For each race, the model mixes two kinds of information: what political experts think, and what traders in a prediction market are willing to pay. It is simple on purpose, so you can follow every step.
Step 1. Start with the expert rating
The Cook Political Report rates each race from Solid to Toss Up. The model turns each rating into a chance that Democrats win:
Step 2. Blend in the market
Kalshi, a prediction market, sets a price on many races. Where a race has one, the model gives of its weight to the market price and the rest to the rating-based chance. Races without a market use the rating alone. The Market D column shows the market's price. You can change both settings above to see how much each one matters.
Step 3. Add up the seats
Expected seats adds up every race's chance. Projected seats counts the races where Democrats are above 50%. Control takes . In the Senate, 34 Democratic-aligned seats are not up this year, and Nebraska's independent Dan Osborn is counted as Democratic-aligned.
How often is it updated?
Market prices: about every 3 hours.
Ratings: an automated check runs each morning on the closest races. It changes a rating only when two different news sources confirm it and the move is one step. Anything less certain is held back for review.
The note at the top of the page shows when each part was last updated.
What this is not
It is not election results. Results are unofficial until certified.
It is a simple blend, not an independent prediction. It adds no polling or candidate analysis of its own beyond what the ratings and markets already reflect.
Races are linked in real life. If one party does better than expected, it tends to happen in many races at once, so the real outcome can land well away from the expected seat totals.
Markets and raters can both be wrong, and a few races have no market price.