Pre-registered proposal: reports/research_proposal_carry_basis_overlay.md. Defaults: look=168h, carry|z|>=1.5, basis|z|>=2.0, multipliers {0.5, 1.0, 1.25}. Window: 2024-06-07 .. 2026-07-31 (785 days, 127 AVAX-gate60 long days) | Hyperliquid AVAX funding cache for CARRY+BASIS; AVAX daily bars from validation.common; _single_coin_decide on AVAX gated by BTC ROC60, validated in reports/universe_generalization.md used to compute the AVAX-gate60 target.
Diagnostics: corr(carry_z, basis_z) = +0.780 | fraction of long days with both signals stretched long (0.5x bucket): 0.4% | fraction with both signals stretched short (1.25x bucket): 0.8%.
| Variant | Days | LongDays | Ret | Sharpe | PF | MaxDD |
|---|
| baseline (gate60, no overlay) | 785 | 127 | -24.00% | -0.05 | 0.98 | -59.48% |
| primary overlay (0.5/1.0/1.25) | 785 | 127 | +12.52% | 0.34 | 1.12 | -46.81% |
| Bucket | Days | Share |
|---|
| 1.0x (no signal) | 102 | 80.3% |
| 1.0x (mixed) | 16 | 12.6% |
| 1.25x (both stretched short) | 6 | 4.7% |
| 0.5x (both stretched long) | 3 | 2.4% |
Bootstrap p(overlay beats baseline, full window): 0.000
Bootstrap p(IS < 2025-01-01): 0.128
Bootstrap p(OOS >= 2025-01-01): 0.000
| Segment | Variant | LongDays | Ret | Sharpe | PF |
|---|
| IS | baseline | 53 | +23.87% | 0.91 | 1.28 |
| IS | primary overlay | 53 | +42.11% | 1.29 | 1.44 |
| OOS | baseline | 74 | -38.65% | -0.64 | 0.77 |
| OOS | primary overlay | 74 | -20.82% | -0.28 | 0.90 |
| look (h) | carry z_enter | basis z_enter | LongDays | Ret | Sharpe | PF | MaxDD |
|---|
| 84 | 1.0 | 1.0 | 127 | +56.01% | 0.70 | 1.28 | -37.11% |
| 84 | 1.0 | 1.5 | 127 | +34.01% | 0.53 | 1.20 | -41.20% |
| 84 | 1.0 | 2.0 | 127 | +20.29% | 0.41 | 1.15 | -45.35% |
| 84 | 1.5 | 1.0 | 127 | +34.18% | 0.53 | 1.20 | -41.44% |
| 84 | 1.5 | 1.5 | 127 | +25.70% | 0.46 | 1.17 | -44.34% |
| 84 | 1.5 | 2.0 | 127 | +18.31% | 0.39 | 1.15 | -45.75% |
| 84 | 2.0 | 1.0 | 127 | +11.46% | 0.33 | 1.12 | -42.98% |
| 84 | 2.0 | 1.5 | 127 | +6.99% | 0.29 | 1.11 | -45.27% |
| 84 | 2.0 | 2.0 | 127 | +4.28% | 0.26 | 1.09 | -46.66% |
| 168 | 1.0 | 1.0 | 127 | +40.77% | 0.58 | 1.23 | -41.64% |
| 168 | 1.0 | 1.5 | 127 | +27.50% | 0.47 | 1.18 | -42.67% |
| 168 | 1.0 | 2.0 | 127 | +12.95% | 0.34 | 1.13 | -46.61% |
| 168 | 1.5 | 1.0 | 127 | +23.29% | 0.44 | 1.16 | -44.60% |
| 168 | 1.5 | 1.5 | 127 | +16.19% | 0.37 | 1.14 | -45.65% |
| 168 | 1.5 | 2.0 | 127 | +12.52% | 0.34 | 1.12 | -46.81% |
| 168 | 2.0 | 1.0 | 127 | +14.62% | 0.36 | 1.13 | -45.82% |
| 168 | 2.0 | 1.5 | 127 | +14.62% | 0.36 | 1.13 | -45.82% |
| 168 | 2.0 | 2.0 | 127 | +12.52% | 0.34 | 1.12 | -46.81% |
| 336 | 1.0 | 1.0 | 127 | +25.47% | 0.46 | 1.17 | -44.12% |
| 336 | 1.0 | 1.5 | 127 | +22.69% | 0.43 | 1.16 | -44.11% |
| 336 | 1.0 | 2.0 | 127 | +10.16% | 0.31 | 1.11 | -47.93% |
| 336 | 1.5 | 1.0 | 127 | +18.75% | 0.39 | 1.15 | -46.45% |
| 336 | 1.5 | 1.5 | 127 | +17.77% | 0.39 | 1.14 | -46.35% |
| 336 | 1.5 | 2.0 | 127 | +10.16% | 0.31 | 1.11 | -47.93% |
| 336 | 2.0 | 1.0 | 127 | +4.21% | 0.26 | 1.09 | -47.25% |
| 336 | 2.0 | 1.5 | 127 | +4.21% | 0.26 | 1.09 | -47.25% |
| 336 | 2.0 | 2.0 | 127 | +0.10% | 0.21 | 1.08 | -48.80% |
- IS baseline Sharpe: 0.91, ret +23.87%, maxDD -33.95%
- IS overlay Sharpe: 1.29, ret +42.11%, maxDD -32.83%
- OOS baseline Sharpe: -0.64, ret -38.65%, maxDD -50.29%
- OOS overlay Sharpe: -0.28, ret -20.82%, maxDD -36.72%
- Pre-registered pass: OOS Sharpe +>= 0.10 AND p < 0.05 AND maxDD not +> 5pp
- Result: PASS
- Inert signal risk: if <5% of long days trigger either 0.5x or 1.25x, the overlay is a no-op and the experiment is a wash. Reported as INERT in the verdict.
- Threshold tuning: the proposal pre-registers look=168h, carry_enter=1.5, basis_enter=2.0. The robustness grid is REPORTED, not selected. Picking the IS winner would be the look-ahead bug class that bit R10-13.
- Hyperliquid only: Binance/Bybit/OKX funding are geo-blocked from this IP (451/403). Hyperliquid started in 2023; the cache is the longest available free source. Binance USDT-M funding IS in the metrics cache but is perpetual-rate only, not the premium index; the BASIS leg specifically needs premium.
- Hyperliquid is one venue: cross-venue funding arbitrage (HL vs Binance) is not what this study tests; this study asks whether HL alone is informationally useful as a sizing overlay on a slow-trend signal.
- CARRY and BASIS are correlated: corr reported above. If corr > 0.7, the composite is one-dimensional and the two-signal framing is misleading; the overlay is then a single-signal gate with two thresholds.
- Multipliers are heuristic: 0.5x and 1.25x are the natural ±50% / ±25% asymmetry, not the result of an optimization. An optimization step would re-introduce the threshold-tuning look-ahead; the honest path is to report the asymmetry you chose and let the verdict be what it is.