About
Scrutinel is an independent statistical forecast of the 2026 United States midterm elections. It estimates the probability of each outcome in every contest on the ballot — the House, the Senate and the governorships — and re-runs itself four times a day.
What it does
Public polls are collected, weighted and averaged; that average is combined with a fundamentals estimate built from presidential approval, the economy and each district’s partisan lean; and the resulting per-race estimates are run through 200,000 Monte Carlo simulations that let related races miss together, because in a real election they do. The full method is set out at the foot of the forecast page.
Where the numbers come from
Every input is public and free: VoteHub for race-level polling and presidential approval, a dedicated generic-ballot aggregate, FEC bulk filings for candidates and fundraising, FRED for economic indicators, US Census ACS for demographics, and presidential results reaggregated onto the 2026 congressional district lines. Pollster ratings and the historical accuracy calibration come from FiveThirtyEight’s public data.
No figure on this site is copied from another forecaster. The polls are third-party data; the averaging, the model and the simulations are our own.
Independence
This site is not affiliated with any campaign, party, candidate, pollster, or news organisation, and takes no view on which outcome is preferable. It is not funded by any political entity. Its running costs are met by advertising, which is served by Google and selected by Google — see Privacy. Nobody buying an ad here has any influence over a number on this page.
Presentation is held to the same standard as the model. Each contest is described from whichever side the figures favour, with the other side’s chance quoted alongside; charts order the two parties by value rather than fixing one above the other; and no colour, heading or turn of phrase is chosen to flatter either party.
What this is not
It is not a prediction of what will happen. A forecast that gives one side a 75% chance is saying the other side wins one time in four — and one-in-four things happen constantly. The published intervals are wide on purpose: they are the honest width of what the data supports this far out, not a hedge.
It is also not betting advice, and the probabilities are not odds anyone is offering you.
Corrections
If you find a wrong candidate name, a misattributed poll, a broken race page or a mistake in the method, please say so — those get fixed quickly, and the source is public so a specific report is easy to act on. Write to hadi.fariborzi@gmail.com.
Contact
Current run 34deabbe21b5, 71 days from the election. Back to the forecast.