The 2026 forecast

Name: 2026 US midterm forecast — ratings + simulated seat distribution
Creator: Akashic
License: https://akashic.app/LICENSE.txt

One committed expectation for every U.S. Senate, House, and Governor race — rated by hand, expressed as calibrated win probabilities and a 50,000-run simulated seat distribution, and scored in public when the results land.

U.S. Senate56.1%Democrats favored to control the Senatemedian 51 D · 49 R 80% range 47–54 Dmarket 43.9%U.S. House79.1%Democrats favored to control the Housemedian 241 D · 194 R 80% range 201–279 Dmarket 86.4%Governors66.1%Democrats favored to hold a majority of statehousesmedian 27 D · 23 R 80% range 22–31 D

The seats that decide it

Every tilt-rated race and the closest leans, most competitive first — the ground where the majorities are actually won. Follow a race for the full call, its money, its polling, and its place page.

AlaskaSenateTilt DflipSullivanvPeltolano market yet AlaskaGovernorTilt DflipWilson·Begich·Kreiss-Tomkinsno market yet AZ-02HouseTilt DflipCranevNezmarket R 63%≠ our call CO-03HouseTilt DflipHurdvRomeromarket R 66%≠ our call CO-04HouseTilt Dflip

U.S. Senate · the rating map

rated: 35 races
projected flips: 7
median seats: 51 D · 49 R

The three chambers, simulated

Each chamber’s full seat distribution — 50,000 Monte-Carlo runs with a correlated national-swing error — beside the market’s implied odds on control.

U.S. Senate →

Democrats are favored to control the Senate — 56.1%

Democrats control

56.1%

Republicans control

43.9%

incl. 8.0% at 50–50 (VP breaks the tie)

Predicted Democratic seats

80% range 47–54

A 50–50 U.S. Senate is Republican-controlled — the Vice President breaks the tie — so Democrats need 51 seats and Republicans need 50.

Model vs. market

Across every 2026 race Polymarket is trading, these are the widest gaps between Akashic’s race rating and the market’s implied odds — our call against the crowd on a single Democratic-to-Republican axis. A market price is trader sentiment, not a forecast; the gap is the conversation.

2028 Place vote forecast laboratory →

Explore modeled Democratic and Republican vote distributions across 30,999 Census Places under the published national turnout and party-environment scenarios.

Foresight election-night desk →

When validated results arrive, the private model becomes a public evidence-backed probability desk. Probabilities remain distinct from human-approved race projections, and blocked contests explain why a number is being withheld.

The accountability scorecard →

Every locked call is scored in public when the results land — Brier, calibration, the named misses, and the seat-forecast error. The record starts with the forecast Akashic actually published in 2026: no retroactive calls and no simulated historical record.

Ratings are set by editorial judgment and expressed probabilistically; the seat distribution is a 50,000-run Monte-Carlo with a correlated national-swing error. See the methodology for the calibration and the simulation, and the scorecard for how every call is scored.