The UK's general election, 15 August 2029: how far the polls can still move by polling day, and what our forecast would look like after a typical miss.
Seats and votes: today against polling day
999 days from the latest poll (20 Nov 2026)| Party | Vote today (GB) | Vote on polling day | Seats if the election were today | Seats on polling day | 80% range |
|---|---|---|---|---|---|
Labour |
27.5 | 29.0 +1.5 | 309 239–376 | 324 | 188–443 |
Conservative |
22.8 | 22.5 −0.3 | 151 88–216 | 130 | 32–261 |
Reform UK |
21.3 | 20.4 −0.9 | 48 28–79 | 32 | 4–173 |
Lib Dem |
10.0 | 10.5 +0.5 | 69 61–75 | 69 | 44–85 |
Green |
9.5 | 9.1 −0.4 | 4 4–6 | 4 | 3–9 |
Restore Britain |
3.0 | 2.9 −0.1 | – – | 0 | 0–0 |
SNP |
2.7 | 2.5 −0.2 | 29 20–38 | 17 | 4–41 |
Plaid Cymru |
1.0 | 1.0 −0.1 | 6 5–8 | 5 | 3–9 |
Seats are the middle of 5,000 simulated elections. "Today" uses the same polls and lens with no allowance for the time left. Vote shares are GB, built from England, Scotland and Wales. 326 seats is a majority.
Chances
| Today | Polling day | |
|---|---|---|
| Labour largest party | 93% | 80% |
| Conservative largest party | 7% | 14% |
| Reform UK largest party | 0% | 6% |
| Labour majority | 36% | 49% |
| Conservative majority | 1% | 3% |
| Reform UK majority | 0% | 1% |
| Hung parliament | 64% | 47% |
By nation
| England | Scotland | Wales | |
|---|---|---|---|
Labour |
29.7 +1.1 | 24.5 +3.9 | 24.2 +4.5 |
Conservative |
24.0 −0.3 | 11.3 −0.3 | 15.0 |
Reform UK |
20.7 −0.8 | 15.8 −1.0 | 23.7 −2.0 |
Lib Dem |
11.1 +0.6 | 8.1 +0.1 | 5.1 +0.1 |
Green |
9.4 −0.4 | 6.8 −0.4 | 8.0 −0.7 |
Restore Britain |
3.2 −0.1 | – | 1.8 −0.2 |
SNP |
– | 29.1 −1.9 | – |
Plaid Cymru |
– | – | 20.3 −1.7 |
Polling-day shares, with the change from today.
Why polls move, and by how much
We took every GB voting-intention poll since 1983 (Mark Pack's PollBase) and, for each general election from 2001 to 2024 (7 elections), averaged the polls at distances from a month to three years before polling day. The same figures for 1987, 1992, 1997 are in the data but not used by the model: it learns from no election before 2000. Then we asked what best predicted the result: the polls as they stood, the party's result at the previous election, or something in between.
The Conservatives, Labour and the Lib Dems
They finished much closer to their previous result than their polls months out suggested. The best predictor, a few months or more before polling day, takes only about 62% of the change in their polls since the last election at face value. Within the last couple of months the polls count for more and more.
Predicting each election from the others, it beat the polls as they stood in 3 to 5 elections (out of 5–7) at every distance from one to three years, and cut the average miss from 4.8 to 4.5 points.
UKIP, the Brexit Party, Reform and the Greens
No such pull back: pulling them towards their previous result made things worse (think of Reform's rise in 2024 or UKIP's in 2015). So the challengers, and the new parties, are carried forward as polled.
Everything is then rescaled to 100%, so points the big three regain come from everyone else in proportion.
How much of the big three's change since the last election to take at face value, by days before polling day. The line is the fitted curve (0.62 + 0.38 × e−days/30); the dots are the best single weight at each distance.
How it did, election by election
Average miss per party (Con, Lab, LD), GB, each election predicted from the others| Before polling day | Elections | Polls as they stood | Previous result | Our method | Beat the polls | Miss (1 SD) |
|---|---|---|---|---|---|---|
| 1 month | 7 | 2.7 | 4.6 | 2.6 | 4 of 7 | 3.3 |
| 2 months | 7 | 4.2 | 4.6 | 3.7 | 4 of 7 | 4.7 |
| 3 months | 7 | 4.5 | 4.6 | 4.0 | 4 of 7 | 5.0 |
| 6 months | 7 | 4.7 | 4.6 | 4.0 | 4 of 7 | 5.6 |
| 9 months | 7 | 3.6 | 4.6 | 3.3 | 3 of 7 | 4.5 |
| 1 year | 7 | 4.1 | 4.6 | 3.7 | 5 of 7 | 4.8 |
| 1.2 years | 7 | 4.3 | 4.6 | 4.2 | 3 of 7 | 5.0 |
| 1.5 years | 7 | 4.7 | 4.6 | 4.4 | 4 of 7 | 5.2 |
| 1.7 years | 7 | 4.9 | 4.6 | 4.3 | 4 of 7 | 5.2 |
| 2 years | 7 | 4.8 | 4.6 | 4.5 | 4 of 7 | 5.6 |
| 2.5 years | 6 | 5.8 | 4.5 | 5.1 | 5 of 6 | 6.0 |
| 3 years | 5 | 5.4 | 4.5 | 5.0 | 3 of 5 | 6.6 |
The last column is how far a party's polling-day share can still move: it sets the extra spread in the simulation (Reform UK ±6.4, Green ±6.4, Conservative ±5.8, Labour ±5.8 points, one standard deviation, for the date chosen), on top of the usual polling error.
What this can't tell you
- It is an average of the past, and elections differ. In 2024 the Conservatives won back none of the ground they had lost in the polls a year out, while Labour's lead shrank a long way; in 2017 Labour gained ten points in the campaign. A new leader, a campaign or a crisis can move things either way, which is why the ranges are so wide.
- The challengers are carried forward as polled. Reform, the Greens and Restore Britain have no history of fading or growing that we can measure reliably; eight cases are too few.
- It is measured on GB polls. We apply the same pull back towards each nation's own 2024 result in England, Scotland and Wales. The SNP and Plaid Cymru are carried forward as polled.
- We don't know the date. The latest the election can be held is 15 August 2029 (Parliament first met on 9 July 2024 and dissolves after five years). Pick a date above to see how much it matters.
Poll data: PollBase by Mark Pack, markpack.org.uk/opinion-polls. GB results 1983–2005 from PollBase, 2010–2024 from the UK Parliament's constituency results.
Figures as of . Published by Polling Forecast.












