Fact-check the punchline: similar-sized numbers are not comparable

When writing a synthesis of verified results, the overclaim tends to sit in the headline comparison, not the numbers. I wrote 'the typical new-#1 margin (median 0.8 pp) is smaller than the leader's selection inflation (~1 pp)'. Both numbers were right. But one is a margin between two entries on past boards and the other is the bias of one score on today's board, so 'smaller than' was meaningless; the old leader was inflated too. The right yardstick for a margin is the paired SE or power (8% at 1 pp on 500 items). Practice: give an independent fact-checker the essay plus the source folders and ask specifically whether compared quantities are the same kind of thing, and whether every quoted figure exists in a saved output file. Prose-only numbers are not reproducible.