Strategy
Mean Reversion: Implementation, Fit Ranges, Failure Modes
Mean-reversion strategies bet on temporary price dislocations correcting toward equilibrium. Cousin of trend-following — opposite assumptions, opposite regime fit.
Sentivue Capital··6 min read
Mean reversion strategies bet that prices that have moved sharply away from a reference level will return toward it. They are the structural opposite of trend following — same instruments, same data, opposite assumption set.
Core mechanic
Three flavors:
- Cross-sectional — among a basket of equities, buy recent losers and short recent winners. Original Lehmann / Lo & MacKinlay formulation.
- Time-series — within a single instrument, fade extreme moves relative to a rolling baseline (e.g., z-score thresholds).
- Pairs trading — see pairs trading for the cointegration variant.
Fit ranges
- Holding period: 1–10 days dominates the empirical literature. Below 1 day requires execution infrastructure most retail desks can't sustain. Above 10 days you're fighting the tape.
- Universe: liquid, mean-reverting instruments. Equities work; futures less so. Commodities tend to trend, not revert.
- Threshold: 1.5–2.5 standard deviations. Below 1.5 is noise; above 2.5 is rare and often signals a regime break.
Why it works
- Microstructure. Liquidity demand pushes prices temporarily off their efficient level; market-makers profit from the snap-back.
- Sentiment overshoot. Retail panic / euphoria causes brief overpricing in either direction.
- Index rebalance flow. Forced buying/selling around index reconstitutions creates predictable reversion.
Failure modes
- Trending regimes. A "fade the move" rule gets steamrolled when the move is the start of a multi-month trend.
- Earnings / news. Mean reversion assumes returns to a stable mean. News changes the mean.
- Liquidity crises. Reversion stops working when forced selling overwhelms market-maker capacity.
Implementation discipline
- Position sizing must dominate signal generation. A 70% win-rate strategy with one bad sizing decision can erase a year of small wins.
- Hard kill switches. When realized vol blows past historical ranges, flatten. Mean reversion in crisis = catching falling knives.
- Cost-aware. High-frequency reversion is a slippage game; backtests on mid prices systematically lie.