Core terms
Survivorship bias
Distortion from studying only the securities, funds or traders that survived to today, leaving out those that were delisted, renamed, failed or stopped posting.
Brown, Goetzmann, Ibbotson and Ross (1992) showed how keeping only surviving funds can make performance look persistent when it is not. In small caps the effect is direct: stocks are delisted, renamed and reverse-split often, so a backtest universe built from today's active tickers quietly drops many of the worst outcomes. The FennEdge Gapper HOD study takes each security's type from reference data as of each historical session, which keeps later delisted or renamed tickers in the sample; it notes that this reduces the bias without removing every classification error.
The same filter shapes chart examples and track records: charts that worked get posted more often, and traders who had a bad run often stop posting, so the visible record skews toward lucky runs.
What traders should verify
Build test universes from point-in-time lists that include delisted and renamed symbols, and treat a feed of posted wins as a selected sample, not a base rate.
Do not conflate
Removing survivorship does not fix split adjustment: adjusted prices after a reverse split can place a stock above a price filter it did not meet on the day. Knowing the bias exists does not say how large it is in a given sample.
Primary source
Primary sourceBrown, Goetzmann, Ibbotson and Ross, 1992