LIMITATION  Data Integrity

Survivorship in Historical Small-Cap Data: A Permanent, Measured Limitation

Does our historical data source cover delisted small caps — and if not, how badly?

Survivorship bias — dead stocks vanishing from history — silently flatters every small-cap backtest. Instead of assuming our historical source was fine, we measured it: we built a roster of tickers our own live systems observed in real time, identified which ones subsequently died (delisted or vanished from the reference asset list), and checked how many still have historical bars.

Result: of 66 known-dead tickers seen live in the window, historical bars exist for 1 — a 1.52% coverage rate. The gap is permanent: the missing history cannot be recovered from this source.

What this means: our historical training data understates failure cases, and we say so. Forward live collection — recording stocks while they are alive — is the only mechanism that captures these names, and it is now how our datasets grow.

What this does NOT mean: that conclusions built on forward-collected data share this bias — the limitation applies to the historical era and is documented wherever that data is used.

← All findings · How we measure

This page describes a research verdict under our published methodology. It is not trading advice, and information claims are not profitability claims. Past results do not guarantee future outcomes.