Methodology

Election Index answers four questions per election: when is it, who counts it, when will each count exist, and which page proves it. It says "not yet" rather than guess.

Count stages

Each race has a sequence of stage windows: exit-poll, election-night, media-call, intermediate-count, final-result, court-validation, government. Each window has an earliest and latest expected time, and a note on which later count replaces it on the authority's live page and when. Windows err late: a question about the final result is only safe after the final-result window closes.

Confidence

announced stated by the authority or the law · precedent derived from the previous election · estimate our estimate. Every window carries its citation.

Expected vs observed

The stage shown as "now" is expected from the calendar. When the Watcher sees a count appear, the stage records observedAt. Until then the API reports observed: false.

Sources and the homepage rule

Every URL is classified by rule: race results, feed, results index, authority homepage, directory, media. Only race results pages and feeds can answer a results question. The rule exists because homepages change: a page that shows one election today will show the next one later.

Source tiers

T1 official final (the certifying authority) · T2 official provisional · T3 reputable secondary (wire services; cites T1/T2) · T4 community (Wikidata, Wikipedia): cross-check only, never the sole source.

How the site runs

A crew of agents does the work. The Scout discovers elections (Wikidata, Wikipedia calendar, observer missions). The Researcher profiles each country's legal deadlines. The Verifier re-checks every claim independently. The Watcher observes counts. The Auditor reports health daily. Nothing is published except through a deterministic gate that checks schema, citations, ordering, sources and licences. People handle only what the agents cannot settle.

Neutrality

No forecasts, no odds, no opinions. Contested results are shown as contested, with the official result and observer findings attributed. We report who certified what, never who "really" won.