Olympic Explorer
Methodology & Sources
The Olympic Explorer is a data layer, separate from our editorial stories. This page documents how medals are counted, where the numbers come from, and how the dataset stays honest.
Source hierarchy
| Tier | Source | Role |
|---|---|---|
| Arbitration | olympics.com | Official control totals; manual cross-check only. |
| Bulk | Wikidata (SPARQL, CC0) | The only automated batch pipeline for results. |
| Cross-check | marcolympics.org | Full results 1896+, fast DQ updates. Requires citation. |
| Cross-check | Kaggle "120 years" | Only ≤ 2016 Rio. Excludes Tokyo/Beijing/Paris/Milano + retests. |
| Editorial | zgaysj.info | Chinese context, local mirror. No structured data, no automated use. |
Counting rules
- Conflict order: olympics.com > marcolympics.org > Wikidata. Differences are logged.
- Team events: counted once for the national total; every member receives individual credit.
- Medal status: awarded / stripped / upgraded. The current effective color is shown (e.g. Beijing 2008 = 48 gold post-DQ).
- Scope: Summer from Los Angeles 1984; Winter from Lake Placid 1980; Helsinki 1952 is participation-only.
- NOCs: Hong Kong (HKG) and Chinese Taipei (TPE) are separate and excluded from China's totals.
- Exhibitions: excluded. Summer/Winter are split; Combined is the sum.
- IDs: pinyin slugs. Milano Cortina 2026 is included.
- Provenance: every result carries a source
{ name, url, retrieved_at }.
Update protocol
Results are fetched via a reconciliation script that compares the draft totals against
hand-verified control totals from olympics.com. A human reviews the reconciliation report
before the draft is promoted. Every change is recorded in a CHANGELOG with a
last_reviewed date.