Corrections
We aim to be accurate, and when we fall short we want to fix it openly. If you spot an error — a wrong number, a misstated result, an out-of-date race — please tell us through contact. A link to a primary source helps us verify quickly.
We correct material errors promptly and record them below with the date. Routine data updates (new polls, scheduled refreshes) are part of normal operation and are not logged here; this page is for corrections to things we previously got wrong. If a race we showed as decided is later un-called or retracted by its sources, we revert it on the site and log the reversal here.
Log
- May 25, 2026
Corrected the Texas U.S. Senate race summary. It previously stated the incumbent was "not expected to be competitive" with "a commanding structural advantage," which our own polling contradicted (a leading primary challenger and a close early general-election poll). The summary now describes the contested primary and competitive polling without handicapping the outcome.
- May 25, 2026
Updated race metadata for roughly 20 races after an audit found several primary dates out of date, the listed incumbents for the Nebraska, West Virginia, and Oklahoma Senate races incorrect, and Louisiana modeled under its former jungle-primary system. The fields and the affected race summaries were corrected.
- May 25, 2026
Rewrote the polling-methodology page. It previously described weighting parameters that did not match the running code and referenced a source file that did not exist; it now documents the actual implementation.