Results against the forecasts

What it does. It sets results against takes registered before the polls closed: the published forecasts, history, and a checklist of 21 yes-or-no checks. The live view reads one state file. Calls come from AP and are entered by hand by Andrew or Karen; nothing on this page is called automatically.

History, in votes. The rungs translate past midterm swings in the national House vote into 2026 seats. The out-party swing averaged 7.0 points over the eight midterms from 1994 to 2022, and 10.0 points when the president’s approval was under 50%. Starting from 2024’s R+2.5, Split Ticket’s seat table turns those swings into seats. Forecasts expect a 1994-size swing; on 2026 maps that is about 240 House seats. That swing, about 11.7 points, is larger than 2018, 2006, 2014. So a night on forecast is already a historically large one. The Brookings figures exclude unopposed districts, and the seat curve is one model’s.

The checklist. Each check has Split Ticket’s pre-election chance. Par, the sum of the chances, is 12.35. The tiers compare the night with past midterms. Each check’s chance is moved to other national environments with Split Ticket’s margin and spread for that race, from its D+9 baseline (Split Ticket’s median of 239 House seats is its D+9 row): a repeat of 2024 (R+2.5) gives about 0.9 checks, a typical midterm swing (D+4.5) 6.3, a 2018-size swing (D+7.1) 9.7, the forecast (D+9) 12.3, a 1994-size swing (D+9.7) 13.3, D+12 16.0 and a 2010-size swing (D+14.7) 18.3. The cut points sit midway between those benchmarks: 4, 9, 12, 15, 18 checks. Par falls in “The forecast wave, 1994-size”. The KS Senate check replaced MN Senate in an audit before the freeze, because MN (91%) carried little signal; NC Senate (98%) stays as one of the races named when the checklist was set, though it carries little signal too. The page says “Too early to tell” until 5 checks have settled and it is past 7:30 PM ET; after that it sets the checks met against the par so far, the sum of the chances of the checks settled so far, and says whether Democrats are running ahead of it, on it, or behind it (a gap of one check or more). The tiers describe the whole night, so during the night the strip shows where Democrats are on pace for: the checks met plus the chances of the checks still open, placed on the tier cut points. The tier is shown as final only once every check has settled, or after 2:00 AM ET.

Per-race chances. Senate and House from Split Ticket. Governors from Polymarket prices (a market snapshot taken Mon Nov 2, 3:49 PM ET), labelled “market odds”, because Split Ticket publishes none. Polymarket lists 21 of the 36 governor races; the other 15 show no chance.

Live odds. Polymarket chamber-control prices are read by your own browser from Polymarket’s public API, every 90 seconds while this page is open and visible, each with the time it was read and a link to the market. This site does not store or relay them. If the read fails, or the editors pause it, the page shows the snapshot from Mon Nov 2, 3:49 PM ET, marked as such. The NYT needle values are typed in by hand from the NYT’s page, each with the time it was read; this page never fetches anything from the NYT. Markets swing hard on early, unrepresentative counts.

Why Kalshi is absent. Kalshi’s data terms bar publicly displaying its prices without written permission, and its API agreement limits use to a member’s own trading. So this page shows no Kalshi numbers. Its markets are at kalshi.com.

What the results so far look like. Once 8 competitive results have settled (called Senate and House races at 15–85% by Split Ticket, plus the settled checks), the page finds the national environment that best explains them: each race’s chance is moved with its Split Ticket margin and spread, as for the tiers, and the most likely environment is read off a grid from R+4 to D+16, with a rough 80% range. Early calls skew toward easy races, every race is assumed to move together, and it uses one model (Split Ticket). Governors are left out: Split Ticket has no governor margins.

Toss-ups and stars. Each tab names the toss-ups that decide it, after the idea in the National Journal Hotline, The Insider's Guide to the Midterms (Cook ratings as of Oct 27): the seats leaning or safe by Split Ticket (60% or more for either side) form the base, and the chips are the races Split Ticket gives Democrats a 30–70% chance, plus Hotline’s toss-ups (its top 10 for the House; Polymarket prices and Hotline’s toss-ups for governors). A chip fills when AP calls the race. The Target states tab follows the chambers, statewide offices and contested followed legislators in 14 states. “Majority reached” appears only when called seats settle control; Republicans keep the Senate at 50–50 through the vice president. The star on a race keeps it in “My races” and alerts you when it is called; stars stay in this browser only.

Every analyst’s bar, hour by hour, is on the Details page.

Built Mon Nov 2, 9:54 PM ET. Frozen at Mon Nov 2, 9:54 PM ET. The checklist and the analysts’ bars are not edited after the freeze; these SHA-256 hashes let anyone check that: