A bot detector that flags a fixed share of every batch is not counting bots. It's grading on a curve. That's the argument in the latest @zerufinance research post. In one campaign cohort of 1,981 wallets, a relative-rank detector flagged 618 as hard bots, about 31%. The median flagged wallet had 1,886 transactions. The "bots" were the most active part of the group. A hand check of a sample found three of five genuinely automated. The other two were a power user and an airdrop farmer, both human. The interesting part is the fix. zScore scores a wallet from its own onchain history against a fixed reference, so the same wallet gets the same score in a cohort of 2,000 or 2 million. Penalties are graduated, so a false positive costs part of a payout, not all of it. Next time someone posts what share of an airdrop was bots, ask what that detector says about a group nobody suspects.