How it works, and what it refuses to claim

Centering is the one attribute a phone can judge objectively, so the app measures it. Point it at a photo of the card and it finds the four border edges by looking for where brightness changes sharply across each axis, then compares the gaps.

A card is perfectly centered at 50. The app treats 55 as the ceiling for a gem mint grade, and steps down from there. When it cannot find four clean edges it says so and returns nothing, rather than reporting a number it made up.

You can skip the photo and answer from memory. Those answers are shaded down by 0.5 of a grade, because people are consistently kind about their own cards and the app should not be.

Two graders can look at the same card and differ. So the app does not predict a grade, it predicts a distribution across the whole scale, and the verdict is computed over all ten outcomes weighted by how likely each is.

The weakest subgrade caps the whole card. A card with flawless corners, flawless edges and a scratched surface is a scratched card, not an average. The cap is forgiven by at most one point, and only when everything else is genuinely far ahead of it.

Some defects end the discussion rather than costing points. A crease does not take two grades off a card; it puts a ceiling on it that no amount of good centering climbs back over.

The sticker price of a tier is not what grading a card costs. Postage both ways, insurance that scales with declared value, sleeves and toploaders all land on top, and postage is charged per package rather than per card.

That last part is why the answer for one card is not the answer for a box. Sending a single card is the most expensive way to grade anything, and the app shows you exactly how much the batch size is costing you.

Then the card has to be sold, so marketplace fees and shipping come off the other end too. The number you see is what you would actually keep.

Prices at each grade come from one of three places, and the app always says which: a real sold price you entered, a market data provider, or the app's own era based multiplier.

That last one is the weakest thing in the model and it is labelled as such everywhere it appears. A vintage gem mint premium the app calls twelve times the raw price could really be eight, or eighteen. So when the entire curve is estimated, the app will not say send unless the edge also clears the cost per card: if the estimate is wrong by the whole fee, you are still roughly where you started.

Pasting one real sold price at the break even grade changes more than pasting five anywhere else, and the app tells you which grade that is.

It is not a grade. No grading company has seen your card, and nothing here binds what one would say about it. It is an estimate built from what you told it.

It cannot see your card. Corners, edges and surface come from your own answers. Print lines, dimples and whitening are things you have to notice and report, and a card you describe generously gets a generous answer.

It cannot spot a fake. Nothing here authenticates anything.

Prices move. Grading rates change, market prices change faster, and a verdict is a snapshot rather than a promise.

An edge under $25 is treated as not worth the wait even when it is positive, because months of your card being somewhere else has a cost the arithmetic does not capture.