Survivorship Bias
A distortion that appears when failed or delisted assets drop out of a dataset, making the ones that remain look representative when they are not.
Any ranking, index, or historical study built from the assets that exist today silently excludes the many that went to zero, were delisted, or whose networks stopped producing blocks. The averages that result describe survivors rather than the population an observer faced at the time. Digital assets are unusually exposed to this, because failure is frequent and delisted assets often disappear from provider histories entirely rather than being retained with a terminal value. The correction is a point-in-time universe: reconstruct the set of assets that existed and met the criteria on each historical date, keep every one of them in the sample regardless of what happened later, and record the date each left.
In der Praxis
A study of the average performance of top-ranked assets is distorted unless assets that later collapsed, such as the Terra ecosystem tokens in 2022, remain in the historical universe at their historical rank.
Das häufige Missverständnis
Survivorship bias is not fixed by using a longer history; a longer record of survivors alone is more misleading, not less.