United States Presidential election7 Nov 2028in 716 days Senate D 99% House D 58% Calendar Track record Research
United States · live results · 3 November 2026
Election 2026 midterms Results

Each election opens in its own mode: the forecast before the votes are counted, the live count while races are still too close to call, and the results once every race is decided.

2026 midterm results

On election night we read the official results feeds of 26 states straight from their election offices, every few minutes, and show each Senate, governor and House race's count against the vote we expect there. A live model turns the count into a chance for each race and for control of each chamber, updated with every read, and we call a race once we are 97% sure of the winner. Our calls are our judgement; the states certify the results, weeks later.

How the live model works

  1. Start from our forecast. Each race starts from our pre-election margin and its uncertainty.
  2. Expect each county. Every county gets an expected margin (its 2024 presidential margin, moved so the counties add up to our forecast) and an expected vote (its 2022 midterm turnout, or its 2024 presidential turnout scaled down where no 2022 race is held).
  3. Read the swing. Where votes are counted, we compare each county's margin with what we expected there. Counties count for more the more of their vote is in: a half-counted county is often all mail or all Election Day ballots, which lean differently.
  4. Project the rest. The swing, shrunk toward zero by our pre-election uncertainty, is applied to the vote still to come in each county. The uncertainty left shrinks with the share of the expected vote still to count. It stays wider while counties are only partly in: their counted votes are often one kind (mail, early or Election Day) and lean one way, and we cannot see which way on the night. How many votes are still to come is uncertain too.
  5. Roll up. For control of the House and Senate we rerun our forecast's own simulations with every race that has votes counted drawn from its live estimate; races with nothing counted keep our forecast, moved by half the swing seen so far.

Where a state's feed gives only statewide totals, or for House districts (our county baselines are statewide), the race's total is compared with our forecast in the same way, with a wider uncertainty. The model says nothing until at least 2% of a race's expected vote is in.

It is behind the same switch as the live results: nothing is computed on the night until the owner switches live results on.

Backtest: the 2024 Senate races, replayed

The count order is simulated. No official feed we can read keeps its election-night versions: Clarity serves only the newest version of an election, and the other feeds only the final count. So each county's real 2024 Senate result is split into three batches (early and mail votes, then Election Day, then late mail and provisional ballots) arriving at random times, early reporters first, and the model reads the count as it would on the night. It starts from what our 2024 backtest forecast said on the eve (fitted without 2024), with the uncertainty that backtest showed in 2018 and 2022 (5.3 points), and knows nothing of 2024 before it is counted: county baselines are the 2020 presidential result and turnout. 26 races, 25 replays each, for three count orders.

Left out: California (two Senate races in 2024; the county file holds one); Connecticut (no county Senate result held); Indiana (no county Senate result held); Missouri (no county Senate result held); New Jersey (no county Senate result held).

Early and mail votes 6 points more Democratic than the county's final result, Election Day votes more Republican (the pattern of 2020-2024)

Share of expected vote countedFavorite wonLeader in the count wonBrier score, liveBrier, forecast aloneResult inside the 80% rangeAverage miss of the projected margin
10%92%82%0.0540.04178%4.7 pts
25%92%87%0.0570.04159%5.0 pts
50%92%90%0.0520.04167%4.6 pts
75%96%96%0.0280.041over 99%1.4 pts
90%96%97%0.0200.041over 99%0.5 pts
over 99%over 99%over 99%0.0010.041over 99%0.0 pts

Races the model favored the wrong side in at half the count (worst replay): Ohio (Democratic chance 70%, result R+3.6); Pennsylvania (Democratic chance 92%, result R+0.2).

Every batch votes like the county's final result (no count-order effect)

Share of expected vote countedFavorite wonLeader in the count wonBrier score, liveBrier, forecast aloneResult inside the 80% rangeAverage miss of the projected margin
10%96%82%0.0350.04198%2.2 pts
25%96%90%0.0340.04199%1.6 pts
50%96%96%0.0320.041over 99%0.9 pts
75%96%95%0.0250.041over 99%0.6 pts
90%96%97%0.0200.041over 99%0.5 pts
over 99%over 99%over 99%0.0010.041over 99%0.0 pts

Races the model favored the wrong side in at half the count (worst replay): Pennsylvania (Democratic chance 62%, result R+0.2).

Early votes 6 points more Republican than the final result (the reverse pattern)

Share of expected vote countedFavorite wonLeader in the count wonBrier score, liveBrier, forecast aloneResult inside the 80% rangeAverage miss of the projected margin
10%90%76%0.0530.04195%2.5 pts
25%86%82%0.0730.04194%3.3 pts
50%86%87%0.0810.04194%3.3 pts
75%93%94%0.0350.041over 99%0.7 pts
90%96%97%0.0210.04199%0.6 pts
over 99%over 99%over 99%0.0010.041over 99%0.0 pts

Races the model favored the wrong side in at half the count (worst replay): Arizona (Democratic chance 35%, result D+2.4); Michigan (Democratic chance 16%, result D+0.3); Nevada (Democratic chance 22%, result D+1.7); Wisconsin (Democratic chance 19%, result D+0.8).

Brier score: the average squared gap between the chance given and what happened (0 is perfect; always saying 50% scores 0.25). "Leader in the count won": how often simply following whoever led the count at that point picked the winner. Replays are seeded, so the page shows the same figures each time.

For each state we looked for the election office's own results feed for 3 November, its format, its terms and robots.txt, and whether it lets a named reader in. We read a state only if all of that is clear; a Cloudflare, Incapsula or Akamai challenge is a refusal and we never get round one. The AP's results are paid and not used; nor are any media feeds. Full notes: research/us/LIVE_RESULTS_SOURCES.md.

StateDecisionFeedPolls close (last, Eastern)
Alaska Read live Division of Elections JSON (first-choice counts) link 1:00 am ET
Arizona Read live Statewide site blocked (Cloudflare); read county by county: Maricopa, Pima, Pinal (about 80% of the vote) link 9:00 pm ET
Arkansas Read live Tally ENR JSON, with counties link 8:30 pm ET
Colorado Read live Clarity ENR JSON link 9:00 pm ET
Connecticut Read live JSON (statewide and district totals; towns, not counties) link 8:00 pm ET
Florida Read live Results extract, tab-delimited, by county link 8:00 pm ET
Georgia Read live Enhanced Voting JSON, with counties link 7:00 pm ET
Idaho Read live Enhanced Voting JSON link 11:00 pm ET
Indiana Read live Indiana Election Division JSON, House districts with counties link 7:00 pm ET
Iowa Read live Clarity ENR JSON link 9:00 pm ET
Michigan Read live No statewide feed; read county by county: Oakland, Macomb, Kent, Genesee, Ingham, Ottawa (about 40% of the vote; Wayne publishes PDFs only) link 9:00 pm ET
Minnesota Read live Media files by county, semicolon-delimited link 9:00 pm ET
Nebraska Read live Results CSV exports (statewide) link 10:00 pm ET
New Mexico Read live Results CSV exports (statewide) link 9:00 pm ET
North Carolina Read live NCSBE JSON, with counties link 7:30 pm ET
North Dakota Read live Results XML exports (statewide) link 10:00 pm ET
Ohio Read live Statewide site blocks our reader; read county by county: Cuyahoga, Hamilton, Summit, Montgomery, Lucas, Butler (about a third of the vote; Franklin publishes PDFs only) link 7:30 pm ET
Pennsylvania Read live Statewide site blocked (Incapsula); read county by county: Allegheny, Montgomery, Delaware, Chester, Lancaster, Lehigh, Luzerne, Erie (over a third of the vote; Philadelphia publishes HTML only) link 8:00 pm ET
South Carolina Read live Clarity ENR JSON link 7:00 pm ET
South Dakota Read live Results CSV exports (statewide) link 9:00 pm ET
Texas Read live Statewide site blocked (Cloudflare); read county by county: Harris, Dallas, Tarrant, Bexar, Travis, Denton, Williamson, Fort Bend (about 50% of the vote) link 9:00 pm ET
Utah Read live Enhanced Voting JSON, with counties link 10:00 pm ET
Vermont Read live JSON by town (statewide totals) link 7:00 pm ET
Virginia Read live Enhanced Voting JSON, with localities link 7:00 pm ET
Washington Read live Enhanced Voting JSON link 11:00 pm ET
West Virginia Read live Clarity ENR JSON link 7:30 pm ET
District of Columbia Open, not read yet JSON link 8:00 pm ET
Hawaii Open, not read yet Text summary files link midnight ET
Louisiana Open, not read yet JSON blobs (House seats are party primaries on 3 November) link 9:00 pm ET
Alabama Not read HTML pages only link 8:00 pm ET
California Not read Incapsula script on the results site link 11:00 pm ET
Delaware Not read HTML only link 8:00 pm ET
Illinois Not read No statewide feed on the night link 8:00 pm ET
Kansas Not read Power BI dashboard only link 9:00 pm ET
Kentucky Not read No data feed found link 7:00 pm ET
Maine Not read No election-night feed 8:00 pm ET
Maryland Not read No live data file found link 8:00 pm ET
Massachusetts Not read Incapsula script; results after the count 8:00 pm ET
Mississippi Not read PDF only link 8:00 pm ET
Missouri Not read Cloudflare block link 8:00 pm ET
Montana Not read Exports only through page forms link 10:00 pm ET
Nevada Not read Incapsula block link 10:00 pm ET
New Hampshire Not read No feed; the site answered Access Denied 8:00 pm ET
New Jersey Not read County clerks report; Incapsula script 8:00 pm ET
New York Not read Cloudflare challenge link 9:00 pm ET
Oklahoma Not read 403 Forbidden link 8:00 pm ET
Oregon Not read HTML; export address answered 404 link 11:00 pm ET
Rhode Island Not read Cloudflare challenge link 8:00 pm ET
Tennessee Not read 403 from CloudFront link 8:00 pm ET
Wisconsin Not read No statewide feed; Cloudflare 9:00 pm ET
Wyoming Not read PDF after the count 9:00 pm ET

Figures as of . Published by Polling Forecast.

You may also like