Knotenlinie H5-3: blind-window test 2016–2020, Italy and the Aegean — preregistration and result

Results version of preregistration H5-3 (version 1, DOI 10.5281/zenodo.23615034, published 2026-10-26 before the blind window was opened). The blind window 2016-01-01 to 2021-01-01 (exclusive), Italy and the Aegean, was opened once on 2026-10-26 and scored with the frozen code (DONDUR.json 382a337843e1786ad382ec8e03e42e50b0618eeb5dad5c3931dcb4d0b2d14e8f). The preregistration files are unchanged from version 1.

Composite verdict: success, carried by the Aegean row alone, which the preregistration labels "not like-for-like". Aegean daily ETAS against the static ESHM13-SEIFA map, M 4.95–8.6: +2.57 ± 0.48 nat/event, z 5.41, N 91 (a daily model against a static map; it does not mean an established daily model was beaten). Italy like-for-like headline row, R-b1-n0.8-mu2 against our reconstruction of ETAS-LM, M ≥ 2.95: −0.010 ± 0.029, N 2254, failure (a tie; this row's power was 1.2%, stated in advance). Physics family (outside the composite verdict): Coulomb rate-and-state stress transfer +0.001 ± 0.004 over the whole window (+0.017 ± 0.004 outside the Amatrice–Norcia sequence, diagnostic only); GSRM 2.1 strain mixture −0.017 ± 0.010. The long-term rows have 2 declustered events and carry no information. Completeness gate passed (yearly Mc 1.3–1.7).

Files added in this version: SONUC.md (result report, Turkish, the record) and SONUC.en.md (English translation); H5-3_sonuc.json (full result, sha256 6958fc83b74e212c285e5356079cd47050d78f0931eb6ed1b0d8ff72c741e1b4) and KISA.json (short result); DONDUR.json, HAZIRLIK.json and the frozen forecast and input files listed in PARMAK-IZLERI.txt, whose SHA-256 match that list; the OpenTimestamps proofs (*.ots) of the fingerprint list and both preregistration texts. Numbers are the output of the frozen code; they were not independently re-computed. Data from 2021 onward stays sealed.