The road to 218.
Explore the history. Inspect the polls. Test the path to a House majority.
Where the House could change
Green: too close to call. Cook Political Report rates these races as toss-ups: neither party has a clear advantage. This is an expert race rating, not a tied poll or an exact 50–50 probability.
Orange: possible party change. Seats in our discussed gain/loss scenarios. Orange takes priority when a seat is also a toss-up.
The green list is our October 9 snapshot of 21 Cook toss-ups. The separate 270toWin consensus layer has 15 toss-ups. Click any seat for polls and details.Seats in this view
| District | History 2018 → 2024 | Our distinction | Pollsters | 270toWin | Polymarket D |
|---|
What do the polls show?
Latest poll per firm and seat. Lead margins are percentage points, not seats or win probabilities.
Pollster comparison
Wedgewood is classified Republican at your request. Affiliation is a working label, not measured bias.
| Seat | Pollster / sponsor | Association | Fieldwork ends | Reported lead | Source |
|---|
Build the scenario, seat by seat.
Start from our discussion baseline, choose gains and losses, and hold all other seats unchanged.
Democratic seat total
Potential Democratic gains Choose the seats Democrats win
Potential Democratic losses Choose the seats Democrats lose
Polls, race ratings and market prices.
A national vote margin, a seat rating and a probability answer different questions.
Average published margin
141 national polls from July–October. Unweighted, with repeated sample variants possible. Excluding left-associated firms: D+6.89.
15 toss-ups
Source map labelled Oct 2, retrieved Oct 18. Different from our Oct 9 Cook snapshot of 21 toss-ups. All ratings are available in the atlas.
Open the reference map ↗The generic ballot over time
Mean published D lead, percentage points · unequal poll mix
Among six right-associated firms with July and September readings, their latest monthly mean rises from D+5.33 to D+8.00. The two comparable left-associated firms move from D+9 to D+7. Changing poll composition limits month-to-month comparisons.
National polling by association
Left / right describe organizational or sponsor associations, not proven polling error. RMG is provisionally classified right-associated. Unknown firms remain unverified.
Democrats lead Republicans by nine percentage points—for example, 52% versus 43%. It does not mean nine seats or a 9% chance of winning.
Explore all 141 national poll entries
| Pollster | Dates | Association | Margin | Sponsor |
|---|
Sources, definitions & limitations
Election history
2018, 2020, 2022 and 2024 general-election winners from MIT Election Lab / Harvard Dataverse, retrieved via a versioned public mirror. Fusion ballot lines are combined by candidate; exhausted ballots are excluded. NC-09 in 2018 was not certified and is marked missing.
Dark blue/red: same party won all four elections. Light blue/pink: same party won both 2022 and 2024, but not all four. Yellow: party alternated at each of the three transitions—there are no such districts in this numbered-seat comparison. Mixed histories not covered by these rules are grey.
Boundaries & ratings
2026 district geometry and consensus ratings come from 270toWin, retrieved Oct 18 UTC. The consensus map is labelled Oct 2. Boundaries can change, and the history is not a reconstruction of past votes within current lines.
Green in our overlay means the 21 seats in the Cook Oct 9 snapshot discussed in this chat. All 21 were in the original 111-seat dataset; 16 have recent polls. Orange means the 16 potential D gains and three potential D losses we selected.
District polling
2026 general-election polls only, excluding national, DC and unspecified districts. Filter by fieldwork end date July 1–October 31; the latest available is October 14. 111 original districts, 96 recent, 163 polls, 66 pollster IDs, 37 seats with multiple pollsters.
Use each firm's latest end date, named matchups where available, then likely-voter samples where available. Same-date scenarios can produce a range. These are candidate leads, not predicted winners. Affiliation labels are working classifications from our discussion; sponsors are shown separately. Wedgewood’s Republican label is explicitly a user assumption.
Markets & scenarios
Polymarket.com prices are a saved public Gamma API snapshot, retrieved Oct 18 UTC—not a live feed and not the separate Polymarket US ask feed shown by 270toWin. A price of 0.905 implies about 90.5% probability. Independent outcome prices can sum above or below 100%. They are not vote shares.
District markets appear only where a party-specific price was found. “No price” is not a 50% probability. Scenarios use a starting D count of 214 and hold other seats fixed; they neither model all 435 races nor allocate vacant seats.
National polling
Data credit: The New York Times, national polling data. This view includes 141 July–October entries. Means are unweighted and entries may include multiple variants of one sample. Published margins are retained even where rounded vote shares differ.
Snapshot, not a forecast
The dashboard combines dated evidence from different sources. Polling disagreement is separate from a forecaster's toss-up rating. A national D lead cannot be translated directly into House seats. Source links let you check subsequent updates.
Download dashboard data ↓