Survivorship bias
When we study only the assets still around today, what is the sample quietly leaving out?
In plain words
Survivorship bias is the error of drawing conclusions from a sample that contains only the things that survived. Markets, funds and exchanges that closed, collapsed or were delisted vanish from the record, so the average that remains describes a group defined partly by having made it.
Why it matters
Risk studies built on survivors understate how badly things can go. Returns look steadier, drawdowns look shallower and diversification looks more reliable than the full history would suggest. A strategy tested only on assets still trading today may never have faced the conditions that removed the others from the sample.
In a real week: ETH, week of 11 February 2019
In the week of 11 February 2019, ETH, one of the assets still trading, rose 6.9% to close at 133.75, yet remained 86.25% below its 52-week high. Its dollar volume was 0.3bn, 0.71 times its prior 12-week average, and its sample history covered 52 weeks. Assets that had already left the market by then contribute nothing to averages computed from this sample.
A common mistake
Treating a survivor's long price history as a complete record. A long history in the data means the asset is still listed; it says nothing about the assets no longer there.