How strongly the sales behind a figure agree with each other. It is printed as A, B or C next to every market comparable, and it is about the EVIDENCE, not about the card: a common card that trades every other day can carry an A, and a rare one that sells twice a year cannot.
A means five sale-days, all inside 60 days of each other, all within 30% of one another, the newest no older than 14 days. B is wider, older or more spread. C is everything else we are still willing to publish — and a card whose sales are thinner than C gets no figure at all, only the reason why.
The grade is what turns the comparable into the FMV: an A sample pays 100% of it, a B 90% and a C 80%. That is not a deduction applied to your card; it is the difference between a figure five tight recent sale-days agree on and a figure three scattered ones suggest. How well a card sells is part of what it is worth.
We measure this rather than assert it. We take cards with enough history, hide each one’s most recent sale, price it from everything older, and compare our figure with the sale that actually happened. On A samples our figure lands closer to that sale than on C, and the case that matters more — the bad miss, where the next real sale comes in a fifth or more below the comparable — is rarest on A and most common on C. The rates answer that gap.
Two honest limits. Those are historical averages across many cards, not a prediction about yours — the market can move against any figure, which is why the buyback rate is fixed and the amount is not. And an A carries no cushion at all, because it is paid the full comparable; the cushion exists where the evidence is thin, which is the whole of the idea.