What forecasters and markets say about the House
Models, expert ratings and prediction markets rarely agree. Here is each one, who is behind it, and how fresh it is.
Polymarket: “Which party will win the House in 2026?” Volume $17,589,109. Snapshot 2026-10-15.
Polymarket: “Which party will win the Senate in 2026?” Volume $6.6M, snapshot 2026-10-15. See all 35 Senate races
Odds as a chart and tablePolymarket
| Outcome | Market price | Source | Snapshot |
|---|---|---|---|
| Democrats control the House | 90.5% | Polymarket | 2026-10-15 |
| Republicans control the House | 9.5% | Polymarket | 2026-10-15 |
Estimate Not a forecast model: the price reflects what traders are willing to pay.
All tracked forecasters8 sources
| Source | Type | What it publishes | Last update we know of | Affiliations |
|---|---|---|---|---|
| Decision Desk HQ | Statistical model | A statistical forecast model plus election-night race calls. | 2026-10-12 | ⚑ known ties |
| Split Ticket | Statistical model | A data-driven forecast model. It is now published through The Argument. | not recorded | ⚑ known ties |
| Cook Political Report | Expert rating | Expert race-by-race ratings (Solid/Likely/Lean/Toss-up). | 2026-10-09 | none on file |
| Sabato's Crystal Ball | Expert rating | Expert race-by-race ratings from the University of Virginia Center for Politics. | 2026-10-08 | none on file |
| Inside Elections | Expert rating | Expert race-by-race ratings (Solid/Likely/Lean/Tilt/Toss-up). | 2026-10-01 | none on file |
| 270toWin consensus | Aggregator | A consensus map built from seven forecasters' ratings. | 2026-10-02 | none on file |
| Polymarket | Prediction market | A prediction market: traders buy and sell contracts on outcomes, and the price is read as a probability. | 2026-10-14 | none on file |
| The Hill | Media outlet | A Washington political news outlet. Its election data and race calls come from Decision Desk HQ. | 2026-10-12 | ⚑ known ties |
A flag means we found a documented relationship between the source and a media company, party or business, with a link to the document. “None on file” means we have not found one, not that none exists. National model probabilities live on each source’s own site, so we link to them with the date we last saw an update. District ratings are on each race page, compiled from Wikipedia.
Where markets and forecasters disagree4 races
Polymarket lists a market for 153 of the 435 races (combined volume $6.7M). In 4 of them the market and the forecaster median point different ways.
| District | Forecaster median | Market D / R | Volume | What differs |
|---|---|---|---|---|
| TX-15 | Toss-up | 80 / 18 | $63K | The market favors Democrats (81%) while the forecaster median calls it Toss-up. |
| NC-11 | Toss-up | 76 / 24 | $25K | The market favors Democrats (76%) while the forecaster median calls it Toss-up. |
| PA-10 | Tilt D | 80 / 20 | $12K | The market favors Democrats (80%) while the forecaster median calls it Tilt D. |
| TX-34 | Tilt D | 84 / 16 | $12K | The market favors Democrats (84%) while the forecaster median calls it Tilt D. |
Estimate Market prices are crowd-implied probabilities and many of these markets are thin; forecaster ratings are a different kind of estimate. Neither is a result.
How each forecaster rates the 435 racescounts by rating
| Forecaster | Safe D | Likely D | Lean D | Tilt D | Toss-up | Tilt R | Lean R | Likely R | Safe R | Rated | Dates listed |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Cook Political Report | 178 | 16 | 12 | — | 21 | — | 9 | 17 | 171 | 424 | 2025-02-06 to 2026-10-09 |
| Inside Elections | 178 | 9 | 5 | 10 | 16 | 8 | 8 | 16 | 174 | 424 | 2025-02-05 to 2026-10-05 |
| Sabato's Crystal Ball | 175 | 15 | 17 | — | 20 | — | 11 | 19 | 167 | 424 | 2025-02-06 to 2026-10-08 |
| Decision Desk HQ | 180 | 27 | 9 | — | 10 | — | 11 | 70 | 117 | 424 | 2026-07-14 to 2026-10-13 |
| The Economist | 63 | 6 | 4 | — | 4 | — | 6 | 13 | 26 | 122 | 2026-04-24 to 2026-09-24 |
| Consensus (median) | 179 | 11 | 17 | 3 | 15 | 2 | 8 | 23 | 166 | 424 | — |
Estimate Each forecaster's own ratings as compiled on Wikipedia (CC BY-SA 4.0); dates are those listed there. Counts include only races the forecaster rated.
How to read thesemarkets vs models vs ratings
Markets
Prices set by traders. They update every minute but can be moved by a few large bettors.
Models
Statistics built on polls, fundamentals and history. Each makes different assumptions, so they disagree.
Expert ratings
Analysts sort races into Solid / Likely / Lean / Toss-up by judgment, not by formula.