2nd opinion
Read the forecast

Breakdowns

Who aligns where — and the math behind it.

Every claim below is a chart. What we can measure today is measured; what ships with the model is sketched, not faked.

Geographic · published

The map of margins

Published data

Geography · the 13 verified races

darker tile = bigger lead

The South runs red, the upper Midwest is the contested middle, and Maine and Iowa are pure coin flips — hover any tile for the expert rating, click to drill into the race.

Published data — RCP polling averages, Sep 28 2026 (Nebraska: Wedgewood, Sep 16). Hover a tile for the Cook rating; click to drill into the race.

Geographic · published

The margin ladder

Published data

Polling margin · the leader's lead, in points

pts · click a race to drill in
Dem leaderRep leaderExpert-rating consensus

Published data — polling leader and margin from the verified averages; tick marks show the expert-rating consensus per race.

Social · published

The slate

Published data

Social · who's running

26candidates
  • 12Democrats
  • 13Republicans
  • 1Independent

Open seats are where candidate quality bites hardest — six of the thirteen have no incumbent, and the WAR-style score exists precisely for them.

Published data — the 26 verified candidates across the thirteen marquee races.

Demographic · pending

Who the voters are

Pending

Vote share by group — college / non-college, urban / suburban / rural, age cohorts — per district, from ACS 5-year estimates crossed with the fundamentals layer.

Nothing fabricated: this visual ships with the model launch, computed from the cited sources.

Economic · pending

Who's funding it

Pending

FEC fundraising totals, cash-on-hand, and small-dollar share per candidate — the money layer of candidate quality, published with the model.

Nothing fabricated: this visual ships with the model launch, computed from the cited sources.

Scenario lab

What would it take?

Drag the what-ifs. The lab reruns 12,000 pretend elections live — same simulation skeleton as our real model.

Illustrative
Teaching toy

Not our forecast — this lab shows how the math works. The starting points are published data; every number that moves is illustrative.

Spot on
Dems underestimatedReps underestimated
No swing
Toward DemsToward Reps
60% together
Each race on its ownAll together

In plain words

If the polls are exactly right, Democrats take the Senate in almost none of the simulated elections — the most common outcome is 47 seats, 4 short of the 51 they need.

The six toss-ups, under your story

  • Texas

    D +2.9

    —

  • Ohio

    D +3.8

    —

  • Iowa

    D +0.5

    —

  • Michigan

    D +3

    —

  • Alaska

    D +2.3

    —

  • Maine

    D +0.9

    —

Chance each race's Democrat wins, given your sliders. Starting margins: published RCP averages, Sep 28 2026.

Democratic seats · 0 of 12,000 sims

0%chance Democrats reach 51 seats

47

median seats

47

most common

Running the simulation… 0 of 12,000

Democratic seats won

Read it like this: each bar is how many of the 12,000 pretend elections landed on that seat count under your story. The tallest bar is the most common outcome. Bars right of the clay line are a Democratic Senate.

Starting points: published RCP polling averages for the six toss-up races (Sep 28 2026); 47–47 consensus board via huffmanwrites.org, Sep 27 2026. The other 29 seats are held fixed — a deliberate simplification, stated plainly. All simulation output is illustrative.