data

• Bulk files, rebuilt nightly as gzipped JSON lines: markets-open (every market open for trading, with price, 24h move and volume); events-open (every open event, with its market count and volume); resolved (every event pdata has pruned, with its final result). The first line of each file names the licence and every field's unit; sizes, row counts and checksums are in index.json, and the last 7 nights are kept at api.pdata.world/bulk/.
• Monthly snapshot of the same three files as Parquet, each tagged with its date, on Hugging Face and Kaggle, and archived on Zenodo with a DOI for citing, 10.5281/zenodo.23211610.
• JSON API — no auth: https://api.pdata.world/api/v1/events (and /markets, /news/movers). Full spec at openapi.json.
• Price history since May 2026 — probability and 24-hour volume over time for an event's top markets: https://api.pdata.world/api/v1/events/{source}/{id}/history, with range 24h, 7d, 30d or all and an optional market_id (a Kalshi market ticker in place of the event id redirects to its event). The same snapshots are the public.market_snapshots and public.event_snapshots tables in the Postgres below. Points older than 7 days are kept as 12-hour buckets, and an event's history is deleted 10 days after it closes. After that its API record holds only the final results (kind: "tombstone"), which the resolved bulk file also keeps.
• MCP server for AI agents: https://api.pdata.world/mcp — see the agent guide.
• Python client, standard library only, with examples and sample files: github.com/noskillcoding/pdata-client.
• Public read-only Postgres (query it directly):
  postgresql://pdata_readonly:pdata_public_2026@pg.pdata.world:6433/pdata_new

pdata, “Cross-platform prediction-market data,” pdata.world, accessed 2026. https://pdata.world/data

For a fixed snapshot, cite the Zenodo record: doi.org/10.5281/zenodo.23211610 (each monthly version has its own DOI as well).

Licensed CC BY 4.0 — reuse freely with attribution.