Where the numbers come from
We run no polls of our own. Every seat number on this site is an average of other people’s, built in three steps.
- We collect every poll that gets published.Israeli pollsters, and the TV channels and papers that commission them.
- We keep the 10 most recent.Ordered by when the pollster was in the field, not when it aired.
- We average each party across those 10.Pollsters differ from each other in a steady way, so first we work out how far each one usually sits from the rest and take that out. Then add up a party's seats across the 10 polls, divide by 10, round to the nearest seat. Under 4 seats a party shows as not clearing the threshold.
What polling gets wrong
Not every pollster asks about every list. When one leaves a party out, our average counts that poll as zero seats for them, so small and new lists read lower here than in the polls that did measure them.
Each one converts vote shares into seats using its own assumptions about the threshold and vote-sharing deals. Averaging their seats inherits all of that.
By more than the seat or two that usually separates a coalition from 61.
The profiles
Profiles are written here from the public record, not from party material, and each one links its sources. The issue scale reads stated positions, not voting records. A candidate who differs from their party is marked personal position.
Placements come from what people have said, not from what we think of them, and the sources on each page let you check that. Quote and link anything here with attribution. If you are on deadline or think something is wrong, get in touch. Press queries are answered first.
The fine print
Polls come from Midgam, Kantar, Direct Polls, Filber and others, as published by N12, Kan 11, i24 News and Channel 14. New ones are pulled daily; the latest fieldwork in the current average is 10 September 2026. Each pollster’s standing lean is estimated over the last 30 polls and taken out before averaging, so the number moves when opinion does and not when a house that reads high happens to publish twice in a week. A house is trusted in proportion to how often it has published. Beyond that the mean is unweighted: no recency weighting inside the 10, no adjustment for sample size or track record.