That is what I was talking about. From Scot Reader's book:
Thus, if one assumes that roughly 200 of each of these rare subjects remain in existence, and further assumes that survival rates for these subjects conform with those of other subjects, the surviving quantity of a typical 350-only subject is somewhere in the neighborhood of 3,000. Finally, assuming for the sake of rough-and-ready calculation uniformity of survival among series, the total number of T206 specimens in existence today is estimable in the general vicinity of 1.6 million. Of course, if there are 400 examples each of Demmitt (St. Louis) and O.Hara (St. Louis) extant instead of 200, the presumed number of T206 cards with us today doubles to 3.2 million, or about 6,000 per subject on average, under this same analysis.
In any case, of the likely hundreds of millions of specimens initially produced, it seems highly probable that the number of T206 cards in existence today is in the low singledigit millions, or a few thousand for a typical subject. This is quite possibly less than one percent of the original production, with the vast majority of these survivors being in lower grade.
If Red Cobb's were double-printed, you'd be at about 10,000 for those. And then you could deduce from relative scarcity how many of the others might be there. Again, this is all rife speculation, but I like his analysis.
You can read the whole darn thing here (and yes, it takes into account relative numbers of cards graded by PSA, etc.):
http://www.oldcardboard.com/t/t206/I...-3-edition.pdf