Mamad
The chance of a wide siren wave in Israel over the next hour, day, week and month, computed from three years of Home Front Command alerts and the last 24 hours of the wire.
Chance of a wide siren wave in the next 24 hours
Why this number
- ▼No wide waves in the last week
- ▼No siren waves of any size in the last 24 hours
- ▼No wide waves in the last 30 days
- ▲Iran: concrete threat or preparations reported
- ▼No wide waves in the last 3 days
- ▼No Home Front Command pre-alerts in the last 24 hours
Base rate
0 wide waves in the last 30 days · last one 152 days ago (458 localities)
No skill score is available until a wide wave occurs within the scoring window.
A wide wave is a run of Home Front Command rocket, missile or hostile-aircraft alerts with no ten-minute gap that covers at least 25 localities, any part of Tel Aviv, Jerusalem, Haifa or Beersheba, or the whole country.
Experimental estimate. Accuracy is not established. Rewritten every ten minutes. Not a warning system: when a siren sounds, follow the Home Front Command.
How the number is made
The event. A wide wave is a run of Home Front Command rocket, missile or hostile-aircraft alerts with no ten-minute gap that covers at least 25 localities, any part of Tel Aviv, Jerusalem, Haifa or Beersheba, or the whole country. The ground truth is the Home Front Command’s public alert history, with stored records dating to 7 October 2023. Missing or delayed source data can affect the estimate.
The history. For every hour in that record we build what was knowable at the time (hours since the last wide wave, waves in the last day, three days, week and month, pre-alerts, the time of day) and whether a wide wave followed within the hour, the day, the week and the month. Four regressions, one per horizon, are fit on those rows and refit nightly.
The news. Once an hour, a closed-question model reads the last 24 hours of the wire’s security coverage and the Home Front Command’s own posts and places Iran, Lebanon, Yemen and Gaza on a five-step scale from quiet to under way, and the official posture from unchanged to shelter. These readings enter the same regressions. Their weights started from a stated prior and move as the wire’s history accumulates. These assumed weights and incomplete news coverage limit the estimate; the displayed percentages are experimental, and their accuracy is not established.
Honesty. Every forecast is stored and, once its horizon has passed, scored against what happened. The score above is that record. Forecast windows overlap, so the count is not a count of independent tests. A model must beat the base rate on historical holdouts to serve, but that does not establish calibration or accuracy today. This page is a forecast of sirens, not a warning: when one sounds, the Home Front Command’s instructions are the only thing that matters.
Read Israeli news in the baba News app.