversion Options IV Rank or IV Percentile at historical extremes

Scanner Columns

Column Description
Z-Score (20D) Distance from 20-day rolling mean in standard deviations
Half-Life (days) Estimated days to 50% reversion based on AR(1) regression
Hurst Exponent Regime indicator; values below 0.5 confirm mean-reverting behavior
IV Rank How far current implied volatility sits relative to its 52-week range

HOW TO USE

  1. Open the Scanner and apply the Mean Reversion preset filter.
  2. Sort by Z-Score (20D) descending to surface the most extreme deviations.
  3. Confirm H < 0.5 in the Hurst Exponent column before acting on a signal.
  4. Check Half-Life to size holding period — longer half-lives require wider stops and more capital patience.
  5. Set entry at the current z-score level, target at −1σ or the rolling mean, and stop beyond the next sigma band.
  6. For pairs trades, navigate to Pairs → select two cointegrated instruments → monitor the live spread z-score chart.

Note

Always define maximum acceptable loss before entering a mean-reversion position. The strategy carries asymmetric loss exposure — price can move substantially further against you before reversing.


LIMITATIONS AND MISCONCEPTIONS

Risk Detail
Mean definition matters A 20-day mean and a 200-day mean produce opposing signals in a trending market
Permanent mean shifts Fundamentally impaired companies or regime changes can render any static mean misleading
Lookback sensitivity Hurst Exponent and Half-Life values change materially with the lookback period selected
Path risk Positions can extend well beyond entry before reverting — size accordingly
Drifting mean A stock in secular decline makes a static rolling mean an unreliable anchor

Tip

Not all deviations revert. Screen for structural breaks and recent fundamental catalysts before treating a large z-score as a signal rather than a warning.