The traders who post results online are the ones who did well; the ones who blew up are silent. A stock screener that only includes currently listed companies excludes everything that went bankrupt, inflating backtesting results.
It is also why a strategy that worked for one famous trader is weak evidence: thousands used similar strategies and were never heard from.
Example: a backtest on today's S&P 500 members from 2005 shows 14% annual returns. Including the companies that were removed from the index over that period drops it to 9%.
Related: backtesting, sample-size, overconfidence, hindsight-bias