What Is Survivorship Bias? The Sneakiest Backtest Error

Survivorship bias is the distortion caused by analysing only the ones that "survived".

An example

Imagine testing a strategy on five years of data for the stocks that trade today. That list does not contain companies that went bankrupt, were delisted or were acquired in that period. The strategy may have lost money on them, but the test never sees it. The result looks better than reality.

Why it matters

The effect is largest for strategies such as "buy the dip": stocks that fell and then failed are not in the test, so the strategy looks much safer than it really is.

How to reduce it

• Where possible use point-in-time lists: data that contains the stocks that traded on each date.
• Avoid applying one "today's list" to the whole past.
• Treat results as optimistic and interpret them with caution; work knowing the gap exists.
• Paper trading does not have this bias; live data uses that day's real universe.

Be honest

If your data has this bias, say so next to the results. For the general method see the Backtesting and Paper Trading Guide; for a Borsa Istanbul example see backtesting on BIST.

This article is for education only and is not investment advice. Past results do not guarantee future results; trading carries a risk of loss.