Loading...
The widely cited 43% PSA 10 gem rate hides a massive selection bias. Our analysis of 32,000+ cards reveals the real odds of getting a PSA 10 -- and how pre-screening changes everything.
In short: The widely-cited ~43% PSA Gem Mint rate applies only to cards that collectors already hand-picked as their best — it is not your baseline. Unscreened cards gem at a far lower rate. Pre-screening with AI is the fastest way to achieve submission hit rates that match or beat that 43% benchmark.
You have probably seen it on Reddit, YouTube, or a grading forum: roughly 43% of cards submitted to PSA receive a Gem Mint 10 grade. The data comes from GemRate and CLLCT, which track PSA population reports and submission outcomes. It is real data. It is accurately reported.
And it is deeply misleading if you use it to estimate your own chances of getting a 10.
The reason is something statisticians call selection bias, and it fundamentally distorts how collectors think about grading odds. Understanding this bias is one of the most valuable things you can learn as a card collector in 2026, because it directly affects how you spend your grading budget.
Selection bias occurs when the sample you are measuring is not representative of the entire population. In card grading, the "sample" is the set of cards people actually submit to PSA. The "entire population" is every card sitting in collections, binders, shoeboxes, and top loaders around the world.
These two groups are nothing alike.
The cards that arrive at PSA have already survived multiple rounds of filtering. The collector pulled the card from a pack and thought it looked good. They inspected it under a light. They maybe checked the centering. They compared it to other copies they own. They decided this specific card was worth spending $25 on grading fees, plus shipping, plus insurance, plus two to six months of wait time.
By the time a card reaches a PSA grader's desk, it has already been pre-screened by human judgment. The obvious duds -- the off-center prints, the dinged corners, the surface-scratched pulls -- were filtered out before submission. What PSA sees is the cream of the crop.
So when 43% of that pre-selected, pre-filtered, hand-picked group receives a 10, it does not mean 43% of all cards in existence are Gem Mint. It means 43% of the cards that collectors already believed were their best met PSA's top standard.
Our State of Card Grading 2026 study analyzed over 32,000 cards through CardGrade's AI, and the data reveals a clear progression that exposes exactly how selection bias works.
When we look at all cards scanned through CardGrade -- every card, regardless of condition, with no pre-selection -- the predicted gem rate is dramatically lower than 43%. This is the true baseline for "a random card from your collection." Most cards simply are not in Gem Mint condition. They have centering issues, minor corner wear, edge imperfections, or surface flaws that keep them out of the top tier.
This number is the reality check. If you grab a random card from your collection and submit it to PSA, your odds of a 10 are far lower than what population reports suggest.
When we apply a quality filter -- looking only at cards that our AI predicts would grade PSA 8 or higher -- the gem rate climbs significantly. This mirrors what happens when a knowledgeable collector pre-screens their cards before submitting. You remove the obvious non-contenders and your hit rate improves.
At the top of the funnel, the GemRate/CLLCT data shows approximately 43% of actual PSA submissions receiving a 10. These are cards that have been filtered not just by a quality threshold but by human judgment, experience, and financial motivation. People do not spend $25 per card on cards they think will grade poorly.
The gap between Stat 1 and Stat 3 is the selection bias gap. It represents the difference between "all cards" and "cards someone decided were worth grading." That gap is enormous.
Here is where it gets practical. If you believe the 43% gem rate applies to your cards, you will over-submit. You will send in cards that feel like 10 candidates but are actually 8s and 9s, because you have anchored your expectations to a number that was never meant to describe unscreened cards.
At $25 per card through PSA's Collectors Club pricing, every card that comes back below expectations is money lost. And the losses add up fast.
Consider this scenario. You have 40 cards you think are solid 10 candidates. Based on the 43% gem rate, you expect about 17 of them to come back as 10s. You submit all 40 at $25 each -- that is $1,000 in grading fees.
But your 40 cards were not pre-screened with the same rigor as the average PSA submission. Maybe 8 of them actually get a 10. Another 12 get a 9. The remaining 20 come back as 8s or lower. Those 20 cards cost you $500 in grading fees for slabs that may not add any value over raw. You expected 17 gems and got 8.
The problem was not your cards. The problem was your benchmark. You compared your unscreened pile to a heavily screened dataset and assumed the same odds applied.
The good news is that selection bias works in your favor once you understand it. The collectors who achieve gem rates at or above 43% are the ones who screen aggressively before submitting. You can do the same thing.
This is the mindset shift. Stop assuming your cards are gems until proven otherwise. Start assuming they are not, and let the evidence convince you. Look at every card as a skeptic, not an optimist.
Our study found that centering is the number one grade-limiting factor across all card categories. A card with 60/40 centering or worse is not getting a 10 at any major company. Use the CardGrade centering tool to get an exact measurement before you evaluate anything else. If centering fails, stop there -- do not waste time inspecting corners and edges on a card that is already disqualified.
This is where technology gives you an unfair advantage. Instead of relying on your eyes and experience (which is what the 43% population already did), use CardGrade's AI grading to get an objective assessment of every card. The AI evaluates centering, corners, edges, and surface simultaneously and predicts grades across PSA, BGS, and CGC scales.
When you filter your collection down to only the cards that AI predicts as strong 10 candidates, your actual submission gem rate can match or exceed the 43% benchmark -- because now you are doing the same rigorous pre-selection that drives that number.
Decide before you start screening: what is your minimum predicted grade for submission? For modern cards where only a 10 carries a premium, set the bar at predicted PSA 10 with high confidence. For vintage or high-value cards where a 9 still adds significant value, you can lower the threshold.
The key is to have a threshold and stick to it. Do not let emotional attachment to a card override the data. If a card predicts at a 9, it is a 9. Submitting it hoping for a 10 is the exact behavior that inflates your costs and deflates your hit rate.
After each submission comes back, compare your actual grades to your AI predictions. This feedback loop helps you calibrate your threshold over time. If you are consistently getting grades that match or beat predictions, your screening process is working.
Let's revisit the 40-card scenario with proper pre-screening. You start with 40 candidates and run them through CardGrade. The AI identifies 15 as strong PSA 10 candidates, 10 as likely 9s, and 15 as 8 or lower.
You submit the 15 strongest cards at $25 each -- that is $375 in grading fees instead of $1,000. Of those 15, maybe 8 to 10 come back as 10s. Your gem rate is now above 50%, and you saved $625 in fees on cards that would not have justified the cost.
Over a year of quarterly submissions, that kind of discipline saves thousands. It is not about grading fewer cards -- it is about grading the right cards.
The 43% PSA gem rate is a real number that describes a specific, heavily pre-selected group of cards. It is not your baseline. It is not your expected outcome. It is the result of thousands of collectors doing exactly what you should be doing: screening hard before submitting.
If you want 43% gem rates, you need 43%-level screening. And the fastest way to get there is to pre-screen with AI before you spend a dollar on grading.
For the complete dataset and methodology behind these findings, read the full State of Card Grading 2026 study. For the broader takeaways, see our key findings summary. And for a full breakdown of what grading costs across companies, check out our PSA grading cost breakdown.
Ready to find out which of your cards actually have a shot at a 10? See CardGrade pricing and start screening smarter.
Population data referenced from GemRate and CLLCT. Rates shift as submission volumes change — verify before planning a batch.
Pre-screen your cards before submitting to PSA, BGS, or CGC. Start free 3-day trial.
Start free trialAlso available as a Chrome Extension
Grade cards while you browse
Install our free Chrome Extension — right-click any card on eBay, COMC, or TCGPlayer for instant grades.
Founder, CardGrade
Jamie Budesky is the founder of CardGrade and the engineer behind its AI vision grading pipeline. An Army veteran and IT specialist (DoD, since 2017), he writes about card grading, AI/ML grading technology, and collecting strategy — grounded in CardGrade's own grading data.
Should you use your iPhone camera or a flatbed scanner for AI card grading? We break down image quality, convenience, accuracy, and when each method makes sense.
Edges and surface quality can make or break your card grade. Learn what graders look for and how to evaluate these often-overlooked sub-scores.
Proper preparation protects your cards and maximizes grades. Learn the step-by-step process for cleaning, sleeving, and packaging cards for PSA, BGS, or CGC.