Mean reversion is the belief that prices, spreads, or indicators that stretch far from a "normal" level tend to move back toward that average. Traders buy weakness and sell strength when they think the move is temporary noise, not a permanent shift.
It is the philosophical opposite of momentum investing. Both can work in different environments; many drawdowns come from applying the wrong one for the regime.
Common signal ideas
- Distance from moving average - price far below a 50- or 200-day average.
- RSI or similar oscillators - oversold or overbought readings.
- Z-score of returns - unusually large down day relative to recent volatility.
- Pairs spreads - two related assets diverge; bet on convergence.
Each signal needs a defined entry, exit, and holding period. "Oversold" without an exit rule is not a strategy.
Where reversion shows up
- Short horizons on liquid ETFs or large caps after sharp one-day shocks (fragile edge, cost-sensitive).
- Pairs trading in related equities (requires careful corporate event risk).
- Volatility mean reversion in options contexts (advanced, not passive ETF buy-and-hold).
Where reversion fails
- Strong trends - what looks overbought keeps rising (momentum dominates).
- Structural breaks - bankruptcy, fraud, index deletion; the mean itself moves.
- Leveraged products - path dependency destroys simple reversion intuition.
Win rate vs payoff
Mean reversion systems often show higher win rates with smaller average wins until a trend day produces a large loss. Risk control and position limits matter more than headline accuracy.
Testing mean reversion rules
- State the universe and liquidity constraints.
- Define signals without peeking at future data (backtesting basics).
- Include costs on turnover-heavy rules.
- Stress start dates and walk-forward slices.
Summary
Mean reversion is a bet on snapback. It rewards discipline in ranges and punishes stubbornness in trends. Combine it with regime awareness, drawdown limits, and honest cost assumptions rather than a single magic threshold.
