PreGradeCards Internal Research — Published May 2026, Updated June 2026

AI Grading Accuracy Benchmark Study
n=10,000 Cards vs PSA Final Grades

PreGradeCards tested its computer vision model against final PSA certification grades on 10,000 cards submitted between January 2025 and May 2026. This page presents the full results, methodology, limitations, and per-category breakdowns — all first-party data.

89%
Within 1 Grade Point
of Final PSA Grade
99.2%
Centering Measurement
Accuracy
82%
PSA 10 Binary
Prediction Accuracy
10,000
Cards Tested
Jan 2025–May 2026

Key Findings

89% within ±1 grade point

Across all 10,000 cards tested, PreGradeCards AI predictions fell within one grade point of the final PSA certification grade 89% of the time. For context: expert human pre-graders typically achieve 72–78% accuracy on the same task.

99.2% centering measurement accuracy

Centering ratio predictions (left/right and top/bottom border percentages) matched physical ruler measurements within 2 percentage points on 99.2% of cards. This compares to typical human centering estimates at 85–90% accuracy.

82% PSA 10 binary prediction accuracy

When asked to predict "will this card earn PSA 10 or not?", the AI was correct 82% of the time. False positive rate (predicting PSA 10 when card received lower): 11%. False negative rate (predicting below PSA 10 when card earned it): 7%.

Surface flaw detection: 76% sensitivity

The AI detected visible surface flaws (print lines, scratches, dimples) with 76% sensitivity. Surface is the hardest criterion to assess from 2D scan images due to lighting dependency — this figure improves to 91% with dual-angle scan input.

Corner accuracy: 93% within ±0.5 subgrade

Corner condition predictions matched BGS-equivalent subgrades within 0.5 points on 93% of cards, making corners the most reliably predicted criterion.

Worst performance: vintage pre-1960 cards

On vintage cards (pre-1960), overall accuracy dropped to 71% due to unique aging patterns, paper stock variation, and surface oxidation not well-represented in the training set. Modern cards (2010–2026) achieve 93% accuracy.

Accuracy by Card Type

Card Category Cards Tested ±1 Grade Accuracy PSA 10 Prediction Notes
Modern Chrome (2010–2026) 3,841 93% 86% Best performance category. Consistent manufacturing = consistent AI patterns.
Pokémon TCG (1999–2026) 2,203 91% 84% CGC Pristine 10 training data added June 2025 improved TCG accuracy 8%.
Basketball Cards (all eras) 1,290 88% 80% Strong performance on Prizm/Select. Vintage pre-1980 drags average.
Baseball Vintage (1960–1979) 887 79% 74% Borderless Topps designs require special centering calibration.
Football Cards (all eras) 742 87% 79% Panini Prizm dominates sample. Consistent with basketball results.
Vintage Pre-1960 612 71% 62% Lowest accuracy. Paper aging, wax staining patterns poorly represented in training data.
Magic: The Gathering 425 88% 81% CGC-graded reference cards used. Black border wear detection strong.

Methodology

Study Period: January 1, 2025 – May 31, 2026  |  Sample: 10,000 cards  |  Reference: Final PSA certification grade

Card selection: 10,000 cards were randomly sampled from PreGradeCards users who (1) analyzed their card using the PreGradeCards AI tool, and (2) subsequently submitted the same card to PSA and shared their final PSA grade. Cards were not screened by expected outcome — all submitted results were included.

AI prediction method: Standard front+back scan using the PreGradeCards web interface. No special preparation by users. The AI model version active at time of submission was used (model updated 3 times during the study period; version-level accuracy is reported above in the card-type breakdown).

Accuracy definition: "Within 1 grade point" means the AI predicted a grade within ±1 of the final PSA grade (e.g., AI predicted PSA 9, card graded PSA 8, 9, or 10 = accurate; AI predicted PSA 9, card graded PSA 7 = inaccurate).

PSA 10 binary accuracy: Separately measured as a true/false classifier: did the AI correctly predict whether the card would achieve PSA 10 Gem Mint or not? This is the metric most relevant to collectors deciding whether a card is worth the submission cost.

Limitations: Self-selection bias — users who submitted to PSA after using PreGradeCards may have higher-confidence cards than the general card population. Scan quality varies by user. Cards analyzed after PSA grade is known are excluded from this study.

How PreGradeCards Accuracy Compares

Method ±1 Grade Accuracy Cost Per Card Speed
PreGradeCards AI This Study 89% $0.19 Instant
Expert Human Pre-Grader 72–78% $5–15 5–10 min/card
Self-Grading (Collector) 55–65% $0 10–20 min/card
PSA Submission (Official) 100% $79.99+ 65+ days

Try the AI That Achieved These Results

10 free credits, no credit card required. See your card's predicted PSA grade in seconds.

Grade a Card Free → Download JSON Data Cite This Research