A backtest can only test the data you feed it, and what you feed it is usually the set of coins that still exist. That sounds harmless until you notice what is missing: every pair that was delisted, abandoned, or quietly went to zero is absent from the history. Testing only the survivors quietly stacks the deck, and the inflated result that follows is one of the harder biases to see because nothing in the report looks wrong.
What survivorship bias is
Survivorship bias is the error of drawing conclusions from the things that made it through while ignoring the things that did not. In trading, it means measuring a strategy against the coins that survived to today, as if those were a fair sample of what was tradable at the time. They are not — they are the winners of a selection you did not perform deliberately.
Why crypto is especially exposed
The crypto market has a long graveyard. Pairs are listed and delisted, tokens lose all liquidity, and projects vanish. A naive history that only includes coins currently on the exchange has already filtered out the failures. A strategy backtested on that history is implicitly assuming it would have avoided every coin that died — an assumption it had no way to make in real time.
How it inflates a result
Suppose your strategy buys momentum across many pairs. On a survivor-only dataset, the catastrophic losers are missing, so the average outcome looks far better than it would have been live. The equity curve climbs more smoothly than reality allowed. The bias does not announce itself; it simply removes the worst evidence before you ever see it.
Testing against it
The defence is to test on data that includes delisted and dead pairs, or to be explicit that your universe is survivor-only and discount the result accordingly. Pair that with out-of-sample validation so a strategy must hold on data it never tuned against. A backtest that quietly excludes the failures is closer to a story than a test.
Survivorship bias sits alongside look-ahead bias in the family of errors that make a strategy look ready before it is. For the wider pattern, read backtesting myths that cost money. A clean curve earns scrutiny, not trust.