“Assuming I’ve got everything straight, the Browns gave up their #1 and #3 picks in 1985 along with a 6 in 1986 to switch first round picks with the Bills in ’86. ”

Pretty much right, except the Bills didn’t have to give up their 1986 1st rounder. Making the trade:

Bills give up:
1985 Supplemental #1 (which they wouldn’t have used had Kosar not been available)

Browns give up:
1985 #1 (Browns originally tried to trade Chip Banks instead, but he refused to report to Buffalo)
1985 #3
1986 #1
1986 #6

So the idea is that the regression allows for both the QB to potentially be overvalued and/or the WR. I think the pairs are really interesting to think about, but there aren’t that many of them. My setup means they have to be in the draft in the first-round in the same year for both QB and WR/TE to be on each side of the regression.

Here are the examples from the tables above (only including players who have at least five seasons logged already): Mirer/Smith (ND, ’93), Collins/Brady (Penn St., ’95), Manning/Nash (Tennessee, ’98), and Russell/Bowe (LSU, ’07). At most one of the pair succeeded relative to draft position in each case, but certainly we wouldn’t want to say too much with this sample.