The Vugol Me dataset

• Candidates. Federal candidates from FEC filings (1980 onward), plus state and local candidates from public ballot records. Each profile shows which source every fact came from. Anything a campaign adds is labeled campaign-provided.
• Races. Every 2026 Senate, House and Governor race and other statewide races, with rosters, primary and runoff results, FEC money totals, election dates and our own rating.
• Polls. Individual third-party polls with pollster, verification tier, field dates, population, sample size, computed margin of error and toplines. Each poll links to the pollster’s own release. Only verified polls enter an average.
• Releases. Immutable, checksummed snapshots of each race’s polling average. Each one stores the input poll IDs, the weights and the model version.
• Not included: Vugol Race Polls. These are self-selected and are never part of the dataset or any average.

• Race polls, results, money and dates: a nightly job, with fixed parsers rather than a language model (sources and rules).
• Poll feed and API: published polls appear as soon as they are approved. API responses are cached at the edge for 60 seconds.
• Releases: issued by staff when a race’s inputs change. A correction publishes a new, superseding release and never edits the old one.

Vugol Me releases and tables are CC BY 4.0. Government records are public domain. Poll toplines belong to their pollsters. Wikipedia-derived text is CC BY-SA 4.0. The data license covers each source class and how to attribute it.

For bulk access, higher rate limits, custom feeds or redistribution in a commercial product, contact us about API and data licensing. Licensees receive the same data on the same terms as everyone else. No license buys influence over a rating, roster or release.

GET /api/public/polls needs no key. It is read-only and returns the same data as Verified Polls. Filters: office, state, pollster, contestKey, format (json or csv), limit (max 500). Limit: 30 requests per minute per IP. Full reference: API docs.

Live sample: the newest verified poll, fetched when this page was built. The options list is truncated to two entries.

{
  "studyId": "f4bdf32c-dd3c-4584-9abb-4fb7b3c57722",
  "contestKey": "US-KS:hyp",
  "pollsterName": "InsiderAdvantage",
  "pollsterTier": "V",
  "sourceUrl": "https://insideradvantage.com/insideradvantage-senate-surveys-texas-paxton-49-talarico-48-kansas-marshall-48-hamilton-47-maine-jackson-49-collins-47/",
  "fieldStart": "2026-10-13T00:00:00+00:00",
  "fieldEnd": "2026-10-14T00:00:00+00:00",
  "population": "likely_voters",
  "sampleSize": 1200,
  "moePct": 2.8,
  "options": [
    {
      "label": "Roger Marshall (R)",
      "canonicalKey": "fec:s0ks00315",
      "value": 48
    },
    {
      "label": "Adam Hamilton (D)",
      "canonicalKey": "fec:s6ks00312",
      "value": 47
    }
  ]
}

Each release carries a SHA-256 over the canonical JSON of its body (keys sorted recursively, no whitespace). The body holds the inputs, weights, model version and outputs. Changing a single byte changes the checksum.

1. Download a manifest: curl -sSL https://vugol.me/releases/2/manifest.json -o release.json
2. Recompute: in Node 22+, verifyManifest(JSON.parse(text)) from @vugolme/core returns true only when sha256(canonicalize(body)) equals the stored checksum.
3. Compare with the checksum printed on the release page, e.g. release #2.

The checksum shows a release has not changed since it was published. It does not mean an independent party has audited it; none has yet (our definitions).