The Coverage Board

What we actually have · how wrong are we → · board.json →

Every number here carries its absolute count and the denominator it was divided by. A bare percentage hides which question was asked, and on this data the question is the whole argument.

Measured at 2026-07-29 20:02:13 UTC. Some counters are still moving: 532 cards were sales-stamped in the last three hours while the drain lane runs. The figures below are that snapshot, not a settled total.

Displayable cards

64,464

Sets

389

Carry a value

58,800

of 64,464

Raw tier

25,457

of 64,464 · 39.49%

Coverage

Measure	Have	Of	%
Card images a front scan on the displayable card	25,345	64,464	39.32%
PSA publishes a population PSA has graded this exact card+grade and publishes a population for it. NOT the sales denominator — see below	38,994	64,464	60.49%
PSA pid mapped a PSA pid exists, so PSA can be asked about this card's sales at all	35,825	64,464	55.57%

Sales checked

This is the number most often quoted the wrong way, so it gets its own section. PSA's sales endpoints are keyed on a pid. A card with no pid cannot be asked about at all — it is unaskable, not "unchecked" — and folding those cards into the denominator measures the wrong thing.

The denominator, named

We divide by the pid-mapped cards: EXISTS (SELECT 1 FROM gpk_pid_card_map m WHERE m.card_id = specs.card_key) — the PSA APR / sales endpoints are keyed on a pid, so a pid is what makes a card sales-answerable at all.

We deliberately do not divide by the 38,994 cards matching specs.psa_pop_by_grade ~ '[0-9]' (pop-record): a published population proves PSA graded the card, not that its sales can be queried; using it would fold in cards we cannot ask about, which is what the framing rule exists to exclude.

Sales checked — share of the cards PSA can actually answer for (pid-mapped)

31,508 of 35,825 · 87.95%

Separate bucket — No PSA pid — PSA can never answer for these

28,639 of 64,464 · 44.43%

these cards are NOT in the denominator above. With no pid there is no PSA sales endpoint to call, so they are unaskable rather than unchecked.

Checked anyway, with no pid to ask on	2,347
Sales-stamped from any source, whole displayable universe	33,855

How NOT to say it

“48.88% of cards have had their sales checked” — 31,508 of 64,464. That figure is arithmetically true and analytically worthless: shown only to name the mistake: dividing by every card measures the wrong thing, because 28,639 cards have no PSA pid at all.

And the trap that is easier to fall into: the subtler error is dividing by the 38,994 cards PSA publishes a population for. A population record is not sales-answerability: it says PSA graded the card, not that PSA can be queried for its sales. That denominator is larger than the 35,825 cards actually carrying a pid, so it understates how much of the askable universe has been checked.

Evidence tiers behind the published values

How much real evidence sits under each value. Full definitions →

Tier	What it means	Cards	Of	%
L1	Recent sales of this exact card	4,694	58,732	7.99%
L2a	Older sales of this card, market-adjusted	8,138	58,732	13.86%
L2b	Sales of this card at another grade or variety	7,376	58,732	12.56%
L3	Model estimate — no direct comp	38,524	58,732	65.59%

Accuracy ledger

75,464 predictions frozen, 2 resolved.

The accuracy ledger in full →

Every figure on this page is read from a stored measurement taken at 2026-07-29 20:02:13 UTC and is reproducible from board.json. Values are estimates, not an appraisal.