Guide & rating key
How the Grading Outlook is worked out, in the same numbers the engine computes with. Every weight, anchor and threshold on this page is read out of the rating model when the page loads — change a weight and this page changes with it.
One card, all the way through
Every number below is what the engine returned for this card — no part of it is written down here.
Umbreon VMAX — Evolving Skies, Holofoil, asked at PSA 10, graded at the Super Express tier. Nothing below is typed: each figure is what the engine returns for this card, from the prices and sales recorded with it.
| Component | What the engine found | Rated |
|---|---|---|
| Profit potential | 27% return at PSA 10 | 40 / 100 |
| Grade premium | $916.31 added over selling it raw ($2,404.28) | 65 / 100 |
| Market depth | 162.0 sales/month at PSA | 100 / 100 |
| Downside | 29% lost if it grades 9 | 36 / 100 |
Weighted together, the Grading Outlook
59 / 100
Its 81 most recent PSA sales were counted once, on 2026-09-14, and cover 15 days. That is the whole PSA market for this printing at every grade, not the 10 alone.
What the outlook answers
What is this card worth at the grade you say it will get, what does getting there cost, and is the difference worth having?
You pick the grader and the grade. We price that exact market — the real sold prices at that grade, that grader, that variant — subtract the fee, the postage both ways and the marketplace’s cut, and tell you what is left.
What grade this card will get.
Not “probably a 9”, not a gem rate, not a confidence signal. Nothing here is trained on your photos or the population report, because a number that looks like odds gets spent like odds. You state the grade; we price it honestly. If the grade you picked loses money, it is shown losing money — the engine will not rescue a pessimistic input with an optimistic one, or the reverse.
Which is why a card with no grade set has no outlook. That is the finished state, not a gap.
The four components
Each is rated 0–100 on its own curve, then weighted, and the weighted contributions add up to the outlook. Any card’s breakdown shows you that same arithmetic, from the same computation.
Profit potential
35% of the outlook · ×0.35How hard the money you put in works.
Grade premium
20% of the outlook · ×0.2How much money is actually on the table.
Market depth
25% of the outlook · ×0.25Whether you can actually sell it once it comes back.
Downside
20% of the outlook · ×0.2What happens to the money if it grades lower than you said.
Each one has its own page: the curve it is rated on, where 524 measured cards land on that curve, and what happens when it cannot be read.
What the number turns into
One computation produces the outlook and the word, so they cannot disagree.
- Grade
- 60 and above. The money, the market and the fall-back all point the same way.
- Marginal
- 40 to 59. It works on paper and something is thin — usually the market, sometimes the drop if it grades lower. Read the breakdown before you post it.
- Skip
- Below 40. Grading does not pay for itself here at the grade you named.
One override sits above all three: a card that loses money by being graded is SKIP at any outlook. A deep market and a big multiple cannot argue a losing trade into a good one.
A missing price is not a zero
The two are opposite facts, and no default value is ever right for either.
“We do not know what this sells for” and “this sells for nothing” lead to opposite decisions. So a price is never quietly filled in with a zero: it is a real number with a real source, or it is explicitly absent and says why. A zero would run straight through the profit maths as a total loss and recommend against grading a valuable card, with no error anywhere.
Absence shows up as an em dash, never a number, and the reason matters because it says whose problem it is:
- — No sales recorded at this grade, explained in the guide
- Nobody has sold this card at this grade in that market. A real finding about the card.
- — Not entered yet, explained in the guide
- You have not told us yet — what you paid, or which grade you expect.
- — Priced in another currency, explained in the guide
- The vendor prices this row in another currency — a Japanese card’s raw price is in yen while its graded sales are in dollars — and no exchange rate could be read today. When the day’s rate is available the row is converted and every figure built on it says at what rate; without one it is refused rather than shown under the wrong symbol.
- — Not covered, explained in the guide
- That grader does not cover this game, or does not price in dollars. Not missing: impossible.
- — Not loaded, explained in the guide
- We failed to load it. That one is ours, not yours, and it means the number is coming back.
- — Does not apply here, explained in the guide
- The quantity does not exist for this row — the bottom rung of a ladder has no grade below it to gain over. Not missing, and no amount of data would fill it.
- — Different market, explained in the guide
- Both numbers exist, but they came from different graders’ markets, so subtracting one from the other would not describe this card. We show each price and refuse the comparison.
- — Not recorded at the time, explained in the guide
- It was true when a submission was sent and nobody wrote it down. Unlike everything above, waiting will not fill this in — what a submission cost is frozen at the moment it leaves, and a figure worked out from today’s fee tables would be one you were never charged.
- — Not fetched, explained in the guide
- The vendor has this number and we chose not to fetch it. A grader’s all-time sale count is context, not an outlook input, so when the card’s most recent sales already hold enough of that grader’s graded cards the rate is built from those and the total is left unbought — the caption says how many of the last hundred sales were counted instead.
When a component genuinely cannot be measured there is no outlook at all, rather than an outlook computed from the rest of the evidence and presented as if it were whole.
The outlook is yours, not a universal one
Your postage, your marketplace cut and your default grader are engine inputs, not preferences — they move every number on every card. Two people can look at the same card and get different answers because they are shipping to different places and selling in different markets, and that is correct.
How the model is built
What you paid never touches the outlook
The Grading Outlook asks whether grading this card pays from here. That is a forward-looking question, and what you paid is already gone. It is not a component, it is not an input to a component, and it cannot reach the outlook by any route.
Both directions of that are refused, which is the part worth saying out loud. Grade it, you are down on it and skip it, you overpaid are the same mistake wearing different clothes: the money is spent either way, and grading may well be the best remaining move on a card you are underwater on. Two people holding the same card get the same outlook — which is the only reason outlooks can be compared or sorted at all.
What you paid is not hidden, it is beside the outlook: every card shows your own profit and loss in its own panel, built from your purchase price. One question about the card, one question about your position, never mixed.
This has been caught leaking into the maths three separate times during the build — twice into a component, once into the recommendation, where the same card came back MARGINAL bought cheaply and SKIP bought dear on an identical outlook. It is now a single shared calculation that has no parameter for it.
Why the curves bend
Money moves by multiplying, so a scale counting in equal steps cannot describe it.
The step from ten dollars to a hundred is a different event from the step from ten thousand to ten thousand and ninety, and a straight line rates them the same. The first attempt here used straight lines and 41% of the cards it was tested against landed inside a single ten-point band — which is not a scale, it is a label.
So three of the four curves level off: each further step up costs several times more than the one before it, not the same amount more. Downside is the deliberate exception — it already has a floor and a ceiling, and bending it would blur losing a little into losing everything.
The curves and the percentiles were calibrated against 524 real card-variants on 2026-07-28, across 3 expansions — not chosen by eye. Each component’s page plots its own curve and lists those measurements run back through it.
When a component cannot apply
“We could not measure this” and “this cannot exist for this card” are different facts. A card sitting at the bottom of its ladder has no grade below it, so it has no downside — not an unknown one, an absent risk, and a favourable thing to know. Refusing to rate it would treat good news as missing evidence.
So a component that cannot apply drops out and the rest share its weight, keeping the outlook on the same 0–100 scale instead of capping it for a structural reason. With all 4 components present nothing moves at all. Worked through, for the case that actually happens:
| Component | Normally | With no downside to price |
|---|---|---|
| Profit potential | 35% | 44% |
| Grade premium | 20% | 25% |
| Market depth | 25% | 31% |
| Downside | 20% | drops out |
The card is then rated on what it actually has, and the breakdown says how many dimensions that was.
Nothing on this page is a second description of the model — it is the model, rendered. If a weight or a curve is retuned, these tables move on the next load, and a guard fails if anyone tries to type a number here that the engine does not hold.