Forecast model
MAE is the average absolute district-margin error; lower is better. Only two holdouts occur after 2016.
Choose a district on the map or with the district finder.
District forecast table
. Margins, probabilities, intervals, and ratings use the model selected above.
| Candidates | Vs. Basic | 80% interval |
|---|
Data sources and freshness
Observed, modeled, missing, and imputed values are distinguished in district details. Supporting data remain downloadable.
How to read this forecast
Basic is the default guardrail: the poll-adjusted presidential baseline plus 20% of CMO expected performance.
Fundamentals+ uses the same inputs and applies the full CMO expected-performance adjustment. It is more responsive to district context and less conservative than Basic.
District probabilities use a zero-centered six-point normal calibration, selected after forward-cycle and leave-state-out tests on 1,188 recent Southern legislative races. Basic and Fundamentals+ use the same calibration applied to their respective margins.
Seat distributions are exact independent-seat calculations conditional on the displayed national environment. A different national environment should be viewed as a separate shared scenario, not hidden inside every district's error band.
Districts with only one major-party nominee are treated as fixed Democratic or Republican seats in chamber summaries. Independent-only and otherwise unresolved districts remain unmodeled and are displayed in gray.
The national-environment projection uses polling through the date shown above and should be refreshed as new polling becomes available.