Window: 2025-07-27 .. 2026-07-26 (365 days) IS split at 2026-01-25 (IS=182d, OOS=183d). Costs: 10bps/side, both legs. Adoptable per window: net Sharpe>0 AND bootstrap p<0.05 AND trades>=10. Top-5 by IS score; OOS verdict on the same params (no re-tuning).
Sweep: 756 runs, of which 1 pass IS. Top-5 IS picks passing OOS: 0/5. Verdict: REJECT: 0/5 top IS picks pass OOS — IS-fit was a window artifact.
Top 5 IS picks — IS vs OOS
| Pair | params | IS sharpe | IS p | IS trades | IS pass? | OOS sharpe | OOS p | OOS trades | OOS pass? |
|---|---|---|---|---|---|---|---|---|---|
| BTC/DOGE | z=30, ze=1.5, h=5 | +2.31 | 0.044 | 14 | Y | -0.18 | 0.565 | 16 | N |
| XRP/LINK | z=30, ze=1.5, h=5 | +1.92 | 0.077 | 18 | N | +0.36 | 0.387 | 15 | N |
| ETH/DOGE | z=60, ze=1.5, h=5 | +1.80 | 0.091 | 11 | N | -2.57 | 0.965 | 14 | N |
| SOL/ADA | z=30, ze=1.5, h=10 | +1.65 | 0.119 | 11 | N | 0 | N | ||
| ETH/XRP | z=30, ze=1.5, h=5 | +1.42 | 0.163 | 15 | N | -0.63 | 0.651 | 16 | N |
Reading
- The IS window picks the best params. If the SAME params also pass the OOS window with the same threshold, the signal is structural; if not, it's a fit on the IS slice.
- This is a single split, so even a 5/5 match would still need a rolling walk-forward before production. A 0/5 match is a hard REJECT for the universe.
Time: 4s.