Mean Reversion (Russell 2000)
Small caps overshoot on short-term moves; fade extreme z-scores over 5–10 days for mean reversion edge.
Thesis (edge)
Russell 2000 exhibits short-term mean reversion after large moves. Fade extreme negative (or positive) 5-day returns when z-score exceeds threshold, with tight risk controls.
Where it works (regimes)
Works in range-bound or moderate-trend regimes. Fails in strong one-way trends (e.g. March 2020). Regime filter (e.g. trend strength) recommended.
Signals
- ( z_5 = (r_5 - \mu_5) / \sigma_5 ): 5-day return z-score.
- Long when ( z_5 < -2 ) (or similar); exit when z reverts or after N days.
Portfolio construction
Size by ATR or recent vol. Max single position 2–3% risk. No leverage.
Risk model
Tail: continued crash (momentum). Stress: 2008, 2020: mean reversion fails. Use hard stop and reduce size in high-vol regimes.
Costs & implementation
Higher turnover than trend. Slippage and spread matter; trade liquid ETFs.
Failure modes
Trending regimes; overfitting z threshold; illiquidity in small caps.
Our Notes & Suggestions
Backtest with regime filter. Consider only fading downside (long only) to reduce short risk. Validate z-score distribution stability.
Our Notes & Suggestions
See the "Our Notes" subsection in the body above for practical guidance, gotchas, and best practices. Always validate regime assumptions and transaction cost assumptions before scaling.
Implementation Checklist
- Compute 5d and 10d return z-scores
- Define entry threshold (e.g. z < -2)
- Position size by inverse volatility
- Stop-loss and max hold (e.g. 10 days)