Does Israel-Hezbollah fighting stay a contained border exchange, widen to strikes on Beirut/Dahiyeh/Bekaa or a ground push, or move toward a political/mediated track, before 31 Oct 2026?

Strikes extend to Beirut, Dahiyeh, or the Bekaa Valley, or a ground push is reported.

Band set 2026-09-17 — Strikes so far are confined to southern Lebanon; no capital-area strike reported as of filing. band history ↓

US-brokered talks or a Lebanese Armed Forces deployment/disarmament step is announced.

Band set 2026-09-17 — No talks or LAF deployment step reported as of filing. band history ↓

Strikes continue near the Lebanon-Israel border at the current tempo; no strikes on Beirut/Dahiyeh/Bekaa.

Band set 2026-09-17 — Strikes reported so far are on southern Lebanon border areas (Nabatieh), not the capital or Bekaa. band history ↓

Resolves YES automatically when news.beirut_dahiyeh.count's count_min rule (3-day window, >=5, >=3 sources) is lit on any evaluable day in the window.

Opens 2026-09-18 · Closes 2026-10-31 · baseline signal_replay (0% hist. freq., n_eff=0.0)

Current band: Unlikely (20-45%)

Voided: Voided during integration (2026-09-17): its baseline was frozen with hist_freq=0.0 before its underlying signal_replay series had any real backfilled history. zaviye_scn_forecast_baseline is write-once by design, so the wrong number could not be corrected in place. Reissued with a '-v2' key and a correctly measured baseline. No reader ever saw this forecast — it predates the engine's launch.

Resolves YES automatically when news.beirut_dahiyeh.count's count_min rule (3-day window, >=5, >=3 sources) is lit on any evaluable day in the window.

Opens 2026-09-18 · Closes 2026-10-31 · baseline signal_replay (82% hist. freq., n_eff=95.0)

Current band: Unlikely (20-45%)

Voided: Voided during concept-audit (2026-09-17): the beirut_dahiyeh precision preview (mandatory before trusting a concept, spec sect4.2) found it badly miscalibrated -- ar's base term list mixed a bare event word ('غارة'/raid) into the required-term list with no context_ar, and he had no context requirement at all, so ANY mention of Beirut/Dahiyeh/Bekaa counted as a hit regardless of topic (the 2020 port-explosion investigation, an AUB news item, an antiquities handover, an opera festival, a local crime story, or a bare raid word anywhere in the world such as a Gaza strike). Real sample (test-signal, 30 days, 20 shown): 0/20 were a genuine strike on Beirut, Dahiyeh, or Bekaa. This forecast's already-frozen baseline (hist_freq=0.821, n_eff=95, method=signal_replay) was measured against that broken series. After fixing the concept (moved the event word out of ar's base list into context_ar; added context_he, both AND'd onto the existing place-name base, matching en's already-correct context_en) and re-backfilling news.beirut_dahiyeh.count since 2026-04-22 with the corrected definition, a signal_replay over the identical window/test (since=2026-06-05, test=streak, threshold=5, min_sources=3) now measures hist_freq=0.053, n_eff=95 -- a ~15x drop, not a rounding difference. Reissued with a '-v3' key and this correctly measured baseline.

Resolves YES automatically when news.beirut_dahiyeh.count's count_min rule (3-day window, >=5, >=3 sources) is lit on any evaluable day in the window.

Opens 2026-09-18 · Closes 2026-10-31 · baseline signal_replay (5% hist. freq., n_eff=95.0)

Current band: Unlikely (20-45%)