PCA Residual Statistical Arbitrage
Rolling latent-factor removal and OU residual mean reversion
Mode FORWARD SIMULATIONCapital EUR 100,000Broker orders DISABLED
Five-year walk-forward backtest
Current mathematical signal
| Symbol | Target weight | Residual z-score | OU half-life |
|---|
Model specification
Past-only estimationReturns are standardized over 252 sessions. PCA removes components explaining approximately 55% of common variance. Each cumulative idiosyncratic residual is fitted as an AR(1) approximation to an Ornstein–Uhlenbeck process. Only residuals with a 2–20 session half-life and |z| ≥ 2 are eligible.