Our sources & editorial method
Public facts, explicit boundaries, and a score that is honest about being an estimate.
Sources for the launch bank
- PubChem: element names, symbols, atomic numbers, and recorded properties.
- Nobel Prize API: individual laureates and their award categories and years, through 2025.
- Wikidata (CC0): film and constellation records. The film collection uses earliest recorded release years and Best Picture nomination or award statements.
- mledoze/countries (ODbL): country, capital, language, currency, and border records. Our game uses UN member entries. The adapted country dataset is available under the same license at country-facts.json, with its license.
- nba_api (MIT) for NBA factual identities, and official NFL, MLB, and NHL sources: factual team names and groupings in the dated launch snapshot. No logos, photography, articles, or highlight footage are used.
How the bank is built
Source snapshots are normalized into small factual records. Each question defines a filter over those records and stores accepted labels, reviewed aliases, and citations. Automated checks verify membership, boundaries, duplicate names, score values, and schedule coverage.
Familiarity tiers are editorial heuristics, not empirical response rates. They combine deliberately selected familiar examples with deterministic ordering in the remaining source set. This produces reproducible game scores, not a claim about a global population.
Known limits
Source records can be incomplete or mistaken. Film genre labels may overlap; release dates may refer to a premiere rather than a local release. Geography classifications follow the named snapshot. Team names follow the season in the prompt. Report a precise issue and source so we can correct future versions.
Our sources do not endorse this game. Gameplay answer lists and scores are kept on the server; the country fact extract is published separately to meet its data-license obligations.