Card Grading NewsA TCG Treasury Publication
News

The Real Gem Rate Is 23.9%: What 28,656 Scanned Cards Show

28,656 scanned cards show a 23.9% unfiltered gem rate vs the cited ~43% — selection bias explains the gap, and centering is the top grade limiter at 36.4%.

Jamie Budesky

By Jamie Budesky · Editor, Card Grading News ·

On August 7, 2026, Card Grading News completed an analysis of 28,656 AI gradings pulled directly from CardGrade's production database — every card graded between February 14 and August 7, 2026, that carried all four subgrades: centering, corners, edges, and surface. The headline number: 23.9%. That is the share of scanned cards predicting a PSA 10. Set it against the roughly 43% gem rate widely cited from GemRate and CLLCT population reporting, and the most important fact in grading economics falls out of the data: the cards in your box are not the cards in the pop report.

The 43% figure is not wrong. It is a measurement of survivors. Every card behind it was already picked up, inspected, judged worth a grading fee, and mailed. CardGrade's dataset, by contrast, is what cards look like before anyone decides they are worth submitting — the unfiltered population. Per CardGrade's own grading data, analyzed by Card Grading News, barely one card in four clears the gem bar when nobody has pre-screened the pile.

Why 23.9% and 43% Are Both True

The gap is selection bias, and it is enormous. A card only enters PSA's population report after its owner looked at it under a lamp, decided the corners were sharp and the centering acceptable, and committed real money — $24.99 to $79.99 per card depending on tier, before the June pause — plus weeks of waiting. Cards with visible whitening, off-center borders, or surface scratches get filtered out at the kitchen table and never appear in any published gem rate. CardGrade's dataset has no such filter. It includes pack-fresh pulls, shoebox finds, inherited collections, and pure curiosity scans. Remove the human pre-screen and the gem rate nearly halves, from ~43% to 23.9%. Neither number is a lie; they measure different populations. Published gem rates measure the output of the submission funnel. This dataset measures its input — and the difference between the two is the value collectors add (or fail to add) by sorting before they ship. The full methodology lives on CardGrade's research page at cardgrade.io/research/state-of-card-grading-2026.

The Full Distribution: What 28,656 Cards Predict

Predicted PSA grade distribution, 28,656 completed AI gradings (Feb 14 – Aug 7, 2026, unfiltered)

Predicted grade

Cards

Share of all scans

PSA 10

6,860

23.9%

PSA 9

10,532

36.8%

PSA 8

5,328

18.6%

PSA 7

1,799

6.3%

PSA 6 or lower

4,137

14.4%

Read the table from the bottom up. 39.3% of everything scanned — combining the PSA 8, PSA 7, and PSA 6-or-lower rows — predicts a grade of 8 or below. With PSA's four value tiers paused since June 2, 2026, the cheapest open door is Regular at $79.99 per card with a 40-50 business day turnaround. That reframes every row as a bet. Predicted 9.5–10: submit — even at $79.99, a gem candidate is the bet the fee structure was built for. Predicted 9: submit only when the market spread between the raw card and a PSA 9 slab clearly covers the $79.99 fee — check cardgrade.io/is-my-card-worth-grading before you decide. Predicted 8 or below: keep it raw. The modal outcome is PSA 9 at 36.8%, which is exactly why blind submission stings: the single most likely result is the grade that most often fails to pay for itself at Regular-tier pricing.

Centering Kills More 10s Than Anything Else

Sole weakest subgrade (ties excluded; 21,767 cards, 76.0% of sample, had one clear weakest subgrade)

Limiting subgrade

All cards

Cards that miss a PSA 10

Centering

36.4%

39.0%

Edges

24.2%

17.0%

Surface

20.1%

23.0%

Corners

19.3%

21.0%

Among the 21,767 cards with one clear weakest subgrade, centering is the sole limiter on 36.4% of all cards and 39.0% of cards that miss a PSA 10 — nearly double the next factor, surface, at 23.0% among the miss-10 group. The mean subgrades tell a subtler story: corners actually post the lowest average at 9.02, against 9.12 for edges, 9.15 for surface, and 9.25 for centering. Corners degrade broadly, but rarely alone — when corners are bad, something else usually is too, so they seldom stand as the single grade-killer. Centering is the opposite: a card can be flawless in every zone and still be dead on arrival because the borders are uneven. That makes centering the one flaw worth checking before anything else, and the one you can measure yourself in seconds with the free tool at cardgrade.io/tools/centering. If the borders fail, no amount of corner sharpness rescues the grade — see the grade-by-grade breakdown at cardgrade.io/guides/psa-grades for exactly what each threshold demands.

The 14.4% That Should Never Ship

4,137 cards — 14.4% of everything scanned — predicted PSA 6 or lower. These are not random junk; every one of them was a card somebody cared enough about to photograph and grade. A meaningful share of what collectors consider submitting simply should never go in the envelope. Whether a PSA 6 slab recoups its fee depends on the individual card, but the math starts $79.99 in the hole, plus shipping, plus 40-50 business days of waiting while PSA works down a backlog approaching 10 million cards — a pile that grew by 1.6 million cards in two weeks after a 20% submission spike this spring. Before June 2, a marginal card could ride Value Bulk at $24.99 and the downside was a cheap lesson. At $79.99, submitting anything that comes back PSA 8 or below is a costly one. The full tier-by-tier pricing picture is at cardgrade.io/psa-card-grading-cost.

The Market This Data Lands In

Context matters because the industry has never been bigger or slower. Roughly 12.5 million cards were graded in 2025, generating an estimated $187-312 million in revenue. PSA holds about 70% of the market, ahead of SGC at ~15%, BGS at ~10%, and CGC at ~5% — the head-to-head cases are laid out at cardgrade.io/compare/psa-vs-bgs. PSA announced the value-tier pause on May 28, 2026, effective June 2, and will not reopen those tiers until the backlog falls from near 10 million to 5 million cards — a process PSA itself estimates could take up to four months, tracked in a monthly public Backlog Tracker and backed by a $200 million capital commitment. In that market, every submission slot is expensive and every misjudged card compounds the queue. An unfiltered 23.9% gem rate says the queue contains a lot of cards that were never going to justify their spot in it.

Scan First, Then Decide

The entire 19.1-point gap between 23.9% and ~43% is pre-screening — which means pre-screening is where the money is made or lost. CardGrade runs 16 inspection zones and returns a predicted grade in under 60 seconds, from $4.99 a month for 10 grades; a single $79.99 Regular submission fee funds sixteen months of scanning at that rate. The workflow that follows from the data is short:

  • Check centering first — it is the sole limiter on 39.0% of cards that miss a 10; measure it free at cardgrade.io/tools/centering.
  • Scan before you ship — get a predicted grade and subgrades before committing $79.99 and 40-50 business days.
  • Apply the rule: predicted 9.5–10 submits, predicted 8 or below stays raw, and predicted 9 goes through cardgrade.io/is-my-card-worth-grading before you decide.
  • Verify anything already slabbed with the cert lookup at cardgrade.io/tools/psa-cert-lookup before buying based on someone else's filter.

Population reports will keep publishing gem rates in the forties, and collectors will keep reading them as odds. They are not odds. They are the score of players who already made the cut. Your box plays at 23.9% until you sort it.

The bottom line

The 43% gem rate measures cards that survived a filter, not cards as they exist — your collection plays at 23.9% until you run your own filter, and at $79.99 per Regular-tier submission, paying PSA to run it for you is the most expensive way to find out.

Frequently asked questions

Why is the 23.9% gem rate so much lower than the widely cited 43%?

The ~43% figure from GemRate/CLLCT reporting describes cards collectors had already inspected, hand-picked, and paid to submit — a pre-filtered population. CardGrade's 23.9% covers every card scanned, with no submission filter applied. The 19-point gap is pure selection bias: weak cards get culled at the kitchen table and never reach a population report.

Are these actual PSA grades?

No — they are AI-predicted grades from 28,656 completed CardGrade gradings between February 14 and August 7, 2026, each running 16 inspection zones with all four subgrades present. That is precisely what makes the dataset useful: it captures the population before the submission filter that shapes every published PSA gem rate.

What flaw most often costs a card its PSA 10?

Centering. Among the 76.0% of cards with one clear weakest subgrade, centering is the sole limiter on 36.4% of all cards and 39.0% of cards that miss a 10 — nearly double surface, the next factor at 23.0%. Corners post the lowest average subgrade (9.02) but rarely fail alone, so centering remains the single most common grade-killer.

Is it worth submitting a card predicted PSA 8 right now?

Generally no. With PSA's four value tiers paused since June 2, 2026, the cheapest open option is Regular at $79.99 with a 40-50 business day turnaround, and an 8 rarely clears that fee plus shipping. Run the card through cardgrade.io/is-my-card-worth-grading before spending anything.

When will PSA's cheaper tiers come back?

PSA says the paused tiers — Value Bulk $24.99, Value $32.99, Value Plus $49.99, and Value Max $64.99 — reopen when the backlog falls from nearly 10 million cards to 5 million. PSA estimates that could take up to four months, publishes a monthly Backlog Tracker, and has a $200 million capital commitment deployed against the queue.

Sources