Why our election results page is still pending
Why our election results page is still pending TL;DR: We use official sources only. State election offices publish preliminary, uncertified results. The FEC and the Clerk of the House publish final totals after certifi
Why our election results page is still pending
TL;DR: We use official sources only. State election offices publish preliminary, uncertified results. The FEC and the Clerk of the House publish final totals after certification. We do not use AP, DDHQ or media race calls. The results page stays pending until the source is named and the numbers are checkable.
Results and sources: elections.lforla.org/results.html.
A race call is not a count
On election night, the first numbers are projections. They are useful, often accurate, and not the same thing as a count. A called race can be reversed by late mail ballots, provisional ballots, recounts or certification disputes.
That does not make media calls bad. It makes them a different data source from an official canvass. If the goal is to score a forecast against the result, the source matters.
The benchmark was sealed before the vote. The result should be at least as stable as the input.
What we use instead
We use four official sources, in this order.
- State election offices publish precinct and county results, usually on election night. These are preliminary and uncertified. We label them as such.
- The FEC publishes federal election results and voting information after the fact: fec.gov.
- The Clerk of the House publishes official House vote counts: clerk.house.gov.
- The EAC documents the canvass and certification process: eac.gov.
The state election office directory is maintained by USA.gov: usa.gov/state-election-office.
No AP. No DDHQ. No cable-network calls. If we cannot name the source, we do not fill the cell.
What the results page will show
When the data lands, the page will publish:
- Control of the House and the Senate, with the seat totals and the source.
- The certification status of each chamber: preliminary and uncertified, or certified.
- Winner accuracy for each model: did the call match the result?
- The Brier score for each probabilistic call, computed from the implied Democratic-control probability.
- A calibration note, with the sample-size caveat. Eighteen models and two chambers are not a calibration study.
The forecasts and the results will sit side by side. The quality score that was sealed before the vote does not change. The accuracy layer is added, not substituted.
Why wait
Certification takes days. In close races it can take weeks, and recounts can take longer. A benchmark that publishes a winner on election night and scores it the next morning is measuring the projection, not the election.
There is also a subtler reason: the scoring rule should not move after the result. If we switched to media calls because they were faster, we would be choosing the source after seeing the outcome. That is the same mistake as editing a forecast after the vote.
The page stays pending until the source is named and the numbers are checkable. That is slower. It is also the only version of the exercise that means anything.
What you can read now
The snapshot is sealed with SHA-256 26c1563ea12200b9436e9ab7fd03b1af7f59fd17aea5697a2bd96a25bd38b2d9 and tagged elections-forecasts-2026-10-31.
Originally published at elections.lforla.org.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.