The Grading Lab
Is it centred well enough to be worth grading? Point your camera at it and find out, before you pay to ask someone else.
How to shoot it
- Card flat on a plain surface that is not the same colour as its border.
- Phone directly above it, not at an angle. A tilted photo has real perspective and the margins it measures are the photograph’s, not the card’s.
- Fill the frame. Even light, no flash bouncing off the holo.
- In a sleeve is fine. A toploader usually is not: its edge reads as a second border.
Every measurement happens in this browser. There is no upload endpoint behind this page, so the photo cannot leave your device even by accident. Reading the card’s Japanese text works the same way.
What the graders actually require
These are the published front-centring tolerances, and they are only the centring component of a grade. A card inside the tolerance for a 10 still fails on a scratch, a soft corner or a print line. What this page can tell you is whether centring has already ruled the top grades out, which is the cheapest question to answer and the one most often answered wrong by eye.
| Grade | Front centring |
|---|---|
| 10 | 60/40 or better |
| 9 | 65/35 or better |
| 8 | 70/30 or better |
| 7 | 75/25 or better |
| 6 | 80/20 or better |
| Grade | Front centring |
|---|---|
| 10 Pristine | 50/50 or better |
| 9.5 | 55/45 or better |
| 9 | 60/40 or better |
| 8.5 | 65/35 or better |
| 8 | 70/30 or better |
| Grade | Front centring |
|---|---|
| 10 Pristine | 55/45 or better |
| 9.5 | 60/40 or better |
| 9 | 65/35 or better |
| 8 | 70/30 or better |
How the measurement works
A card is a rectangle with a coloured border, and the artwork sits inside it as a second rectangle. Centring is the ratio of the margins between the two. So the page does not try to recognise anything: it takes the brightness of every pixel, measures how fast brightness changes, and adds those changes up along each row and each column. A straight edge puts a spike in that sum at exactly its position, because a hundred pixels in a line add up and noise that is not in a line does not. The outermost spike on each side is the card. The next one inwards is the artwork frame. Everything after that is division.
No model, no training data, no server. The whole measurement is in src/lib/grade/centring.ts, it is a few hundred lines of arithmetic, and it is tested against synthetic cards built at known ratios so that a card constructed at 75/25 has to come back as 75/25.
What it cannot do
- Surface, edges and corners. Most of a grade, and none of it visible in the geometry of one photograph. This measures centring, and centring alone.
- Back centring. Graders look at both faces and weight the front more heavily. Shoot the back separately if you want both numbers.
- A photo taken at an angle.Perspective changes the margins, so the number you get is the photograph’s centring rather than the card’s. Shoot from directly above.
- A card in a toploader.The holder’s edge reads as a second border and the measurement finds it instead. A penny sleeve is fine.
- Full-art and borderless cards. There is no border to measure, and graders judge those on the thin edge of the print itself. The page will tell you it could not find a frame rather than invent one.