Use cases & endpoints
Survivorship-bias-free data — delisted names retained
Survivorship bias is the quiet tax on backtests built from convenient data. If a price history only contains companies that still trade, every strategy tested on it was implicitly told in advance which companies would survive — and the ones that did not are exactly where losses would have come from.
The bias is strongest precisely where research is most interesting: small caps, distressed names, and any period containing a crisis. A screen that looks brilliant over a decade can be mostly an artifact of the delisted names that were never in the sample.
Hindcast retains delisted and renamed securities with their final observations and the reason they left, and makes universe membership queryable as of a past date. A screen run over an old period sees the securities that were listed then, including the ones that later failed — which is the only way for the result to describe a decision that could have been made at the time.
hindcast — session
previewhindcast> which securities were in the universe on a past date?
→ tool: universe { market:"...", as_of:"..." }
✓ listed members on that date, including later-delisted names
source: listing and delisting records retained in full
What's covered
- Delisted securities retained with final prices and delisting reason
- Universe membership queryable as of any past date
- Renamed and re-tickered securities linked to one entity
- Corporate actions kept as events for reproducible adjustment
- Recycled symbols disambiguated by date
- Applies to holdings and fundamentals joins, not only prices
Join the waitlist — this page's dataset is already selected.
Questions
How much does survivorship bias actually change results?
Enough to reverse conclusions in the strategies most affected, and the effect is largest exactly where the data is hardest to get. The honest framing is that the size of the bias is unknown for any given study until the delisted names are put back in.
Does this matter outside equity backtests?
Yes. Fund datasets have the same problem — closed funds disappear from vendor snapshots — so holdings and performance research needs the same treatment to avoid measuring only the survivors.
Related
Join the waitlist. Vote the roadmap. First in, first served.