Vol. I · No. 1A Daily Projection · Read As ResultsElection Night · Nov. 3, 2026

Call the Map

"All the results that fit — one night at a time."

Complete ReturnsSenate · Governors · House · LegislaturesNot yet rolled

A model-based projection of the 2026 U.S. midterms, refreshed daily and presented as if the votes were already counted — for infotainment, not a real tally. * marks a primary not yet decided (most likely nominee shown). Tap any race for detail; hit New Scenario to reshuffle within the model’s uncertainty.

Simulation rolls one plausible election night. Forecast runs the model 10,000 times and reports its single most-likely result with win probabilities — the accountable version you can grade after Nov. 3.

Call the Map · Forecast Model Brier-scored after Election Day

Distribution of Democratic Senate seats · 10,000 simulations

U.S. Senate—

—–—

100 seats · 51 for control

U.S. House—

—–—

435 seats · 218 for control

Governors—

—–—

50 states · 36 up tonight

State Chambers—

—–—

99 chambers · 15 in play

The national mood

—

Roll to draw one of a million nights.

House popular vote

—

national two-party vote

How unusual was tonight

—

upsets against the ratings

The Senate

35 of 100 seats up

The Governors

36 of 50 mansions up

The House

435 seats · ~40 battlegrounds

U.S. House — all 435 districts · real 119th-Congress boundaries (2025 mid-decade maps pending in some states) · hover any district

R+40District margin · battlegrounds outlined in blackD+40

Hover or tap any district for its result.

Roll the night or run the forecast to color the map.

The State Legislatures

99 chambers · 15 competitive

Flagship Race — North Carolina Senate

Cooper (D) vs. Whatley (R) · fully built

The proof-of-concept for every race: a real 100-county map with genuine 2024 baselines, poll averages, demographic crosstabs, fundamentals, and a result that regenerates on every roll — counties correlated by a shared statewide swing. This is the depth each of the ~60 competitive races will eventually reach.

R+40County margin — roll to generateD+40

Hover or tap a county for its result.

Result

Roll the night (or switch to Forecast) to generate.

Recent polling average

Fundamentals

Exit-poll crosstabs (model estimate)

Build Log & Next Steps the model audits itself

On every load, the project inspects its own data and code, scores how complete each system is, and decides what's most worth building next. This is the to-do list the site keeps for itself — and for you.

System status

The model's own feature ideas

How the night is made

A forecast wearing a results page's clothes. The presentation is journalism; the machinery is a model — shown here in the open.

Roll, don't set

Each roll draws one national environment from a continuum of a million nights, weighted to the 2026 consensus: a Democratic-leaning midterm under a Republican president, spring generic ballot near D+7, with Trump's approval underwater. You can't pick the weather — you roll it, and the night prints its own national popular vote.

Every race then draws its own margin around a hand-set baseline, scaled by an elasticity (how hard the national mood moves it) and nudged by real inputs: incumbency, open seats, 2024 presidential margin, candidate quality, fundraising, and a left-insurgency factor for the DSA/Mamdani-backed candidates who swept the spring primaries — energizing in blue seats, riskier in swing ones. A shared state swing ties each state's races together on the night.

Calibrated to real upset rates

Pool weights trace to the Cook Political Report's own rating accuracy, 1984–2022 final ratings:

Rating	Holds	In model
Safe	≈ 99.9%	≈ never flips
Likely	≈ 97%	~3% upset
Lean	≈ 94%	~6% flip
Toss-Up	≈ 50/50	true coin flip

An upset counts only when a result defies both its rating and the night's tide — so a Lean-R seat falling in a blue wave is the wave, not a surprise. Every competitive race is tappable for its exact result, exit-poll crosstabs, sources, candidate data, and a written account of how the model got there.