| Variant | Segment | n | Sharpe | Ret (cum) | MaxDD | p vs base | CorrMkt | CorrEqMom | Resid |
|---|
| GLD buy-hold (baseline) | IS | 2799 | 0.47 | +128.62% | -45.56% | nan | 0.04 | 0.05 | 0.33 |
| GLD buy-hold (baseline) | OOS | 2678 | 0.90 | +316.76% | -26.40% | nan | 0.07 | 0.15 | 0.82 |
| GLD 126d-TSMOM long/cash | IS | 2799 | 0.63 | +152.64% | -26.32% | nan | 0.02 | 0.05 | 0.47 |
| GLD 126d-TSMOM long/cash | OOS | 2678 | 0.76 | +190.26% | -25.98% | 0.505 | 0.04 | 0.13 | 0.72 |
| EqMom baseline (12-1 top decile, hold21, net10bps) | IS | 2363 | 0.62 | +195.59% | -54.71% | nan | 0.88 | 1.00 | -0.06 |
| EqMom baseline (12-1 top decile, hold21, net10bps) | OOS | 2667 | 1.06 | +1004.91% | -38.55% | nan | 0.77 | 1.00 | 0.44 |
| EqMom x fast-credit gate (20d widening > 1std -> flat) | IS | 2363 | 0.93 | +301.59% | -32.95% | nan | 0.66 | 0.78 | 0.59 |
| EqMom x fast-credit gate (20d widening > 1std -> flat) | OOS | 2667 | 0.95 | +525.58% | -31.61% | 0.497 | 0.52 | 0.85 | 0.49 |
| EqMom x inversion gate (T10Y2Y < 0 -> flat) | IS | 2363 | 0.62 | +195.59% | -54.71% | nan | 0.88 | 1.00 | -0.06 |
| EqMom x inversion gate (T10Y2Y < 0 -> flat) | OOS | 2667 | 0.82 | +425.16% | -38.55% | 0.488 | 0.71 | 0.92 | 0.14 |
| Blend (OOS) | Sharpe | MaxDD |
|---|
| EqMom+GLD_bh 50/50 | 1.29 | -23.45% |
| EqMom+GLD_trend 50/50 | 1.23 | -23.80% |
- GLD 126d-TSMOM long/cash: REJECT (uplift -0.15, p=0.505, resid 0.72, corrEqMom 0.13)
- EqMom x fast-credit gate (20d widening > 1std -> flat): REJECT (uplift -0.11, p=0.497, resid 0.49, corrEqMom 0.85)
- EqMom x inversion gate (T10Y2Y < 0 -> flat): REJECT (uplift -0.24, p=0.488, resid 0.14, corrEqMom 0.92)
Uplift +0.23 Sharpe, maxDD -23.5% vs -38.6%, but paired p=0.51. Not significant. GLD buy-hold standalone (OOS Sharpe 0.90, corrEqMom 0.15, resid 0.82) has diversifier properties; blend stays MARGINAL candidate, no sleeve. GLD trend rule adds nothing over buy-hold (1.23 vs 1.29).