2026 House forecast · v0.25-r4
Democrats are favored to win House control.
Site release updated November 3, 2026 · House estimates and the frozen House simulation remain on the November 3, 2026 model · polls through November 1, 2026
Experimental model. The COYNTOSS House model is experimental. It has not completed a historical backtest, its win probabilities are not yet historically calibrated, and it is being actively developed, so its methods, ratings and probabilities may change from one release to the next. Treat these numbers as a work in progress, not a finished forecast.
- Median D seats
- 238
- 80% range
- 212–268
- Toss-ups
- 33
218 for a majority
Democratic seats
of 435 districts
Every district by forecast rating
Geometry and forecast data loading…
THE HOUSE, SEAT BY SEAT
All 435 districts, from most Democratic to most Republican
Loading the seat chart…
THE CLOSE RACES
Where polls and fundamentals pull the model
The closest districts that have district polls, within five points. The hollow circles are the two published ingredients; the solid dot is the published model margin.
Loading districts…
THE BATTLEGROUND
The toss-up districts
Ordered by how close the projected margin is. Bars show each side's exported chance of winning.
SEATS THAT COULD CHANGE HANDS
Where the projected leader is from the other party
Districts held by an incumbent whose party trails in the model margin. Open seats are not included.
RACES TO WATCH
Closest districts
Every district from D+15 through R+15, ordered by absolute projected margin.
| District | Candidates | Margin | Rating | Open |
|---|
EVERY STATE
Open a state's House forecast
Each bar is one state's districts from most Democratic to most Republican, colored by rating.
PUBLISHED FROZEN OUTLOOK
Seats and uncertainty
- Average Democratic seats
- 239.3
- Median Democratic seats
- 238
- Central 80% of simulations
- 212–268
Average seats are the mean across simulations; the median is the middle outcome. The central 80% range is the 10th to 90th percentile, with outcomes outside it still possible.
Model inputs November 3, 2026 · simulation run November 3, 2026 (UTC) · 10,000 draws · run V0.25-R4-20261103-S20261103 · polls through 2026-11-01 · PROVISIONAL / VALIDATED FUNDAMENTALS (EXP-H001).
CONDITIONAL SCENARIOS · PROVISIONAL
What if the national environment changes?
These exported diagnostic reruns change the national generic-ballot assumption while holding the other model inputs fixed. They are conditional scenarios, not a new forecast or a replacement for the published frozen outlook above. Even the baseline rerun can differ from the official snapshot.
Current inputs · national margin D+8.3
- Diagnostic average D seats
- 239.2
- Diagnostic D control chance
- 83.7%
- Diagnostic central 80%
- 212–269
Exact seat frequencies
All scenarios are listed below. Interactive selection becomes available when the matching scenario data loads.
| Scenario | National margin | Average seats | Control chance | Central 80% |
|---|---|---|---|---|
| Generic ballot 4 points more Republican | D+4.3 | 225.4 | 63.4% | 201–252 |
| Generic ballot 2 points more Republican | D+6.3 | 232.1 | 74.6% | 206–260 |
| Current inputs | D+8.3 | 239.2 | 83.7% | 212–269 |
| Generic ballot 2 points more Democratic | D+10.3 | 246.8 | 90.3% | 218–278 |
| Generic ballot 4 points more Democratic | D+12.3 | 255.0 | 94.7% | 224–288 |
Scenario assumptions and source
PROVISIONAL / UNCALIBRATED · 50,000 draws · seed 20261103 · source run V0.25-R4-20261103-S20261103 · source release 2026-11-03-01.
- Each generic-ballot shift changes an unconstrained district center by shift × (1 − its existing polling blend); verified same-party ballots retain their fixed party control. +/-100 encodes ballot eligibility, not a candidate vote-margin estimate.
- Polling, district leans, candidates, uncertainty and the shared-error settings stay fixed.
- All scenarios use the same random draws, so their differences reflect the specified shift.
- No coefficient was chosen to match VoteHub or another forecast.
- District centers/blends: updated-site/data/house-forecast.json.
- District total and residual sigmas: data/house-source-export/probability-inputs.json.
- Shared-error settings and generic ballot: data/house-diagnostic-settings.json, copied from the identified workbook run.
- Seeded diagnostics use a separate random stream from the official frozen simulation; small numerical differences are expected.
- Conditional scenarios, not a new official election forecast or a calibrated prediction.
- The House model and its uncertainty settings have not completed a historical backtest.
- The official frozen snapshot remains separate from these diagnostic reruns.
Methodology
How the House forecast works
Each district combines its fundamentals (presidential results on the current map, the national environment, incumbency and past House results) with district polling where it exists; a seeded simulation turns those estimates into seat and control chances. This is a beta model, and its win probabilities are not yet historically calibrated.