Strict forward window 2026-07-31..2026-08-31 (not used in the original study). Protocol frozen: look=168h, carry|z|>=1.5, basis|z|>=2.0, multipliers {0.5,1.0,1.25}. Per-coin verdict mirrors validation/carry_basis_overlay_forward_oos.py.
| Coin | orig OOS | fwd days | long | base SR | overlay SR | base ret | overlay ret | n_active | p | forward verdict |
|---|---|---|---|---|---|---|---|---|---|---|
| ETH | PASS | 31 | 12 | 3.03 | 3.20 | +9.51% | +11.58% | 1 | 0.351 | CONFIRMED |
| BTC | INERT | 31 | 12 | 4.49 | 4.31 | +13.33% | +12.88% | 1 | 1.000 | AMBIGUOUS |
| SOL | PASS | 31 | 12 | 4.92 | 5.02 | +20.61% | +22.57% | 1 | 0.351 | CONFIRMED |
| XRP | PASS | 31 | 10 | -1.96 | -1.96 | -5.16% | -5.16% | 0 | 1.000 | INERT |
| LINK | PASS | 31 | 14 | 3.55 | 3.69 | +18.63% | +21.86% | 1 | 0.351 | CONFIRMED |
| AVAX | PASS | 31 | 0 | 0.00 | 0.00 | +0.00% | +0.00% | 0 | 1.000 | INERT |
| DOGE | PASS | 31 | 5 | -4.91 | -4.91 | -10.50% | -10.50% | 0 | 1.000 | INERT |
| ADA | PASS | 31 | 3 | -4.14 | -4.14 | -3.36% | -3.36% | 0 | 1.000 | INERT |
Aggregate
CONFIRMED 3 | AMBIGUOUS 1 | MARGINAL 0 | INERT 4 | REJECT 0 (of 8 coins with data).
Sample-size warning: every CONFIRMED coin here fired on 1 long day(s) in the 31-day window (n_active column). A one-day 'confirmation' is not evidence -- treat the aggregate as INERT-to-thin, not as a pass. The overlay stays gate-off until a longer forward record accumulates.
Reading
- The original study PASSed 7/8 coins (BTC inert) on the mined OOS. This forward window is the strict test: a coin CONFIRMS only if the overlay Sharpe >= baseline on data the study never saw.
- 31 days is small per coin; the multi-coin tally is the point -- consistent CONFIRM across coins with more than a handful of active days would be real evidence; a tally dominated by n_active<=1 is not.
- Caveat carried from the single-coin study: the overlay fires on few long days (INERT if <1%), and a forward verdict is one data point, not a promotion gate by itself.