[sphere and ash]

The article is not just pointing out authentication errors--no one expects 100% accuracy--but the ONE-SIDEDNESS of one authenticator's errors. Every authenticator has an error rate, and if you are really conducting analyses these errors should be randomly distributed. The whole point of the article is that the errors here are not randomly distributed, but distorted by bias (everything from one source is good, everything from another source is bad).

I would bet you will find similar biases in card grading. I have never seen a grading population report, but I would bet it does not fit a normal (i.e., bell curve) distribution. My hunch is that large, favored customers cause grade inflation at the highest grades.

The question for anyone buying anything (and particularly autographs, because the rate of fraud is higher) is this: can you survive your mistakes?