So originally i just wanted to make a post like a normal person, but after seeing a few people quite vocal about not using wins for self-buy prices i thought- hey, this actually sounds like something interesting and fun to mess around with. Too lazy for fancy formatting, though.
So i took records from some lower tier PLs (would have liked to use PUPL 6 too but for some reason that is good and exists it is set to not allow copying so) - win counts, differentials, prices. The numbers are correlation coefficients, here it should be like at least 0,2-0,3 to be relevant, depending on the sample size. The left column is correlation between cost of the player and their win count / cost and their differential - this was mostly just for fun, but the takeaway is that more expensive players generally will get more wins, but not necessarily a better differential.
Middle column is about the correlation of win count/differential in previous iteration with price/differential in the next iteration of the tournament. Obviously, only the players that participated in both iterations are counted. Both higher win count and differential mean generally higher price in the next iteration for the tournament, but win rate is a noticeably better indicator. As for wincount in previous tour - differential in next and differential in previous - differential in next -- well, it's something i was curious about and it seems that someone's good or bad record in one iteration of the tournament means nothing for the next one. Fun fact.
Don't take this too seriously, it's not really my.. area, but what i'm trying to say is: if the aim is to make manager self-buy prices as close as possible to what they would have cost in an auction, wincounts are better than differential. Like, i feel that this is pretty obvious - someone who is generally considered to be a good player and trusted enough to not be subbed out even if they have a bad season and end up 3-7 will still likely get bought for a decent price, while someone who gets a nice 3-1 debut will still probably be relatively cheap. And of course for like a 1-0 sub having the same price as a 4-3 player makes no sense at all. The records are not the sole reason of why these players were bought - it's just an indicator.