For 30 years, the sports card hobby has relied on the opinion of anonymous individuals in dark rooms. A "Gem Mint 10" was whatever that person felt it was on that specific Tuesday morning. In 2026, that is no longer acceptable.
companies like TAG, AGS, and now PreGradeCards are using Computer Vision (CV) and Photometric Stereopsis to measure cards with sub-pixel accuracy. We aren't just looking at the card; we are mapping its topography.
Human vs. AI: The Consistency Gap
The biggest problem in grading is not strictness; it's consistency. If you crack a PSA 9 and resubmit it, you have a 30% chance of getting a different grade. That is a broken system.
Repeatability Study (100 Resubmissions)
*Data based on blind control group resubmissions of the exact same cards.
How AI Grading Works
It's not magic. It's math. Here is the 3-step process used by modern grading engines.
Surface Mapping
The card is blasted with light from 12+ angles. Shadows reveal distinct surface imperfections invisible to the naked eye (dents, scratches).
Edge Detection
Algorithms measure the precise pixel distance between the card image and the border to calculate centering to the 0.01%.
Defect Weighting
The computer scores every defect based on severity and location, generating a final aggregated score (e.g., 946 / 1000).
The Future: Hybrid Grading
Will PSA disappear? No. But they will adapt. The future is "AI-Assisted Human Grading."
In this model, the AI does the heavy lifting—measuring centering and flagging surface flaws—while a senior human grader makes the final "Eye Appeal" judgment. This combines the speed and accuracy of robots with the subjective appreciation of art that collectors love.