Election Night 2026

What this page shows. Every number starts from the 40,000 simulations behind the Strength In Numbers 2026 midterm forecast. Each simulation is one complete outcome for every Senate and House race. As results come in, simulations that agree with them gain weight, so related races move together, and the chances, seat ranges, histograms and maps on this page come from the reweighted set. The page checks for new numbers automatically.

How results count. Once a race has votes counted, those votes are locked in: the model never re-estimates them, and only the votes still to be counted are projected. The forecast, updated with results elsewhere, and the county model make that projection. The county model compares the counties that have finished counting (in states that count fast, also those with at least 95% of their expected vote counted and a count that has stopped changing) with their results in the last presidential election, and its say grows as more of a race’s counties finish. The model does not project a race from its running totals, because the votes still out often lean differently from the votes counted first. Where a county counts a large batch of ballots centrally and reports it late, races with many of those ballots rest on the forecast until the batch is in. A call fixes the seat for the winning side. “Projected final” is the model’s estimate of the final two-party margin, the counted votes plus the projection of the rest, with its 95% range; the chances come from the same estimate. “What these terms mean,” above the races table, in each race’s details and on the county analysis tab, defines each label.

Chances and ratings. “Dem chance” is the chance that the Democratic side wins. Independents count on the side they run in place of: Dan Osborn (Nebraska Senate), Seth Bodnar (Montana Senate) and Bill Hill (Alaska’s House seat) with the Democrats, Kevin Kiley (California’s 6th District) with the Republicans. Ratings use the site’s buckets on the rounded chance: Toss-up 41–59%, Lean 60–74% (26–40% for Republicans), Likely 75–97% (3–25%) and Safe 98% or more (2% or less). Called races show “Dem win,” “Rep win” or “Ind win.”

Runoffs. In Georgia, and in Louisiana’s House races, a candidate needs a majority of the vote to win outright; otherwise the top two meet in a runoff in December. For a race headed to a runoff, the chance is the chance of winning the runoff; for one that may need a runoff, the chance of winning either way. The model takes the runoff’s margin to be the November count’s, give or take about eight points. The November count is shown separately: the first round’s projected final margin with its 95% range, and the leader’s share of the vote.

The county analysis compares each county’s results with its vote in the last presidential election. Counties still counting can lean toward whichever kinds of ballots are counted first, so they are marked separately; in states that count fast, counties with at least 95% of their expected vote counted and a count that has stopped changing count as done and are drawn with a dashed outline. Projected finals appear once the county model is active for the race; before that, once a race has votes, its counties still counting show estimated finals, spread from the race’s estimate. A county’s election-night count can be in before it is final; the tab marks such a county “count in” and uses it in its swing charts, while the model waits for the final count (in states that count fast, until its count has stopped changing).

Where the results come from. State and county election offices’ results pages, plus results entered by hand from official sources and news organizations’ calls. Some vote counts come from civicAPI when a state’s own results page is behind.