Survivorship Bias
Also known as: survivor bias, survival bias, surviving-sample bias
What is it?
Survivorship bias is the error of measuring only the things that are still around, then treating that filtered sample as if it were the whole population. It is built into almost every performance list a trader reads. A copy-trading leaderboard ranks the accounts still trading; the ones that blew up were delisted and are not in the average. A stock index contains today's members, so a strategy backtested on it never had the chance to buy a company that went to zero.
38 have already closed and been delisted
The only ones you ever see
The 12 may be genuinely good. The point is that their average was computed after 88 were removed - illustrative figures.
Suppose 100 signal services launch, 62 are still running at six months, 31 at a year, and 12 make the leaderboard you are looking at. The 68% average win rate on that page is the average of twelve survivors, computed from a starting cohort of one hundred. The distinction that matters is between the average of the survivors and your expected outcome from joining. Those are different numbers, and only the first one is ever published.
Survivorship bias is not a claim that the twelve are frauds - they may be genuinely good. It is that the denominator which would tell you your odds has been quietly deleted before you were shown the figure.
Why it matters: Leaderboards and strategy lists only display what survived, so the average return on the page is not the return you can expect from joining.
It inflates every published track record and backtest that draws on a list of current members, and it never shows up as an error in the numbers themselves.
Real-world example
A copy-trading leaderboard showing 40 profitable strategies looked like strong odds until the platform's own archive revealed that 300 more had been started and closed over the same period.
How SignalBots handles it
SignalBots reports a strategy's full result history rather than only its winning stretch, so a track record is read against the trades that went wrong as well as the ones that worked. See /risk-warning.
Pro tip
Before trusting any ranking, ask how many entries were removed from it. The denominator sets your odds and it is almost never printed next to the average.
Common pitfalls
Backtesting a stock strategy against today's index members, which silently deletes every company that was delisted or went to zero.
Frequently asked questions
How does survivorship bias show up in a backtest?
Most often through the instrument list. If the universe is built from assets that exist today, the test never gets the chance to buy the ones that failed, so returns come out higher than anything that was actually achievable at the time.
Can I correct for survivorship bias?
You can reduce it by using a point-in-time dataset that includes delisted instruments, and by asking any platform for the full count of accounts opened rather than the ranked list. Neither removes the uncertainty, and your capital remains at risk.
Is a long track record proof against it?
It helps, because surviving many years is harder than surviving a quarter. It is still a survivor's record though, and it says nothing about how many similar traders started at the same time and stopped.
How is this different from cherry-picking?
Cherry-picking is a deliberate choice to show favourable results. Survivorship bias needs no intent at all - the selection is done by failure itself, which removes the losers before anyone compiles the list.
Does it affect prop-firm pass rates too?
Yes, in the same way. Marketing tends to feature funded traders, who are the small share that cleared the evaluation, while the accounts that breached a rule and closed are not represented in the figures shown.
Trading involves substantial risk of loss. Historical and backtested results do not guarantee future performance. Read the full risk warning.
