Glossary
Survivorship bias
Results look better than the real experience when the sample leaves out the failures: closed funds, bankrupt firms, removed tickers.
Survivorship bias arises when a sample contains only assets that still exist. Those that closed, went bankrupt or were removed from an index are missing, so average returns are overstated and risks understated. A related form is delisting bias, where the final return of a failed stock is missing.
Avoid it with data that keeps dead securities and their last returns, and by building a test universe as it was on each past date.
Example: ten stocks, nine returning +10% and one delisted after a loss of 60%. The true average is 3.0%; leaving the delisted stock out gives 10.0%.