2026 midterm results · Updated
Our final forecast
The forecast as it stood on
Who Shows Up forecasts every House, Senate and governor race by asking whose voters will actually turn out. It reads the electorate's enthusiasm from more than 100 special elections, blends it with polls and fundamentals, and simulates the election 20,000 times.
Your races
Enter your ZIP code to see the House, Senate and governor races on your ballot, our final odds in each, and where a single vote is likeliest to decide control. It uses the new district maps where states redrew them for 2026.
How many seats each party wins
Each bar is how often a seat count came up across 20,000 simulated elections. Blue bars are outcomes where Democrats win control; red where Republicans do.
The races most likely to decide the Senate
Ranked by tipping point: how often, across all simulations, each race is the seat that hands a party control of the Senate. How this works.
The national picture
How the forecast has moved
What makes this forecast different
- Turnout first. Special-election overperformance is one of the two biggest inputs to the national picture. Every simulation also randomizes how much of the national swing comes from turnout versus voters changing sides.
- Independent. The model uses no pundit ratings and no other forecasts. Prediction-market prices from PredictIt appear on the Compare page for comparison only and have no effect on its numbers.
- Checked against past elections. We reran the full model on 2018, 2020, 2022 and 2024, using only what was known in late September of each year. It picked the winner in
of races, and the results landed inside its 80% ranges of the time. Run on 2010 through 2016 with fewer ingredients (no campaign money, candidate records or special elections), it still picked of winners. Each new ingredient had to make those past forecasts more accurate to be included; ideas that didn't were left out (see what we tried). - Honest about polls. September generic-ballot polls have overstated Democrats in 14 of the last 15 elections, campaign-sponsored polls lean toward their sponsor by about four points, and fresh polls count more than stale ones. The model corrects for all three using measured numbers.