You decide: the House
Call races, or move turnout, the national mood and the polls' error, and the forecast's own simulated elections say what happens to the House.
The marks are past elections: hover or tap one for what happened that year, click it to set the slider there.
Each dot is one district. Click a close district's dot to call it (on a phone, use the list below, which has a search box). Paler colours are closer races.
| Race | Candidates | Forecast (D) | Yours (D) | Call it |
|---|
How this works
The House forecast is 20,000 simulated elections, run with the same model and the same national and regional connections between races as the published forecast. With nothing changed this page counts those simulations and gives the published numbers. Calling a race keeps only the simulations in which it went that way, so the other races move with it as far as the model says they are linked. The sliders recompute every simulated election's margins before counting.
The sliders are arithmetic on top of the forecast, not part of the model. The forecast predicts margins, and turnout, the national mood and polling error are already inside its uncertainty. A slider asks what happens if one of them came out a given way on top of everything else, in every simulated election. Turnout scales Democratic voters relative to Republican voters; only that ratio moves a two-way margin, so there is one slider, not one per party. National environment moves every race's margin by the same amount, polled or not. Polling error moves only races that have polls, on top of the environment, so setting both to the same side moves the polled races twice.
The marks on the sliders are the record, not limits. Each is a past election, midterms labelled and presidential years as the lighter ticks; hover or tap one for what happened that year. Turnout marks are the change in the Democratic-to-Republican makeup of voters from the previous election of the same kind, by the network exit polls (turnout by party is not published, because ballots are secret). Environment marks are how far the national House vote landed from the final generic-ballot polls. Polling-error marks are the average error of the final House district polls, by cycle, from this site's own poll record. The ranges run past the largest shift on record. The figures and their sources are in data/you_decide/history.json.
A majority is 218 of 435 seats. Districts whose winner no slider setting can change are counted as they fall in every simulation; the list shows the 417 whose winner can change. Alaska's independent is counted as not Democratic, as on the House forecast.
Simulations from the forecast of 2026-10-22 (run a4b74f53).