Democrats heavily favored to retake the House of Representatives
Le Millénaire · 200,000 simulations · Engine calculation: October 3, 2026 · election on November 3, 2026
Chances of controlling the House of Representatives
D
Democrats
94.8%
House majority
239.4
expected seats · 90%: 217 to 264
R
Republicans
5.2%
House majority
195.6
expected seats · 90%: 171 to 218
Majority at 218 seats out of 435. Generic Ballot (likely voters): D +7.35. Probabilities conditional on a Democrat-versus-Republican contest in each district. Updated October 3, 2026.
The engine produces the 435 predictive distributions and the national balance of power. It is recomputed every day with the latest polls. The probabilities assume a Democrat-versus-Republican contest in each district: uncontested seats, same-party contests and third parties are not modeled. This is a provisional forecast, not a scientific certification.
United States · 435 districts · same selection in both maps
435 embedded forecasts · map colored by the engine's probabilities.
P(D) · conditional model
Safe RLikely RLean RToss-upLean DLikely DSafe D
01NAVIGATE
Tap a district for its sheet. Enable the pan tool to move the map with your finger.
One square = one seat. Grouping by state is schematic. The “Geographic map” button shows district boundaries; both views open the same sheets.
Reference boundaries: Census as of January 1, 2026; Missouri as of January 1, 2025. An official boundary source does not, on its own, certify the prior's matching or all later redistricting changes. The map does not alter any model data.
Alphabetical shortcuts to four states; no electoral ranking.
THE HOUSE AT THE NATIONAL LEVEL
Democrats and Republicans
200,000 simulations · D/R conditionnel
Conditional D/R simulation. The figures assume each district is won by D or R. Uncontested cases, same-party contests and third parties are not certified in this calculation. Uncertainty is estimated on three cycles; the probability of control of the whole House has no complete historical validation.
D
Democrats
239.4
mean simulated seats
90% interval
217 – 264 seats
House control
94.79%
ABSOLUTE MAJORITY218out of 435 seats
Threshold applied to each of the 200,000 draws.
R
Republicans
195.6
mean simulated seats
90% interval
171 – 218 seats
House control
5.21%
Engine results embedded in the document. No result is recomputed in the browser.
The height of each bar shows how often that number of seats occurs in the simulations. The mean is not the most frequent scenario. The two parties' intervals do not add up bound to bound.
House engine forecasts 2026 for the 435 districts: central D and R scores, central lead, conditional D/R probabilities and mechanical classification. Third parties remain NON_MESURE.
State
D score
R score
Lead
P(D)
P(R)
Rating
View
Alaska
50.9%
49.1%
D +1.8
58.0%
42.0%
Toss-up
↗
Alabama
36.2%
63.8%
R +27.6
0.1%
99.9%
Safe R
↗
Alabama
47.2%
52.8%
R +5.6
26.6%
73.4%
Lean R
↗
Alabama
30.9%
69.1%
R +38.2
≤ 0.1%
≥ 99.9%
Safe R
↗
Alabama
21.0%
79.0%
R +58.1
≤ 0.1%
≥ 99.9%
Safe R
↗
Alabama
39.7%
60.3%
R +20.6
1.0%
99.0%
Safe R
↗
Alabama
36.5%
63.5%
R +27.0
0.1%
99.9%
Safe R
↗
Alabama
64.4%
35.6%
D +28.9
≥ 99.9%
≤ 0.1%
Safe D
↗
Arkansas
29.9%
70.1%
R +40.2
≤ 0.1%
≥ 99.9%
Safe R
↗
Arkansas
46.0%
54.0%
R +8.0
18.6%
81.4%
Likely R
↗
Arkansas
38.4%
61.6%
R +23.1
0.5%
99.5%
Safe R
↗
Arkansas
30.6%
69.4%
R +38.8
≤ 0.1%
≥ 99.9%
Safe R
↗
Arizona
52.1%
47.9%
D +4.2
68.0%
32.0%
Lean D
↗
Arizona
48.2%
51.8%
R +3.6
34.3%
65.7%
Lean R
↗
Arizona
76.1%
23.9%
D +52.2
≥ 99.9%
≤ 0.1%
Safe D
↗
Arizona
57.1%
42.9%
D +14.2
94.4%
5.6%
Likely D
↗
Arizona
44.3%
55.7%
R +11.4
10.0%
90.0%
Likely R
↗
Arizona
54.2%
45.8%
D +8.5
82.9%
17.1%
Likely D
↗
Arizona
66.9%
33.1%
D +33.9
≥ 99.9%
≤ 0.1%
Safe D
↗
Arizona
47.2%
52.8%
R +5.6
26.5%
73.5%
Lean R
↗
No district matches.
Try a state name, its abbreviation or an identifier such as CA-13.
435 forecasts in the document
All
D score: D vote share D/(D+R). P(D)/P(R): conditional model probabilities. The classes are derived mechanically from these probabilities. P(other) remains unmeasured; the 435 sheets also show the historical reference.
WHERE THE MAJORITY IS DECIDED
The districts in play
MECHANICAL THRESHOLD · P(D) 25–75%
This box mechanically isolates the districts where the conditional D/R model gives each party at least 25% chance of winning. It is not an editorial ranking.
43competitive districtsP(D) 25–75%19highly competitive districtsP(D) 40–60%11near 50/50P(D) 45–55%211–224D–R seats in 2024House result
District
D score 2026
R score 2026
P(D)
P(R)
2024 · D/R (D+R = 100)
50.9%
49.1%
58.0%
42.0%
49.5 / 50.5
47.2%
52.8%
26.6%
73.4%
54.6 / 45.4
52.1%
47.9%
68.0%
32.0%
48.1 / 51.9
48.2%
51.8%
34.3%
65.7%
45.5 / 54.5
47.2%
52.8%
26.5%
73.5%
43.5 / 56.5
47.5%
52.5%
28.5%
71.5%
44.7 / 55.3
49.9%
50.1%
49.2%
50.8%
47.4 / 52.6
48.9%
51.1%
40.4%
59.6%
43.9 / 56.1
49.7%
50.3%
46.9%
53.1%
42.8 / 57.2
47.7%
52.3%
30.6%
69.4%
42.7 / 57.3
48.1%
51.9%
33.4%
66.6%
43.5 / 56.5
48.3%
51.7%
35.1%
64.9%
45.2 / 54.8
49.7%
50.3%
47.0%
53.0%
57.8 / 42.2
47.5%
52.5%
28.5%
71.5%
40.5 / 59.5
49.3%
50.7%
43.8%
56.2%
55.0 / 45.0
50.5%
49.5%
54.3%
45.7%
54.5 / 45.5
48.7%
51.3%
38.6%
61.4%
42.1 / 57.9
49.4%
50.6%
44.8%
55.2%
50.3 / 49.7
50.5%
49.5%
54.4%
45.6%
44.1 / 55.9
52.7%
47.3%
72.9%
27.1%
48.1 / 51.9
50.4%
49.6%
53.3%
46.7%
46.8 / 53.2
47.9%
52.1%
31.9%
68.1%
41.4 / 58.6
49.2%
50.8%
42.9%
57.1%
43.8 / 56.2
48.6%
51.4%
37.9%
62.1%
46.0 / 54.0
50.0%
50.0%
50.1%
49.9%
50.9 / 49.1
47.5%
52.5%
29.1%
70.9%
0.0 / 100.0
49.3%
50.7%
43.6%
56.4%
43.2 / 56.8
47.3%
52.7%
27.2%
72.8%
0.0 / 100.0
50.8%
49.2%
56.9%
43.1%
44.8 / 55.2
48.6%
51.4%
37.7%
62.3%
38.0 / 62.0
47.2%
52.8%
26.3%
73.7%
41.4 / 58.6
51.0%
49.0%
58.5%
41.5%
50.4 / 49.6
47.1%
52.9%
26.0%
74.0%
40.5 / 59.5
48.5%
51.5%
36.9%
63.1%
43.5 / 56.5
51.3%
48.7%
61.3%
38.7%
43.6 / 56.4
48.0%
52.0%
33.0%
67.0%
41.7 / 58.3
51.3%
48.7%
61.5%
38.5%
52.8 / 47.2
50.5%
49.5%
54.4%
45.6%
51.3 / 48.7
49.7%
50.3%
47.3%
52.7%
67.4 / 32.6
50.2%
49.8%
51.5%
48.5%
43.6 / 56.4
48.7%
51.3%
38.8%
61.2%
42.5 / 57.5
49.9%
50.1%
49.2%
50.8%
44.8 / 55.2
52.7%
47.3%
72.8%
27.2%
48.6 / 51.4
2026: central engine outputs, calculation of October 3, 2026. 2024: D/R shares normalized on D+R; rounded archives from the project folder and targeted recheck of the Clerk for Alaska, Minnesota and North Dakota. These are not constant-boundary swings.
COMPARE THE ELECTORAL GEOGRAPHIES
Results 2024 ↔ Forecast 2026
435 results 2024 · 435 projections 2026
Move the divider to switch from the last election to the projection. Each map uses the 435 districts and the same color scale: the Democratic share of the D+R vote, not the probability of winning. Choose a district to read its figures.
100%
2024 · HOUSE RESULTSBoundaries of the 119th Congress
2026 · FORECASTProject folder geography
Loading geographic boundaries…Both complete series are already in the file. Boundaries are downloaded from Census.
All 2026All 2024
Same scale on both maps
R +40TieD +40
Darker = larger vote margin
One district, two readings
Boundaries can change between elections. The same number does not guarantee the same boundaries.
2024 · Result among D+R
D —R —
Rounded archives from the project folder; targeted recheck of ten districts.
2026 · Central D/R score
D —R —
Win probabilities shown separately upon selection.
Geographic comparison, not a uniform swing: 2024 uses the boundaries of the 119th Congress; 2026 uses the project folder's Census boundaries with the Missouri fix retained. These vintages do not certify later changes. The votes 2024 are normalized on D+R; third parties are excluded and elections without a D/R opponent do not measure a margin between two sides. Sources and limits ↗
The values of the 435 districts do not depend on the table's pagination.
THE NATIONAL SIGNAL
What the polls measure
MESURE · October 3, 2026
The Generic Ballot
A national measure of voting intentions for Democratic and Republican House candidates, without naming a local candidate.
The populations remain separate: likely voters (LV), registered voters (RV) and all adults are not interchangeable.
Descriptive aggregate from the identified data feed. A D−R margin is neither a seat split nor a probability of winning. No daily series is invented. [2]
10district waves in the snapshot
0source primaire verified in this corpus
These local reports remain outside the engine. Their scores are not presented here as validated measurements. The count describes the project folder's file, not all published polls. [4]
THE MILLÉNAIRE METHOD
An explicit baseline. Documented tests.
RECHECK · 21 SEPT. 2026
01
Start from the results
The electoral baseline separates the presidential reference from local persistence. The project folder's parameters remain unchanged.
02
Separate the information
The national signal, district polls and local effects remain separately identified. Unverified data are not included.
03
Measuring past error
The gaps shown here are recomputed on the archives of 2010, 2014 and 2018. A historical performance is not a guarantee for 2026.
Historical model error, in points
Same sample at each horizon · MAE and RMSE of two-party D share.
Horizon
Sample size
MAE M0
MAE M1
MAE M2
RMSE M2
J − 60
1,016
5.628
4.966
3.241
4.331
J − 30
1,016
5.628
4.966
3.251
4.323
J − 14
1,016
5.628
4.966
3.212
4.291
1,016 cycle-district pairs observed at three horizons, i.e. 3,048 rows, but only three distinct elections. The control 2022 to J−50 is not the complete integrated backtest. [5]
Parameters and status of the layers +
Retained parameters
Local persistence: incumbent α = 0.57; open seat α = 0.18. National sensitivity: β = 1. These coefficients remain unchanged; only the uncertainty components were estimated for the engine.
District polls
Specification M3 pre-registered, τ = 4 points, meets its historical criterion 2018. Replication on a second cycle remains open; its overall status remains A_MESURER.
Uncertainty
Historical residuals were used to estimate national, regional, state and district variances by nested moments, after correcting for noise in group means. The national component was recalibrated on October 1, 2026, of 0.980 to 1.457 point: this is the engine's measured national error at T−60. The region components (0.478), State (1.430) and district (3.937) are unchanged. Overall validation remains limited to three cycles. Unit: points of two-party D share.
Layers not included
Finance, conditional special elections and fundamentals: NON_MESURE. Second additive incumbency correction and group elasticities: REJECTED in the source registry.
One engine run. A provisional forecast.
200,000 correlated draws are recomputed every day. This page shows the outputs of engine M2, with the canonical prior unchanged, published as a provisional forecast. M3 remains excluded; electoral cases outside the D/R contest, geography and full national calibration are not certified by this run.
UPDATES & TRACEABILITY
A page linked to its data
Displayed dataOctober 3, 2026The date of a measurement is never replaced by the page generation date.
Page synchronizationDaily pipelineCollection of polls, recomputation of 200,000 draws, checks and publication every day. Displayed calculation: October 3, 2026.
House engine calculationRan · 200,000 drawsEvery day: collection of public polls, engine recomputation, automated checks, then publication. A day without a new poll does not change the figures.
The results are embedded in this file. A check of an external feed must never erase them; it is not a new engine calculation.
[2] Engine outputs: forecast_us_2026.json, calculation of October 3, 2026 (HOUSE-GB-REFRESH-20261003). Generic Ballot under rule Common V3, polls through October 3, 2026.
[3] Local reference.Corrected canonical prior V3. The page uses the historical field pres24, not the prior's score 2026. Values rounded to two decimals on screen.
[4] District polls.Local aggregate V1, project snapshot v8. No score from an unverified primary source is shown.
[5] Validation.Historical validation handoff, recheck of September 21, 2026. The metrics M0/M1/M2 come from archived values, without re-estimation.
About this edition. House V7: carried over from V6, shared banner of the Berlin reference, cartogram and geographic views. Model and prior unchanged. Viewing some Drive files requires authorization.
Comparator 2024/2026 — V1.7. The series 2026 comes from the 435 objets forecast.districts, never from the paginated table. For 2024, the archived scores are rescaled to D+R = 100; ten districts (AK, MN, ND) are corrected using the Clerk statistics (printed pages 2, 37, 54). The other rows remain rounded archives, without full primary revalidation. The percentages 2026 are conditional D/R.