Call
the Map
FORECAST55 DAYS OUT
← Back to The Wire

analysis · General Election

Welcome to Call the Map

Call the Map is a forecast of the 2026 U.S. midterms: all 435 House districts, 35 Senate races and 36 governorships, from one model, updated every day.

Call the Map is a forecast of the 2026 U.S. midterms: all 435 House districts, 35 Senate races and 36 governorships, from one model, updated every day.

It is free to read. No advertising, no paywall, no account, and nothing sold.

What it is, honestly

This started as a hobby project and is still mostly one. It is built and run by one person. I hope it can become something bigger, and whether it does depends almost entirely on whether people find it useful enough to support.

I am not going to tell you this model is better than anyone else's, or close to it. I do not know that, and the people who have been doing this for years have earned their reputations. What I can tell you is that I gave it everything I had, that I checked the things I could check, and that where I could not check something I have said so on the page rather than in a footnote.

What it does

Every number here is computed rather than curated. One model produces all 506 races and both chambers together, so the map, the scoreboard and the race pages cannot disagree with each other — and when a race moves, the site says whether the race moved or the model did.

The method is published in full — every equation, every fitted coefficient, and the validation record it was tested against before it published a single 2026 number. It is all on the methodology page. If you want to argue with a number, the working is there to argue with.

Mosaic of joint control outcomes from 100,000 simulations of the published edition: Democrats take both chambers 38.8 percent, Republicans take both 22.2 percent, and the two split-control outcomes together account for 39.0 percent.
The four ways November can land, as joint frequencies over the same 100,000 simulations that produce every other number on the site. Split control is the single most likely family of outcomes.

Where the forecast is weak, it says so. The Senate model reads about three points too Democratic out of sample, and that is disclosed rather than quietly corrected. Polls a race does not have are not invented. Prediction-market prices sit beside the forecast for comparison and carry no weight in it at all.

Where to start

The front page has both chambers, the map and the closest races on one screen. Any race page shows the forecast, the evidence that produced it, and the polls it admitted. The methodology is the whole model, and the poll archive is every poll with a link to its source.

If it is useful, keep it going

The running costs are real and they come out of one pocket — hosting, map tiles, data feeds. I have got this about as far as I can on my own, and further improvement now depends on help.

Nothing here is held back for supporters. There is no members' tier and no private numbers, so support buys time and data rather than access.

If you can: Ko-fi takes the smallest cut, and Buy Me a Coffee works too. If you would like your name on the supporters list, say so in the message and tell me exactly what name to use — and whether you are happy for the amount to be shown. If you say nothing, I will assume you would rather stay anonymous.

If you can't, that is completely fine. Read it, use it, and tell someone who would find it useful.

Corrections, bug reports and methodological objections are all welcome, in that order of usefulness. I would rather be corrected than be wrong in public, and I am always open to suggestions.

See you on election night.