Experimental v1.0
MT senate 2026
Election date 2026-11-03. Forecast leans R at 83.6%.
lr-2026-06-10-chal-recal-twospeedpartial fundraising data The candidate-fundraising input for this race is mid-cycle receipts through 2026-06-11 (the latest FEC filings, pre-primary). It refreshes on the quarterly FEC filing calendar and the share can shift as new candidates file.
The divergence, explained
Every race is a structural anchor, a behavioral signal, and a national thermostat. The anchor is the state lean, incumbency, and personal-vote alone; the behavioral block (the 10 mood dimensions) moves it from there; the thermostat is the deep-estimated midterm swing against the president's party.
Why this number? interactive
The real model is a base rate, then structure, mood, and the midterm thermostat, then a calibration step. Toggle any factor to see the model's true response. This changes only this explorer, not the published forecast.
Open this race in the interactive explorer to move the model's levers and watch the win probability respond.
Top dimension drivers
| Dimension | Direction | Contribution to log-odds | Standardized value |
|---|---|---|---|
| CivicTrust | toward D | +0.313 | 0.888 |
| challenger_d_share_c | toward R | -0.158 | -0.480 |
| Economy | toward R | -0.152 | 0.398 |
Methodology
The forecast applies a logistic regression trained on prior cycle outcomes (state-lean + standardized dimension scores) to the state's current dim snapshot. See /methodology for the model spec and /calibration for the live calibration track record. This is the public summary; the authenticated deep dive lives at /app/race/senate-2026-MT.