Methodology
How every verdict is computed
Every card resolves to one call — grade it, chase it, keep it raw, or wait for data. Here is the model behind the call, and the point where we stop and say we don’t know.
A verdict is a comparison, not an opinion. On one side is the value a card is likely to carry once it is graded. On the other is everything it costs to get there. When the first clears the second at the real odds, we say grade it. When it doesn’t, we say keep it raw. Nothing here is best-case.
The comparison
The graded side is what each realistic outcome actually sells for — the PSA 10 price, the PSA 9 price, each with its own recent sample. The cost side is everything it takes to get there: the raw price you could take today, plus the grading fee, plus the shipping and selling costs on the slab. We compare the two. A card earns a positive verdict only when the graded outcomes clear the all-in cost with enough room that even the miss doesn't hurt — not when the top-grade price alone looks large. We don't pretend to know the odds of any single grade; we show what each outcome pays and where break-even sits, and the verdict follows from whether the downside is covered.
The deciding inputs: the ladder and the cost
Two observable numbers do most of the separating between cards with identical top-grade prices and opposite verdicts. The first is the multiple — what the PSA 10 sells for relative to raw, the size of the prize. The second is the spread at PSA 9 — whether the most common miss still clears break-even (raw plus the grading fee), or falls under it. A big multiple with a covered downside reads worth grading; a big multiple that collapses the moment you miss reads chase; a thin premium reads keep raw. Every input is a recent sold price you can check yourself.
The four verdicts
Every card resolves to one of four calls. The thresholds come from the comparison above, not from feel.
The expected graded value clears the all-in cost with a comfortable margin, and the grade ladder behaves — the top grade commands a real premium and the lower grades hold enough value to protect the downside. The odds, not just the ceiling, are on your side.
The expected value is positive, but it rides on hitting the top grade. Miss it and the PSA 9 and below barely clear the cost. These are cards where your own copy’s condition tips the decision — a clean copy is a real chase; a played one is not.
The expected graded value does not clear the cost, or the premium between raw and graded is too thin to justify the fee and turnaround. The card is worth more in your hands, sold as-is, than it is worth gambling on a grade. Keeping a card raw is a real answer, not a consolation prize.
When there aren’t enough recent graded sales to trust the comps, we don’t issue a verdict. A number built on one or two stale sales is worse than no number, because it looks like certainty and isn’t. We hold the call until the data supports one.
The thin-market rule
Most cards on the site trade thinly. A single optimistic sale can drag an average somewhere it shouldn’t go, and a card with two graded sales in six months has no trustworthy comp at all. So we set a floor: below a minimum count of recent graded sales, the card shows “not enough data” and carries no estimate. It is the least satisfying answer and the most important one — the standard that keeps the other three verdicts worth trusting.
Every verdict is computed before any buy link is placed. The call never depends on what happens to be for sale — the data decides first, and the links come after.
The data behind it
Raw prices are eBay’s ungraded sold price, not TCGplayer’s market price. Graded figures are recent sold averages from eBay, counted per grade so the sample size travels with the number. Everything is in US dollars. Grading cost is modeled as a flat $25 — PSA’s Value Bulk fee per card, the cheapest real path since PSA’s 2026-02-10 rate revision, fee only (shipping and membership aren’t included). PSA temporarily paused Value-tier submissions effective 2026-06-02, with reopening tied to reducing its grading backlog to 5 million cards (approximately 14 million as of late July 2026; PSA’s own projection is 5–6 months). The $25 figure is the published Value Bulk fee and is retained as the cost basis; the lowest tier currently accepting submissions is Regular at $79.99. Refresh cadence isn’t uniform — some signals on the site update daily, but the core sold-price figures behind a verdict update on a slower cycle right now. Where a card’s own condition would move it up or down the grade ladder, that is the one variable the model can’t see — the method gives you the market’s answer; your eyes give you the input. Card images are provided by TCGdex (tcgdex.dev).
The raw figure at the top of a card page is a 90-day trailing median, not an all-time average — the ladder further down keeps the all-time figure for grade-to-grade comparison, labeled separately so the two are never mistaken for each other. Below five recent sales it’s marked thin sample; at five or more, if the sales themselves disagree too widely to trust the middle one (a relative spread over 0.65, by the same robust measure used to keep a single wild sale from swinging a median), it’s marked low confidence. Neither marker withholds the number — both just say how much weight to put on it.