CTM Forecast Lab: First Experimental 2026 House Probability Estimates
CAPITOL TIMES MEDIA | FORECAST LAB | OCTOBER 10, 2026
EXPERIMENTAL MODEL PREVIEW — NOT THE APPROVED CTM ELECTION FORECAST.
Capitol Times Media is making three research-stage, polling-only U.S. House win-probability estimates publicly accessible as a limited methodology preview. These calculations are real outputs from our documented historical model, but they are not the official CTM election system's released predictions. The production forecast remains in validation, and the Forecast Center's official probability publishing gate is still closed.
The figures below estimate the chance of winning each two-major-party contest, not projected vote share. Figures are rounded to whole percentages for readability. They must not be treated as precise probabilities or election-day outcomes.
Pennsylvania's 7th Congressional District
EXPERIMENTAL CHANCE OF WINNING: Bob Brooks (D) 74% | Ryan Mackenzie (R) 26%.
The model's internal weighted polling margin is Democratic +5.82 points. A Franklin & Marshall College survey fielded September 9–20 found Brooks at 45% and Mackenzie at 40% among 357 registered voters. A second, older Democratic-affiliated survey also contributes to the internal weighted margin.
Original poll: Franklin & Marshall College.
New York's 17th Congressional District
EXPERIMENTAL CHANCE OF WINNING: Cait Conley (D) 52% | Mike Lawler (R) 48%.
The model's internal weighted polling margin is Democratic +1.10 points. Emerson College Polling/PIX11 reported Conley at 48% and Lawler at 46% in a September 27–29 survey of 400 likely voters, with approximately 7% undecided. A near-even research probability should not be interpreted as a decisive advantage.
Original survey: Emerson College Polling.
Washington's 3rd Congressional District
EXPERIMENTAL CHANCE OF WINNING: Marie Gluesenkamp Perez (D) 41% | John Braun (R) 59%.
This research model uses a Democratic −1.05-point polling margin from a single September 17–27 DHM Research survey of 543 likely voters commissioned by Oregon Public Broadcasting. With leaners included, that survey reported Gluesenkamp Perez 47% and Braun 48%. This isolated poll-only output omits important incumbency, district and national context.
Important disagreement: Cook Political Report's October 9 House ratings moved Washington's 3rd District toward Democrats, illustrating how this narrowly specified poll-only research result can diverge from broader expert assessments. The number above is not a final CTM prediction or endorsement of a candidate.
Original polling: Oregon Public Broadcasting. Independent ratings: Cook Political Report, October 9.
How the experimental model was calculated
The research formula is P(Democratic win) = clip[normalCDF((weighted Democratic-minus-Republican two-party polling margin − 0.7361) / 7.7645), 2%, 98%]. The poll-error mean and spread were estimated using 74 reconstructed survey observations from 38 House districts in a 2022 historical snapshot, with only polling available by October 9, 2022 included.
The frozen formula was then evaluated on 33 House races in the separate 2024 election dataset. It produced a Brier score of 0.163, log loss 0.476, and correctly classified winners in 75.8% of that selected historical sample. These retrospective results do not establish prospective 2026 accuracy, and the training and test samples are small, selected groups of polled districts rather than all U.S. House seats.
Archived poll-data reference: Historical FiveThirtyEight House poll CSV copy.
Why these numbers are preliminary
This is a deliberately limited poll-only model. It does not apply a validated 2026 national generic-ballot-to-district swing coefficient and does not model expected turnout, candidate quality, incumbency, fundraising, district-specific demographic composition, economic conditions or the political environment. A hypothetical five-point district shift substantially changes the numbers in all three races; that stress test is not a confidence interval.
No House or Senate chamber-control forecast can be inferred from this three-race preview. The model has not been approved for publication as CTM's official production forecast. Future releases may revise these research figures materially after incorporating better data, validation and model approval.
Visit the Capitol Times Election Forecast Center for the broader race tracker. Official numerical forecasts will appear there only after the model clears its separate approval and publication checks.
Editorial disclosure: The pilot estimates describe model uncertainty, not certainty about outcomes. No political candidate or party sponsors these numerical calculations.
