Transparent measurement
AI card grade accuracy
CardRevive does not publish a guessed “accuracy” percentage. This page uses paired AI estimates and returned professional grades only after evidence review and separate anonymous-aggregate consent.
Building the verified cohort
There are currently 0 qualifying paired outcomes. CardRevive withholds public percentages until at least n=20; a smaller sample would look precise without being dependable.
Definitions and cohort rules
- Exact-grade accuracy
- The AI final estimate equals the returned numeric grade on the normalized 1–10 scale.
- Within-one accuracy
- The absolute difference between the estimate and returned grade is no more than 1.0.
- Qualifying outcome
- A valid PSA, BGS or TAG result with separate aggregate consent, an earlier company-specific AI estimate, and certificate or slab-label evidence reviewed by CardRevive.
- Duplicate and invalid handling
- One outcome per card and company; duplicate certificate fingerprints, rejected evidence, impossible grades, missing predictions and withdrawn consent are excluded.
Cohort dates: Not yet available to Not yet available. 0 supplied rows were excluded from this calculation. Breakdowns require at least n=10 in each displayed cohort.
Limitations and possible bias
These results describe this self-selected cohort. They are not a random sample of all CardRevive estimates, cards, or customers and should not be treated as a guaranteed result for another card.
- Self-selection. Owners choose whether to submit a card, report its result, and separately consent to anonymous aggregation. People who make those choices may differ from people who do not.
- Outcome-return and nonresponse bias. Only cards that were submitted, returned and reported can qualify. Cards that were never submitted, have not returned, or were not reported are absent, so their outcomes may differ.
- Verification limits. Evidence review helps exclude invalid and duplicate records, but it does not remove self-selection, nonresponse, or human-review error.
- Reporting incentive. CardRevive awards one grading credit for every ten returned-grade reports. That incentive may affect who reports or how often; evidence review and separate aggregate consent are still required for inclusion here.
- Small samples. The n=20 overall and n=10 breakdown minimums are publication thresholds, not guarantees of representativeness, stability, or statistical significance. No confidence intervals or significance tests are shown.
Contribute a real outcome
Grade a card, then add its returned result from the private report. Evidence stays private unless you separately allow publication.