IS/OOS 2016-01-01, sqrt(365), 501 equity names. Independence = corr to EqMom + corr to market + market-neutral residual Sharpe. A real NEW edge = positive OOS Sharpe, p<0.05, AND (low corr to EqMom OR residual Sharpe>0).
| Factor | OOS Sharpe | p | Corr EqMom | Corr Mkt | Resid Sharpe |
|---|---|---|---|---|---|
| ResidualMom (top10%) | 1.50 | 0.000 | 0.98 | 0.79 | 0.88 |
| BAB long-only low-beta | 0.92 | 0.008 | 0.43 | 0.62 | 0.20 |
| BAB long/short | -0.90 | 0.993 | -0.57 | -0.64 | -0.17 |
| TOM overlay (market) | 0.61 | 0.045 | 0.35 | 0.43 | 0.10 |
Reading
- Residual momentum strips market beta before ranking: if its residual Sharpe>0 it is NOT the same factor as EqMom (which is raw momentum).
- BAB (low-beta long / high-beta short) is a documented real anomaly; by construction it has LOW corr to the market. The long-only slice is the deployable, no-short diversifier.
- TOM is a calendar timing overlay ON the market (corr~1 to market) -> not a new factor, only a seasonal tilt; useful as an overlay, not a sleeve.
- Cost note: top-decile monthly rebalance at 10bps/side is already ~2.4% drag/yr; BAB long/short is far costlier (2x turnover) and needs shorts.
Caveats
- Annualization uses sqrt(365) (house crypto convention). For these EQUITY daily returns the finance-standard is sqrt(252); the reported Sharpes are therefore ~20% OVERSTATED vs a 252-day basis. Cross-round ranking is unaffected (all Rounds use sqrt(365)), but do not compare directly to quant-site/backtester figures that use sqrt(252) without rescaling.