Forecast vs. result
26th Knesset election, 27.10.2026. Project 120's final forecast (based on the 128 polls officially published on the Central Elections Committee website, the last from 23.10) against the official results.
Blocs
| Bloc | Forecast | 90% range | Actual |
|---|---|---|---|
| Right bloc | 58 | 52-64 | 53 |
| Left bloc | 47 | 39-52 | 48 |
| Arab parties | 15 | 11-20 | 19 |
The forecast gave the right bloc a 24% chance of a 61-seat majority; it won 53. All three blocs landed inside the published ranges, the right bloc at the bottom edge - mainly because of Shas and Arab-sector turnout (details below).
Parties
| List | Poll average | Forecast | 90% range | Actual | % polls | % forecast | % actual |
|---|---|---|---|---|---|---|---|
| Likud | 23 | 26 | 21-31 | 27 | 17.4% | 19.5% | 20.5% |
| Yashar | 22 | 21 | 18-23 | 21 | 16.8% | 15.7% | 16.1% |
| Beyachad | 13 | 12 | 10-14 | 12 | 10.1% | 9.1% | 9.3% |
| Shas | 8 | 10 | 8-12 | 6 | 6.2% | 7.6% | 5.3% |
| Joint List | 9 | 9 | 5-14 | 11 | 6.7% | 6.8% | 8.8% |
| United Torah Judaism | 8 | 8 | 7-9 | 7 | 5.9% | 6.0% | 5.5% |
| Otzma Yehudit | 8 | 8 | 6-10 | 7 | 6.3% | 6.0% | 5.7% |
| Yisrael Beitenu | 8 | 7 | 6-9 | 7 | 5.9% | 5.5% | 5.9% |
| The Democrats | 9 | 7 | 0-14 | 8 | 7.1% | 5.5% | 6.4% |
| Ra'am | 5 | 6 | 5-7 | 8 | 4.4% | 5.0% | 5.9% |
| Religious Zionism | 7 | 6 | 5-8 | 6 | 5.5% | 5.1% | 5.1% |
| Reservists & Economic | 0 | 0 | 0-5 | 0 | 3.0% | 2.9% | 2.7% |
| Amcha Yisrael | 0 | 0 | 0-4 | 0 | 2.8% | 2.8% | 1.7% |
"Forecast": the 120-seat map shown on the site. Green: within one seat. Red: outside the published range. "Poll average": per-pollster average of the 20 polls published in the final week (18-23.10, 9 pollsters), converted to seats with the same Bader-Ofer method and surplus agreements. Blue and White received no votes; Amcha Yisrael and the Reservists & Economic list missed the threshold, as forecast.
What worked
- Likud. Most pollsters measured Likud below the result, some by 4-6 points. The model's adjustment - halfway between the historically accurate pollsters and the rest, plus the polls' historical Likud miss - raised it from 23 seats in the poll average to 26. Actual: 27.
- Most other lists - Yashar, Beyachad, Yisrael Beitenu, Otzma Yehudit, Religious Zionism and United Torah Judaism - landed within one seat of the forecast.
- Overall the forecast missed by 12 seats in total, against 18 for the simple poll average; in vote share, 0.82 points per list on average against 0.98.
What missed
- Shas - over-estimated because of the previous elections. In all five previous elections (2019-2022) the polls under-measured Shas, so the model added about 1.6 points. This time the polls over-estimated Shas (6.2% vs. 5.3% actual), and the correction doubled the error: 10 seats forecast, 6 actual. This is the largest miss, and alone accounts for most of the right-bloc gap.
- Record, unexpected turnout in the Arab sector. Turnout in Arab localities reached about 80%, against about 53% in 2022 - a historic high, and above turnout in Jewish localities. No poll anticipated it: the Joint List and Ra'am won 14.7% together, against 11.2% in the poll average, and 19 seats instead of 15. Ra'am won 8, above its published range (5-7).
- The Democrats. Here too the historical correction (polls used to over-estimate the left) lowered them too much: 7 forecast, 8 actual, while the poll average gave 9.
What we learned
- Apply a historical correction only when its direction is stable. When a party's polling error flips sign between elections, widen its uncertainty instead of shifting the forecast.
- Shared uncertainty for Arab-sector turnout. Sector turnout moves the Joint List and Ra'am together and should enter the model as one factor, with wider ranges.
- Small lists. A list that withdraws should leave the model automatically, and lists polling below the threshold tend to lose votes at the last minute (Amcha Yisrael: 1.7% actual vs. 2.8% in the polls).
Sources: official results on the Central Elections Committee website for the 26th Knesset (votes26.bechirot.gov.il, as of 30.10.2026; 5,690,682 valid votes, 12,926 polling stations). Seats computed by Bader-Ofer with a 3.25% threshold and the signed surplus agreements. Forecast: the last run before the election, 27.10.2026. Arab-locality turnout computed from polling-station results in localities where the Joint List and Ra'am together won over 70% of the vote. All data and code are open in the project repository.