COYNTOSSby @achasecoyne
2026 results

Archive

Model Changes

A dated record of changes to how COYNTOSS models and averages work. Newest first. Entries are added, never rewritten.

Changelog

  1. State approval

    State approval estimated for every state, with district, county and precinct views

    • Before, the approval map showed only the 11 states with a state poll in the past year. Every state and DC now has an estimate: the national average shifted by the state's 2024 lean (log-odds of Trump's two-party share there minus nationally) times a slope fitted to the state polls of the past year (0.99 from 16 polls).
    • States with their own polls blend that estimate with the polls' gap from the national average on their end dates, weight n / (n + 1) for n polls, and are outlined on the map. Their numbers can differ from the raw state-poll average because older polls are moved with the national trend since they ended.
    • The district, county and precinct views apply each state's estimate to 2024 presidential results with the same calibration the 2026 forecast maps used. They are estimates, not measurements.
    • National averages are unchanged.

  2. Sources

    July–September 2026 polls filled in from pollster releases

    • VoteHub and Wikipedia carried few national approval and generic-ballot polls with fieldwork from July through September 2026. We added 49 poll results transcribed from the pollsters' own releases (Economist/YouGov, CBS News/YouGov, Reuters/Ipsos, Washington Post-Ipsos, Echelon Insights, Quinnipiac, AP-NORC and Emerson), each linked to the document it came from.
    • One stored Wikipedia row was wrong: it listed the July 25–27 Economist/YouGov all-adult numbers (34% approve, 62% disapprove) as registered voters. The release copy (39%, 59% among registered voters) replaces it.
    • The current Trump approval average is unchanged at 37.8% approve, 59.8% disapprove. The final 2026 generic ballot average moved from D 50.1, R 42.6 (D+7.5) to D 49.6, R 41.7 (D+7.9), and now rests on 24 polls instead of 10.

  3. Polling averages v1.2

    Smoother Trump approval average in quiet stretches

    • When few approval polls are out, the half-life now stretches (up to 30 days) until it covers the 10 newest polls. Before, one or two polls after a quiet month could move the line by more than 3 points in a day.
    • Our sources have almost no approval polls from July through September 2026, which is where the old line jumped most.
    • The current average moved from 38.0% approve, 59.6% disapprove to 37.8% and 59.8%. The 2026 generic ballot average is final and was not recomputed.

  4. Polling averages v1.1

    More 2028 matchups and betting-market odds

    • Matchups tested in a single poll in the past year now appear when both candidates are in their party's primary average. Pairings with a fringe candidate still need two polls.
    • The 2028 page and homepage now show Kalshi's nominee markets, the midpoint of the best bid and ask for each candidate. They are shown as Kalshi's prices, labeled and linked, and are never blended into a COYNTOSS average.
    • Candidate photos come from Wikimedia Commons, public-domain or Creative Commons images only, credited on the methodology page.

  5. Polling averages v1

    New national polling averages and the off-season site

    • Added COYNTOSS averages for Trump job approval (national and by state), the 2028 Democratic and Republican primaries, 2028 head-to-head matchups and the generic congressional ballot.
    • Weights use recency (half-lives of 10 to 60 days by tracker), sample size, a pollster-frequency adjustment and a half weight for partisan sponsors. No house-effect or pollster-quality adjustment yet.
    • Polls come from VoteHub's public poll API and Wikipedia poll tables. Only raw poll results are used, never another site's average.
    • The 2026 forecasts moved to the Archive; the homepage now leads with the trackers.

  6. 2026 forecast 2026-11-03-02

    2026 cycle closed; final forecast frozen

    • Release 2026-11-03-02 is the final 2026 forecast. Every 2026 page stays at its URL, unchanged.
    • Added the 2026 in review page, which compares the final forecast with called results. The forecast is never re-run.

  7. Senate forecast 2026-11-03-02

    Election Day forecast update

    • Final update from the frozen workbook simulation runs.
    • The map gained a winners-only view and striped patterns for projected seat flips.

  8. Site

    Navigation and state-page graph order

    • State and governor pages show the win-probability graph first, then the polling average, then a condensed model-margin history.

  9. Senate forecast

    Senate forecast reconstructed from Sept. 9

    • The Senate forecast was rebuilt from Sept. 9, 2026 using a corrected workbook: Abacus Data polls removed as bad data, the Montana treatment repaired, and polls limited by their release date.
    • Reconstructed points are labeled on the charts. The original releases stay archived; nothing was deleted.

2026 forecast update notes

The notes published with each 2026 forecast update.

Show all 36 update notes

  1. Two new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of November 3. 2 poll records were added: 2 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 75.27% chance of Democratic control, up from 74.66%, and 51.960 expected Democratic seats, up from 51.911. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  2. 42 new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of November 3. 42 poll records were added: 40 enter the polling averages and 2 remain excluded under the workbook release-date plan. The frozen simulation reports a 74.66% chance of Democratic control, up from 72.99%, and 51.911 expected Democratic seats, up from 51.781. Probabilities are published for all 35 Senate races.

    No Senate race rating changed.

    The House forecast now uses the House workbook as of November 3. The frozen simulation gives Democrats a 83.79% chance of House control (previously 82.95%) and 239.3 expected Democratic seats (previously 238.6).

    The Governor forecast now uses workbook run GOV-20261103-100K-S260831-R1 as of November 3. The frozen simulation gives Democrats a 78.80% chance of a majority and 27.43 expected Democratic governors (previously 75.71% and 27.19). 32 new Governor poll records were added.

  3. 14 new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of November 2. 14 poll records were added: 14 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 72.99% chance of Democratic control, up from 71.94%, and 51.781 expected Democratic seats, up from 51.706. Probabilities are published for all 35 Senate races.

    Senate rating changes: Nebraska Likely R to Lean R.

    The House forecast now uses the House workbook as of November 2. The frozen simulation gives Democrats a 82.95% chance of House control (previously 83.96%) and 238.6 expected Democratic seats (previously 239.0).

    The Governor forecast now uses workbook run GOV-20261102-100K-S260831-R4 as of November 2. The frozen simulation gives Democrats a 75.71% chance of a majority and 27.19 expected Democratic governors (previously 71.05% and 26.81). 22 new Governor poll records were added.

  4. Six new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of November 1. 6 poll records were added: 6 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 71.94% chance of Democratic control, up from 70.11%, and 51.706 expected Democratic seats, up from 51.603. Probabilities are published for all 35 Senate races.

    Senate rating changes: Maine Toss-up to Tilt D. The governor forecast is unchanged.

    The House forecast now uses the House workbook as of November 1. The frozen simulation gives Democrats a 83.96% chance of House control (previously 83.85%) and 239.0 expected Democratic seats (previously 238.8).

  5. Seven new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of November 1. 7 poll records were added: 7 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 70.11% chance of Democratic control, up from 69.82%, and 51.603 expected Democratic seats, up from 51.581. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The House forecast is unchanged.

    The Governor forecast now uses workbook run GOV-20261101-100K-S260831-R2 as of November 1. The frozen simulation gives Democrats a 71.05% chance of a majority and 26.81 expected Democratic governors (previously 70.61% and 26.78). 9 new Governor poll records were added.

  6. One new Senate poll

    The Senate snapshot now uses the RETROACT workbook as of October 31. 1 poll record was added: 1 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 69.82% chance of Democratic control, down from 70.85%, and 51.581 expected Democratic seats, down from 51.664. Probabilities are published for all 35 Senate races.

    Senate rating changes: Nebraska Lean R to Likely R.

    The House forecast now uses the House workbook as of October 31. The frozen simulation gives Democrats a 83.85% chance of House control (previously 80.54%) and 238.8 expected Democratic seats (previously 236.2).

    The Governor forecast now uses workbook run GOV-20261031-100K-S260831-R2 as of October 31. The frozen simulation gives Democrats a 70.61% chance of a majority and 26.78 expected Democratic governors (previously 64.65% and 26.40). 46 new Governor poll records were added.

  7. Six new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of October 30. 6 poll records were added: 5 enter the polling averages and 1 remain excluded under the workbook release-date plan. The frozen simulation reports a 70.85% chance of Democratic control, up from 70.56%, and 51.664 expected Democratic seats, up from 51.642. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  8. Two new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of October 30. 2 poll records were added: 2 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 70.56% chance of Democratic control, down from 70.61%, and 51.642 expected Democratic seats, down from 51.642. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  9. One new Senate poll

    The Senate snapshot now uses the RETROACT workbook as of October 30. 1 poll record was added: 1 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 70.61% chance of Democratic control, up from 69.14%, and 51.642 expected Democratic seats, up from 51.515. Probabilities are published for all 35 Senate races.

    Senate rating changes: Kansas Tilt R to Toss-up. The governor and House forecasts are unchanged.

  10. One new Senate poll

    The Senate snapshot now uses the RETROACT workbook as of October 29. 1 poll record was added: 1 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 69.14% chance of Democratic control, up from 69.08%, and 51.515 expected Democratic seats, down from 51.521. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  11. 12 new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of October 29. 12 poll records were added: 12 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 69.08% chance of Democratic control, down from 69.14%, and 51.521 expected Democratic seats, down from 51.540. Probabilities are published for all 35 Senate races.

    Senate rating changes: Montana Safe R to Likely R. The governor and House forecasts are unchanged.

  12. Six new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of October 29. 6 poll records were added: 6 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 69.14% chance of Democratic control, up from 68.01%, and 51.540 expected Democratic seats, up from 51.448. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  13. 15 new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of October 28. 15 poll records were added: 15 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 68.01% chance of Democratic control, up from 66.79%, and 51.448 expected Democratic seats, up from 51.383. Probabilities are published for all 35 Senate races.

    Senate rating changes: South Carolina Likely R to Lean R. The governor and House forecasts are unchanged.

  14. Seven new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of October 26. 7 poll records were added: 7 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 66.79% chance of Democratic control, down from 68.08%, and 51.383 expected Democratic seats, down from 51.455. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  15. Seven new Senate polls

    The Senate snapshot now uses the RETROACT workbook as of October 24. 7 poll records were added: 7 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 68.08% chance of Democratic control, down from 68.62%, and 51.455 expected Democratic seats, down from 51.489. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  16. New Senate polls and an updated frozen simulation

    The Senate snapshot now uses the RETROACT workbook as of October 22. 19 poll records were added: 19 enter the polling averages and 0 remain excluded under the workbook release-date plan. The frozen simulation reports a 68.62% chance of Democratic control, up from 66.96%, and 51.489 expected Democratic seats, up from 51.376. Probabilities are published for all 35 Senate races.

    No Senate race rating changed. The governor and House forecasts are unchanged.

  17. Poll records reconciled across the three forecasts

    The Senate snapshot now uses the RETROACT workbook as of October 20. It has 349 poll records, including 319 in the polling averages; 20 new records were added and six Wedgewood Polls records were removed. The frozen simulation reports a 66.958% chance of Democratic control and 51.376 expected Democratic seats. Probabilities are published for all 35 Senate races.

    The governor workbook adds 20 poll records and removes five Wedgewood Polls records. One new poll enters the October 18 averages and 19 remain excluded by the workbook release-date plan. Six race probability snapshots are withheld because their margins differ from the frozen simulation inputs; the governor model remains dated October 18.

    The House workbook adds five poll records and removes two Wedgewood Polls records, bringing the feed to 281 records. All five new polls remain excluded under the October 18 model release plan; the House model and simulation remain dated October 18.

  18. Poll updates across the three forecasts

    Three new Senate poll records were added. All three remain outside the polling averages under the workbook release plan. Democrats’ chance of Senate control moves from 66.57% to 66.84%, and expected Democratic caucus seats move from 51.381 to 51.394.

    Eight new governor polls were added, and all eight enter the averages. Nevada changes from Tilt R to Toss-up and Vermont from Likely R to Lean R. Democrats’ chance of a governor majority moves from 63.712% to 64.647%; expected Democratic governors move from 26.32977 to 26.40002.

    The House forecast adds three poll records. FL-05 changes from Solid R to Likely R, IN-01 from Likely D to Solid D, and TX-22 from Solid R to Likely R. Expected Democratic seats move from 235.3924 to 236.1762, and the chance of House control moves from 79.35% to 80.54%.

  19. House polls and expert ratings update

    The House forecast adds 129 poll records across 100 districts, bringing the total to 275 records and 219 included polls. The model now incorporates Inside Elections ratings at up to 30% weight, with a six-point adjustment cap and eligibility for races within 18 points. The extra poll-blend discount has been removed.

    Democrats’ expected House seats move from 241.71 in the previous published House forecast to 235.39; their chance of House control moves from 87.19% to 79.35%. On the same updated polling inputs, the expert setting moves expected seats from 242.33 to 235.39. No House rating category changes. Nick LaLota, Rob Wittman and Bryan Steil become favored over Chris Gallant, Shannon Taylor and Mitchell Berman, respectively.

    The expert category-to-probability mapping is provisional. Prior backtesting did not establish better House seat-total accuracy from expert ratings. Senate and governor forecasts are unchanged.

  20. Six new Senate polls; the ratings hold

    Six polls join the Senate forecast: InsiderAdvantage in Texas, Kansas and Maine; Bowling Green/YouGov in Ohio; Rasmussen in Michigan; and Talk Business/Hendrix in Arkansas. That brings the number included in the polling averages to 310.

    Texas edges from D+1.4 to D+1.2. Ohio moves from D+2.3 to D+2.6, Michigan from D+3.9 to D+3.8, Kansas from R+2.9 to R+3.0 and Arkansas from R+16.1 to R+15.3. Maine remains D+1.8 after rounding. None changes rating. Democrats’ chance of Senate control rises from 66.37% to 66.57%; their expected seat count rises from 51.373 to 51.381. The House and governor forecasts are unchanged.

  21. Iowa remains close; Texas inches toward Talarico

    Six Senate polls were added across five states, including two in Iowa. The others cover Minnesota, Florida, New Mexico and Texas. All six enter the polling averages, for a total of 304 included polls.

    Iowa moves from R+0.8 to R+0.5 and Texas from D+1.3 to D+1.4. Minnesota slips from D+5.7 to D+5.6; Florida moves from R+6.2 to R+6.1; New Mexico from D+16.2 to D+16.6. The ratings do not change. Democrats’ chance of Senate control rises from 64.66% to 66.37%. The House and governor forecasts are unchanged.

  22. Mississippi becomes more competitive; five governor ratings change

    The update adds 11 Senate polls, 59 governor polls and four House polls. Mississippi Senate moves from R+16.9 to R+12.8, shifting from Safe R to Likely R. Rhode Island moves from D+26.1 to D+30.0.

    In the governor forecasts, Alaska changes from Lean R to Tilt D, Kansas and Texas from Lean R to Tilt R, Florida from Lean R to Likely R, and Vermont from Lean R to Likely R. New Mexico moves from D+12.3 to D+9.1; Illinois from D+15.4 to D+17.9. House probabilities remain experimental and have not been historically calibrated.

  23. Maine’s second district tightens after two polls

    Two polls were added in ME-02. The forecast moves from D+3.1 to D+2.3, and Paul LePage’s chance of winning rises from 37.8% to 40.0%.

    The update also corrects nine incumbency classifications and three candidate names. Nine California districts have finalists from the same party, so that party’s seat is fixed in the simulations. The House probabilities remain experimental and uncalibrated.

  24. Texas’s 15th district moves to Toss-up

    Two polls were added in TX-15, along with 18 other House poll records. The Texas Southern survey comes from August; field dates and sample details for the Normington Petts poll come from an archive because the original posts were inaccessible.

    TX-15 moves from R+4.6 to R+0.4, changing from Lean R to Toss-up. NC-01 moves from D+0.3 to D+2.7. KY-06 changes from Likely R to Toss-up, while TX-28 changes from Toss-up to Lean D. Democrats’ chance of House control rises from 87.06% to 87.82%. The model’s coefficients and uncertainty settings are unchanged.

  25. Twenty-two Senate polls enter the forecast

    The added polls cover Alaska, Arkansas, Florida, Georgia, Iowa, Kansas, Maine, Michigan, Minnesota, North Carolina, Ohio, South Carolina and Texas. Two additional polls, in Minnesota and North Carolina, are recorded but do not yet enter the averages.

    Iowa moves from R+0.65 to R+0.96 and Minnesota from D+5.32 to D+5.68. No rating changes. Democrats’ chance of Senate control moves from 66.15% to 65.61%. The House and governor forecasts are unchanged.

  26. Five NYT/Siena polls leave the Senate ratings unchanged

    New York Times/Siena polls were added in Alaska, Iowa, Kansas, Ohio and Texas. Alaska moves from D+1.29 to D+1.82 and Kansas from R+3.96 to R+3.68. No race changes rating. Democrats’ chance of Senate control rises from 62.64% to 63.83%.

  27. Small moves in Kansas, Maine and South Carolina

    Eight polls were added across Alaska, Kansas, Maine, Michigan, Minnesota, Nebraska and South Carolina. Kansas moves from R+4.48 to R+3.96, Maine from D+1.74 to D+1.47, and South Carolina from R+6.13 to R+5.45. No rating changes. Democrats’ chance of Senate control moves from 63.06% to 62.64%.

  28. A revised Senate forecast and a fuller history

    The Senate forecast now excludes Abacus Data polls and counts polls from their public release dates. Montana is modeled as Seth Bodnar, an independent, against Republican Kurt Alme. Daily historical estimates now cover September 9 through September 30.

    Kansas changes from Likely R at R+5.50 to Lean R at R+4.48. Montana changes from Safe R at R+17.24 to Likely R at R+13.66. Democrats’ chance of Senate control moves from 61.04% to 63.06%. Win probabilities for Massachusetts and Rhode Island were not published in this update.

  29. South Carolina remains a Republican favorite

    New InsiderAdvantage polls were added in Iowa and Maine. The forecast gives Josh Turek a 41.5% chance in Iowa and Troy Jackson a 70.9% chance in Maine. In South Carolina, the Republican’s chance is 90.7%, against 9.3% for Annie Andrews.

  30. Georgia Senate moves to Tilt D

    Georgia moves from D+2.73 to D+1.23, changing from Lean D to Tilt D. Democrats’ chance of Senate control rises from 59.75% to 60.38%, with expected seats moving from 50.93 to 50.95.

    This update includes four Senate polls and three governor polls. The governor majority forecast is unchanged.

  31. Ohio moves to Toss-up

    Ohio Senate moves from D+2.07 to D+1.82, crossing from Tilt D to Toss-up. Nebraska governor moves from R+13.91 to R+17.03, changing from Likely R to Safe R.

    Other Senate changes are smaller: Kansas moves from R+6.53 to R+5.27, Maine from D+2.03 to D+2.25, Michigan from D+3.32 to D+3.35, Nebraska from R+5.32 to R+5.13, and North Carolina from D+7.90 to D+7.84.

  32. Marist finds Democrats ahead in both Texas races

    Marist’s September 17–20 Texas survey has James Talarico ahead of Ken Paxton, 50–44, and Gina Hinojosa ahead of Greg Abbott, 49–46. Both polls entered the September 22 averages. Texas Senate moves from D+1.07 to D+1.23 and remains a Toss-up; the governor forecast moves from R+6.43 to R+5.70 and remains Likely R.

  33. Iowa and Michigan governor forecasts move toward Democrats

    Thirteen polls were added: seven Senate, five governor and one House. The Iowa and Michigan governor margins move 2.50 and 2.49 points toward Democrats, respectively. The statewide poll archive now contains 437 records, and the House archive contains 115.

  34. The national polling input moves to D+8.65

    The forecast now uses Silver Bulletin’s D+8.65 generic-ballot average. The poll update adds 40 records and removes two Alaska duplicates, for a net increase of 38. Iowa Senate changes from Tilt R to Toss-up; New Hampshire changes from Lean D to Likely D.

  35. New Hampshire moves to Lean D

    New Hampshire Senate changes from Likely D to Lean D, with its projected margin moving from D+5.61 to D+4.70. Five polls enter the averages. Florida and Montana polls under review remain excluded.

  36. New polls in Maine, Michigan, Texas and Minnesota

    CNN has Troy Jackson ahead of Susan Collins in Maine, 48–45. Michigan’s polls disagree: CNN has Abdul El-Sayed ahead, 47–44, while Detroit News/Glengariff has Mike Rogers ahead, 46–44. YouGov has James Talarico ahead in Texas, 48–43, and TIPP has Peggy Flanagan ahead in Minnesota, 44–42.

    This is the first date shown in the forecast history. Later historical reconstructions can differ from the estimates published at the time; the charts identify which series they use.