Survivorship bias in calls, rankings and research

An account can appear consistently successful after deleting failed predictions or discussing only tokens that remained active. A leaderboard can exclude closed accounts. A token study can analyse only assets meeting a performance threshold. Each selection changes the conclusion the data can support.

Define the population first

Write down which calls, accounts or tokens should have been included before looking at their outcomes. Record the period, entry rules, exit rules and treatment of missing data. Keep inactive and failed examples when they belong to that population.

MASTR's July essay explicitly cautions against interpreting a manipulation indicator found within a selected high-performing sample as the fraud rate of all memecoins. The denominator is part of the claim.

For a promoter's record, reconstruct all recommendations over a fixed period rather than accepting a collection of winners. For copy trading, include the trades a follower missed and the transactions that failed. A result may still be impressive after these corrections; the point is to measure the result that was actually promised.

来源

1. MASTR: copy-trading research snapshot, 9 August 2026
2. MASTR: Crypto Survival Guide, four original panels
3. A Midsummer Meme's Dream: research on memecoin manipulation indicators

资料核对日期：2026年9月5日. Historical cases retain the date and legal status of the cited record.