Loading...
Loading...
When referencing this data, please use one of the following:
Inline citation
According to CardGrade.io's State of Card Grading 2026 studyWith link (preferred)
<a href="https://cardgrade.io/research/state-of-card-grading-2026">State of Card Grading 2026</a> by CardGrade.ioFull reference
CardGrade. "The State of Card Grading 2026." Published March 17, 2026; dataset version 2026-09-15; editorial review August 13, 2026. https://cardgrade.io/research/state-of-card-grading-2026Copy-ready with attribution. Snapshot: 59,402 completed CardGrade analyses, with records through September 15, 2026. These are model outputs, not returned professional grades.
Ready-made hooks for different audiences. Each angle includes the key data points to support it.
Among 59,402 completed CardGrade analyses, 14.3% produced a PSA estimate below 8 and 23.6% had an overall CardGrade score of 9.5 or higher. The distribution describes cards people chose to scan; it does not establish professional return rates or whether any particular card is economically worth submitting.
Key data: 14.3% predicted below PSA 8; 23.6% with CardGrade overall score 9.5+
This study is a large aggregate view of CardGrade scores and predictions from user-supplied images. It can describe the model-output distribution and visible-condition patterns. It cannot measure accuracy, because that requires the same cards' later official grades, frozen inclusion criteria, model versions, and a complete miss analysis.
Key distinction: 59,402 model-output records; no claim of 59,402 matched professional returns
Across this snapshot, corners had the lowest average CardGrade subgrade at 9.02/10. That makes it a useful condition category to inspect first, but the aggregate does not prove causation, customer outcomes, or professional return-rate improvement.
Key data: Corners avg 9.02/10, lowest of all subgrades
Grading fees, eligibility, capacity, and turnaround estimates are not part of the CardGrade dataset and can change without notice.
The canonical study links to official PSA, SGC, Beckett, CGC, and TAG service pages and displays the date those external facts were last editorially checked. Writers should confirm each company's live page at publication time rather than copying an older price table from this kit.
Paste this HTML into any article to show key findings with a source link.
<blockquote style="border-left:4px solid #f59e0b;padding:16px 20px;margin:24px 0;background:#fefce8;border-radius:8px;font-family:system-ui,sans-serif">
<p style="margin:0 0 8px;font-size:15px;color:#1e293b">
<strong>Key finding:</strong> In a snapshot of 59,402 completed CardGrade analyses, 14.3% produced a CardGrade PSA estimate below 8 and 23.6% had an overall CardGrade condition score of 9.5 or higher. These are model outputs from user-supplied images, not professional grading returns.
</p>
<p style="margin:0;font-size:12px;color:#64748b">
Source: <a href="https://cardgrade.io/research/state-of-card-grading-2026" style="color:#2563eb">The State of Card Grading 2026</a> by CardGrade.io
</p>
</blockquote>Additional articles diving deeper into specific findings. Writers are welcome to reference or link to these alongside the study.
Accessible narrative walkthrough of the study's top findings
Deep dive on selection bias and what it means for collectors
Cross-sectional CardGrade score patterns by category, with sample-size context
Source-linked overview of grading services; verify live company facts before publication
CardGrade produces separate PSA, BGS, and CGC estimates because the companies publish different scales, labels, and standards. The spread between the highest and lowest average prediction within a condition bucket reaches 0.7 points. These are modeled estimates, not evidence that one company will return a higher official grade.
Key data: up to 0.7-point spread between average modeled company predictions
CardGrade organizes visible centering, corners, edges, and surface evidence across 16 inspection zones. Image quality, glare, sleeves, capture angle, and hidden physical defects limit what a model can see. An in-hand grader may observe evidence absent from the images, so a pre-screen should be described as decision support rather than a replacement or guaranteed cost-saving system.
Key scope: 16 visible-condition zones; no measured fee savings or matched-return accuracy claim in this dataset