SLABBYPREPSee a sample verdict · Subscribe · $3/month →

How SlabbyPrep measures centering, and where every number comes from

SlabbyPrep is part of CrabTree Collections, a trading-card store at 6600 Hiatus Road, Tamarac, Florida, USA. Questions or corrections: crabtreecollections@gmail.com. This page explains what the tool measures, how well, and the source and date of every standard, fee and price it uses.

The measurement

Find the card. Two independent methods look for the card in the photo: one separates the card’s colours from the background, the other floods the background inward without crossing strong edges (which survives dark frames on dark mats). The better outline wins. Straight lines are fitted to its sides, so rounded corners do not pull the corners in, and each side is snapped to the strongest edge at full resolution.
Straighten it. The four corners are mapped to an upright card of the size you chose, 63 × 88 mm (Pokémon, Magic, sports and most others) or 59 × 86 mm (Yu-Gi-Oh), which removes the tilt of a hand-held photo.
Find the borders. On each side the outer edge is found again in the straightened image, and the inner edge of the border is the strongest colour step just inside it, agreed across many rows or columns. The widths become millimetres from the card size, and the shares from the two widths on each axis: 3.3 mm and 2.7 mm is 55/45.
Check it yourself. Every line can be dragged with a magnifier; the figures follow your correction.

All of this runs in your browser. The photo is never uploaded.

How accurate it is

Measured on synthetic photographs with known border widths, generated with the same code that runs the tests on every change:

On a clean, well-lit synthetic photo the measured left/right and top/bottom shares land within 1.5 percentage points of the truth (tests/pipeline.test.ts).
Across 40 random synthetic cards with varied backgrounds, tilts and noise, at least 36 are found automatically and the mean error is under 0.8 points per axis (tests/pipeline.test.ts).
Border widths on Yu-Gi-Oh-size cards (59 × 86 mm) land within 0.12 mm (tests/card-size.test.ts).

Hard photos

Synthetic photos built to break the measurement, 40 per case, each with exact known borders (measured with scripts/cv-bench.mjs on 2026-11-02; a sample of each case is pinned in tests/cv-hard.test.ts). When a photo is too hard, the app says so and asks you to check the lines instead of showing a confident figure; the table counts how often a reading was shown without that warning, and how close those readings were on the worse axis.

Case	Photos	Shown without a warning	Within 1 point	Within 0.5 point
Clean photo	40	36	36	35
Tilted up to 25°	40	35	35	34
Glare band across the card	40	38	37	36
Uneven light	40	38	38	38
Sensor noise and heavy JPEG compression	40	31	31	29
Motion blur	40	35	34	31
Out of focus	40	34	33	30
In a penny sleeve	40	37	36	36
In a top-loader	40	39	38	37
Yu-Gi-Oh size, several problems at once	40	15	15	15
Several problems at once	40	13	12	10
On a wood table	40	17	16	16
On a patterned cloth	40	13	12	9
On a busy playmat	40	0	—	—
Border the same colour as the surface	40	8	7	7
Full-art card, art close to the border colour	40	10	8	8
Borderless card (no centering to measure)	40	0	—	—

Across the 640 photos with a border, 399 readings were shown without a warning and 388 of them (97%) were within 1 percentage point; 6 were more than 2 points off (one each with motion blur, defocus, glare and a sleeve, and two full-art cards). Everything else came with a warning to check the lines. On busy backgrounds the outline is usually not found automatically: place the corners by hand. Borderless cards are always flagged, because they have no border to measure.

Real photos add things synthetic ones only imitate, such as lens distortion and printing that runs into the border. Shoot straight on, out of the sleeve, on a plain surface, and check the lines in the magnifier before you rely on a figure near a grade line.

What it does not measure

Corners, edges and surface: you rate them, and they can only hold or lower the grade the centering allows. Grader judgement: PSA and CGC write “approximately” and TAG writes “~” on their lines, and every company keeps discretion. No figure here is a promise of a grade.

The verdict

For each company, the best grade your centering and condition allow is the top of a ladder of likely outcomes: the top grade at a probability set by your condition ratings (or the one you choose), the next grades below it at decreasing shares. Each outcome is priced from the market prices you enter or look up, the cheapest tier whose value cap covers the top outcome’s price is chosen, and the expected sale value after that fee is compared with the raw price. Companies that do not accept the card (TAG and Yu-Gi-Oh!, for now) are never recommended. Selling fees and shipping are not deducted; allow for them.

Where the standards and fees come from

Each company’s own pages, read on the date shown. Each company is re-read monthly where its site can be read; Beckett and SGC block automated reading, so their figures come from the dated official copies shown.

Company	Centering standard	Fees
PSA	standards · 2026-10-28 Read from PSA’s grading standards page.	fees · 2026-11-02 Read from PSA’s Trading Card Grading page (its pricing, Premium tiers included) on 2026-11-02: every fee, cap and turnaround as on 2026-10-29. SlabbyPrep counts Value ($39.99, cards up to $500) as not open because it sits under PSA’s notice “Value Services Are Temporarily Paused”, though PSA still lists it with Get Started.
Beckett (BGS)	standards · 2026-07-14 From Beckett’s own grading page as archived on 2026-07-14 (beckett.com blocks automated reading; it was offline from 2026-09-09 until its online submission center was restored, by 2026-10-15).	fees · 2026-10-15 Prices from Beckett’s own grading page as archived on 2026-07-14; availability from Beckett’s maintenance page of 2026-10-15. Beckett.com blocks automated reading, so its restored submission center could not be checked.
CGC Cards	standards · 2026-10-28 Read from CGC Cards’ grading scale page.	fees · 2026-10-28 Read from CGC Cards’ Services and Fees page (Collector-member prices).
SGC	standards · 2026-09-17 From SGC’s own grading scale as captured on 2026-09-17 (gosgc.com blocks automated reading).	fees · 2026-09-17 From SGC’s own Services and Pricing table (standard-size cards) as captured on 2026-09-17; gosgc.com blocks automated reading.
TAG	standards · 2026-10-28 Read from TAG’s grading rubric. Back figures shown are TAG’s TCG figures; sports backs are looser.	fees · 2026-10-28 Read from TAG’s pricing page and its help article on available services (updated 2026-10-21).
ACE	standards · 2026-10-28 Read from ACE’s grading scale page.	fees · 2026-10-28 Read from ACE’s pricing page (GBP), converted at 1 GBP = 1.3288 USD (ECB rate, 2026-10-28).

Where the prices come from

Trading-card prices come from PriceCharting’s price guide, with their permission, one card per lookup, credited with a link to the card’s page wherever a price appears. Sports cards are not in that guide; their prices are entered by hand, with links to recent sales. Nothing is stored.

How to check centering by hand · every line by grade · the verdict’s arithmetic · fees · privacy