Calibrated on 24,317 graded cards · v1.2.4

Don't pay to grade
cards that won't pay you back.

GradingMetric is the institutional-grade intelligence layer for card submissions. Front + back image, eight seconds, and a calibrated probability across PSA, BGS, CGC and SGC — plus the economics behind whether to send each one.

24,317
Cards calibrated
< 3%
Calibration error
4
Graders covered
$2.4M
Pre-grade evaluated
PROSPECT '21
ALL-STAR '23
— ROOKIE GOLD — M. KESTRAL RC · #218 · 2024 SIGNAL ▲ FOIL · /99 SP★
Capital Score
84
Submit · PSA 10 likely
Best fit
PSA 1062%
Used by professional submitters managing $10M+ portfolios
PPinnacle Holdings AAtrium Capital MMonolith Trading VVermont Vault RRiverhouse
The Product

A trading desk for your collection.

Upload a batch. Get probability across every grader, defect-level analysis, recommended action, and the economic case for each card — all in one frame.

app.gradingmetric.com / batches / spring-modern-32
Workspace
Credits
342 remaining

Dashboard

Performance — May 2026
+ New Batch
Cards Analyzed
1,847
↗ 312 this month
Submit Rate
38%
+4.2 pp vs apr
Batches
24
8 active
Avg Capital
71
Above 70 threshold
Batch Cards Submit % Avg Score Status
Spring Modern · 32 32 44% 76 ● ready
'24 Bowman Chrome 18 28% 62 ● review
Vintage Hockey · Lot 7 11 73% 84 ● ready
Auction Pulls · MJ 86 7 14% 42 ● hold
The Capital Score

One number tells you
whether to spend the postage.

Most tools tell you what grade your card will get. That's half the picture. The Capital Score combines grade probability, grader-specific tolerances, current market multiples, and submission fees into a single 0–100 economic signal.

70+ is a clear submit. Below 50, you're paying postage and grading fees to lose money. Every score ships with a confidence band so you know when to trust it.

— Capital Score formula
CS(card) = Σ[
  P(grade) × ΔPrice(grade)
] Fees Shipping
  normalized to [0, 100]
CARD · 7F31A2B6 · 2024 Topps Chrome — Kestral RC
Submit
84 / 100
0 50 70 · submit 100
P(PSA 10)62%
P(PSA 9)31%
PSA 10 market$640
PSA 9 market$110
Submit fee$25
Expected return+$391
Confidencehigh
Best fitPSA
How it works

Three steps. Eight seconds per card.

From phone camera to submission shortlist. No spreadsheets, no comparable hunting, no guessing across grader scales.

01

Photograph front + back

Snap each side under any lighting. The vision pipeline normalizes glare, centering reference, and surface texture before the model ever sees it. Heic, jpg, png, webp.

02

Get four grader matrix

One inference pass returns full grade distributions for PSA, BGS, CGC and SGC — every probability calibrated against our 24k-card ground-truth corpus.

03

Submit only what pays

The Capital Score ranks each card by expected economic return. Export the submission list as CSV, paste into your grader portal, ship.

Coverage

Every grader speaks a different language.

We model each one's house preferences — PSA's centering tolerance, BGS's sub-grade arithmetic, CGC's surface bias, SGC's snug-fit rules — so the right card goes to the right house.

PSA

Professional Sports Authenticator
110
Centering tol.55/45
Half gradesno
Avg turnaround45 d
Best formainstream

BGS

Beckett Grading Services
110 BLK
Sub-grades4
Half gradesyes
Avg turnaround28 d
Best forpristine

CGC

Certified Guaranty Company
110 P
Surface focushigh
Half gradesyes
Avg turnaround21 d
Best formodern

SGC

Sportscard Guaranty Corp.
110 PRI
Vintage biasyes
Half gradesyes
Avg turnaround15 d
Best forvintage
Methodology

Calibrated. Auditable. Not vibes.

The model is trained on real submission-to-grade pairs and continually back-tested. Every prediction ships with the calibration data behind it.

24,317
Ground-truth pairs

Each card in our corpus has photographed front + back, full sub-grades, and a confirmed grading slab — no second-hand reports, no inferred labels.

2.7%
Brier calibration error

When we say a card has a 60% PSA 10 chance, it actually grades PSA 10 within 58–62% of the time. Measured weekly against fresh submissions.

+18.4%
Submission yield uplift

Median submitter using our shortlist sees an 18.4% higher net return per dollar of grading fees compared to their unfiltered baseline.

Calibration plot

When we say 70%, we mean it.

Predicted PSA 10 probability vs. observed rate across 4,200 held-out submissions. The closer the curve to the diagonal, the better-calibrated the model.

Mean abs. error
2.7%
Held-out set
4,200
Sharpness
0.41
0% 100% PREDICTED P(PSA 10) OBSERVED
Pricing

One credit. One card. Never expires.

No subscriptions. No usage tiers you'll forget about. Buy credits when you have a batch, sit on them when you don't.

Try

5 credits
$9.00
$1.80 per card
  • 5 card gradings
  • Never expires
  • AI-powered analysis
  • CSV export
Buy

Starter

15 credits
$25.00
$1.67 per card
  • 15 card gradings
  • Never expires
  • AI-powered analysis
  • CSV export
Buy

Elite

100 credits
$140.00
$1.40 per card
  • 100 card gradings
  • Never expires
  • AI-powered analysis
  • CSV export
Buy

Enterprise & vault partner pricing — contact us

Who it's for

Built for everyone who submits seriously.

Whether you move one card or five hundred a week, the workflow is the same: scan, see the economics, submit only what pays.

Dealers & bulk submitters

Triage a 500-card intake in an afternoon. Sort by Capital Score and only spend grading fees on the cards that earn their postage back.

Vault & auction operators

Export the submission shortlist as CSV straight into your grader portal. Cut hours of comparable-hunting off every consignment.

Solo collectors

Check any card against the same calibrated model the desks use — before you commit to a grading fee on a hunch.

Questions

Things people actually ask us.

Is GradingMetric affiliated with PSA, BGS, CGC or SGC?
No. We're independent. We model each grader's published standards and back-test against our own corpus of confirmed slabs — there's no insider data, no quid-pro-quo, no preferred-grader incentives. The recommendations are honest.
How accurate is the model on cards it's never seen?
We measure this constantly. On held-out submissions, our calibration error sits below 3% Brier score, which means a stated 70% probability lands within 67–73% in practice. Confidence intervals are returned per-card so you know when to second-guess.
Does the model work for vintage cards?
Yes, but accuracy degrades for pre-1965 issues where surface defects matter more than centering. We surface lower confidence scores on these and explicitly recommend SGC's vintage scale when the math supports it.
What about altered or tampered cards?
We flag likely trimming, recoloring, and tape residue in the defect breakdown. We can't replace a professional authenticator, but we'll tell you when a card has signals worth investigating before you ship.
How are credits priced?
One credit = one card analyzed. Credits never expire. The more you buy at once, the lower the per-card price. The Trial pack gives you 2 free credits with signup so you can verify the analysis against cards you've already had graded.
Can I use this for trading cards beyond sports?
We support modern sports, vintage sports, and TCG (Pokemon, Magic, One Piece). Non-sport entertainment cards are in private beta — contact us if you want access.

Stop guessing. Start submitting smart.

Two free credits with signup. Run them against cards you've already graded and see for yourself.