Model vs. markets
Where our forecast differs from what betting markets imply — and how to read the gap. A disagreement means one of us is missing something. Our calibration record, below, is how we keep score.
Which party wins most seats
Probability of finishing with the most seats at the next general election.
If the election were held today: our nowcast puts Labour at 88%. That nowcast is also the forecast's trunk: its anchor scenario, Labour holds the lead, which the other five scenarios branch from and which moves with the polls.
Largest disagreement on the page: we rate Labour's most-seats chances 27 points higher than markets do.
Overall majority
Whether any party wins 326 of 650 seats — or nobody does.
If the election were held today: our nowcast puts No Overall Majority at 60%.
Re-cut 16 September 2026 — the scenario set was re-registered as rev2-w2-t0. Figures before and after this date come from different scenario sets, so the model-vs-market history shows a step there; we do not smooth it.
Across 15 resolved contests, our stated probabilities have been better calibrated than market prices — Brier score 0.091 vs 0.148 (lower is better). When we disagree with the market, this record is your reason to take us seriously — and our reason to stay honest.
The market, read as scenario weights
Not yet computed. This panel will show the scenario weights that best reproduce the market prices above — worked out with a stated method, and shown beside ours so you can see which scenario we and the market disagree about.