Glossary

Survivorship bias

The error of studying only the companies that still exist, which leaves out the ones that failed.

If a backtest uses today's list of companies and looks back in time, it never meets the companies that went bankrupt, were delisted or were taken over at a low price. Their losses vanish from the results, so the past looks better than it was.

The fix is a dataset that includes every company that existed on each past date, with each failed company kept in until the day it stopped trading.

When you read a backtest, ask whether delisted and bankrupt companies are included.

  • Backtest: A test of an investment rule on past data, to see how it would have done.
  • Look-ahead bias: Using information in a test that was not yet available on the date of the decision.

All terms →