The model against the markets
Every race below has a contract on Kalshi, Polymarket, or both. The market number is the price of the Democratic side's contract (the independent's where one stands in for a Democrat), the midpoint of the best bid and ask, so a 70 means traders will buy a dollar paid on a Democratic win for about 70 cents. Prices are not quite probabilities: the two sides of a race usually add to a bit more than a dollar, and thin markets drift. Model chances are from the Nov 4 edition; prices were read Nov 4, 9:36 AM ET. The model does not use market prices anywhere.
Kalshi has no 2026 Senate-control contract, so its number is the summed price of every bin at 51 seats or more on its Democratic seat-count ladder. The model counts 50 seats plus an unaligned independent as Democratic control, which the ladder may not; that is worth about a point.
Kalshi's number is the summed price of every bin at 218 seats or more on its Democratic seat-count ladder; Polymarket's is its party-control contract.
Senate: 2 races priced
One dot per race: across, the model's chance that the Democratic side wins; up, what the market charges for the same outcome. Dots above the line are races the market likes better for Democrats than the model does. Kalshi prices 2 of these races and Polymarket 2.
Governors: 2 races priced
One dot per race: across, the model's chance that the Democratic side wins; up, what the market charges for the same outcome. Dots above the line are races the market likes better for Democrats than the model does. Kalshi prices 1 of these races and Polymarket 2.
House: 68 races priced
One dot per race: across, the model's chance that the Democratic side wins; up, what the market charges for the same outcome. Dots above the line are races the market likes better for Democrats than the model does. Kalshi prices 22 of these races and Polymarket 68. Most House districts have no public polls, so the model's number there is fundamentals alone; the markets have whatever their traders know.
| Race | Candidates | Model | Kalshi | Polymarket |
|---|---|---|---|---|
| NC-6 | Jefferson v. McDowell | 1% | — | 100%+98 |
| TN-4 | Broderick v. DesJarlais | <1% | — | 97%+97 |
| NJ-2 | Mullock v. Drew | 2% | — | 99%+97 |
| NC-13 | Barringer v. Knott | 3% | — | 99%+96 |
| NC-9 | Ojeda v. Hudson | 6% | — | 100%+93 |
| NC-14 | Womack v. Moore | 7% | — | 99%+92 |
| NC-5 | Hubbard v. Foxx | <1% | — | 92%+91 |
| CO-4 | Laubacher v. Boebert | 8% | — | 98%+91 |
| TX-9 | Gutierrez v. Mealer | 10% | 100%+89 | 99%+89 |
| NV-2 | Benitez-Thompson v. Flippo | 13% | — | 99%+86 |
| MI-10 | Hines v. Bouchard | 22% | 99%+77 | 99%+77 |
| MI-4 | McCann v. Huizenga | 23% | 99%+77 | 99%+76 |
| OH-10 | Knickerbocker v. Turner | 27% | — | 99%+72 |
| VA-1 | Taylor v. Wittman | 35% | 99%+65 | 100%+66 |
| FL-9 | Soto v. Green | 34% | — | 99%+65 |
| CO-5 | Killin v. Crank | 33% | — | 96%+62 |
| MN-1 | Johnson v. Finstad | 29% | — | 91%+62 |
| SC-1 | Lacore v. Honeycutt | 36% | — | 98%+62 |
| MT-1 | Forstag v. Flint | 52% | 99%+47 | 99%+47 |
| AZ-6 | Mendoza v. Ciscomani | 53% | 99%+47 | 100%+47 |
| NC-3 | Smith v. Murphy | 7% | — | 51%+44 |
| NY-17 | Conley v. Lawler | 56% | 99%+44 | 99%+43 |
| AZ-2 | Nez v. Crane | 27% | 69%+41 | 68%+40 |
| WA-3 | Perez v. Braun | 62% | 99%+37 | 99%+37 |
| FL-7 | Dalton v. Elijah | 38% | — | 2%−36 |
Chances are the Democratic side's. The small number beside each market price is its gap from the model in points, bold at 10 or more. Market prices link to the market. Showing the 25 largest gaps of 68. .
What a gap means
A market price folds in everything its traders believe, including the polls this model averages, so agreement is not independent confirmation. Disagreement is more informative. Where the market is well above the model for the Democrat, traders are acting on something the model does not see (a late poll, local knowledge, candidate quality, a scandal), or the model's fundamentals are stale for that race, or the market is thin and a few bets moved it. Where the market is below, the reverse.
The House is where to expect the biggest gaps. Most districts have no public polls, so the model projects them from partisanship, incumbency, fundraising, and the national environment, with the wide intervals that deserve. Many of those districts have Polymarket contracts with tens of thousands of dollars traded, which is enough to move a price but not a lot of information. The model's own track record is on the track record page; how it is built is on the methodology page.