Someone fell prey to the small sample size trap, and the problems with comparing percentages over same.
33% of 5'7" draft picks (-5) outperformed their draft position! But only 8% of the 6'5" draft picks (+5)did!
Of course, if you look at the raw numbers, that's 33% of... 3. Which means
1 draft pick at that size outperformed their position. Meanwhile, that 8% is of 48. Which translates to 4 players.
Seriously, when you're comparing 1 to 4, even as percentages, your data is not statistically significant.
The only range in which you're really comparing similar numbers is 5'11" through maybe 6'3", and the percentage differences look like they're all within a single standard deviation, which means you have absolutely no signal coming out of that analysis whatsoever.
Binning the columns to 1" of height each is really rather silly, and seems designed to inflate the percentages on the lower numbered columns. For instance, that third column at -4 has 18 total draft picks. The percentage on that is about 17%, which translates to: 3 players. If we total up the first three columns, you get 1+1+3 out of 3+3+18, or 5/24, which translates to ~21%. Which isn't nearly so impressive as having two columns at 33% each.
Even the tremendous 40+% of the -3 column starts falling apart when you look at the top of the column and see that's out of 52 total. You can add the bottom four columns (-6, -5, -4, -3) together (total is 76) and you end up with fewer than the +4 column on its own (103).
If that guy really has a Masters of Business Economics like his Twitter profile says, well, I'd say it terrifies me about what they teach in economics degrees these days, but really, it would explain a lot about.
The funny thing is, I don't even necessarily think his hypothesis is necessarily incorrect (that smaller prospects are undervalued), but his methodology is such utter garbage, it drives me nuts.
but i'm willing to take that chance.
