Datasets
Market data for AI agents — EOD prices & deep history
Most price history quietly excludes the companies that failed. Vendors drop delisted tickers, recycle symbols, and adjust for splits in ways that are hard to reverse — and a universe built from what still trades today will make almost any strategy look better than it was. The cheapest data is usually the most expensive mistake in a backtest.
Hindcast is building an end-of-day layer where delisted names stay in the record with their final prices and the reason they left, symbols are tied to the entity graph rather than to the string on the exchange, and corporate actions are stored as events so both adjusted and unadjusted series can be reconstructed rather than assumed.
This is deliberately not a tick-data product. It is the history layer an agent needs to answer questions about the past accurately, alongside the fundamentals, filings and macro series it is already querying — one endpoint, one identifier space, one set of citations.
hindcast — session
previewhindcast> daily closes for a ticker in 2015, unadjusted, with actions listed
→ tool: eod_prices { ticker:"XYZ", start:"2015-01-01", end:"2015-12-31", adjust:false }
✓ daily OHLCV plus the corporate action events in the window
source: exchange end-of-day records with action history retained
What's covered
- Equities, ETFs and futures — daily open, high, low, close and volume
- Delisted and renamed securities retained with their final observations
- Corporate actions stored as events, so adjustments are reproducible
- Symbols resolved through the entity graph rather than raw ticker strings
- B3 and US venues on the same schema for cross-market queries
- Point-in-time universe reconstruction for any past date
Join the waitlist — this page's dataset is already selected.
Questions
Is intraday or tick data included?
Not at the start. The focus is deep, clean daily history because that is what agents and research backtests actually consume. If intraday matters to your use case, the vote on this page is how that gets prioritized.
How is survivorship bias avoided?
By never deleting anything. Securities that delisted keep their rows, and universe membership is queryable as of a past date, so a screen run over 2011 sees the 2011 universe rather than today's survivors.
Related
Join the waitlist. Vote the roadmap. First in, first served.