Dre at Wages of Wins has a post claiming that parity is impossible in the NBA. I decided to be a little contrary and see if it were possible (and I see now that people in the comments have suggested doing similar things to what I’ll do below). Dre’s claim is that since the spread in ability amongst the top players is bigger than the spread amongst good or average players, any team who has a top player will have to be good, and in fact better than teams without a top player. Since some teams have managed to acquire two or three top players, there is no way parity can happen.
Parity, of course, hasn’t happened for two reasons: teams have some flaws in evaluating players, and teams are differentially willing to spend money to get the best players. That has led to a system where last year teams won between 17 (Minnesota) and 62 (Chicago) games. What if we used a different system? We could, for example, have a draft every year for every player. In 2011, there were 452 players. With 30 teams, that neatly adds up to 15 players per team with two left over (two teams get 16 players). Let’s say that teams used Wins Produced as their measure of talent.
I ordered all players last year by the number of wins they generated (from Kevin Love’s 25 to Andrea Bargnani’s -6.5), so there’s also some amount of knowing what would happen (i.e. who would play well and who would play poorly, although note I didn’t change their minutes; whoever got Bargnani would still decide to give him 2353 minutes). One way to draft would be your typical fantasy snake draft: team 1 gets picks 1 and 60, team 2 gets 2 and 59, and so on until every player is taken. If you do that and add up team wins, they range from 31 to 52.6, and wins increases neatly with team number: the team that picks first in the first round does best, down to the team that picks last in the first round.
So there’s something to Dre’s claim; even when you try to even things out a bit by having a snake-style draft, the top team does best. That’s because in the first round (i.e. the top 30 players) there’s a spread of about 15 wins produced whereas in the second there’s only a spread of about 2. The later-picking teams can’t make up that big initial difference. Instead of a snake draft, we could have teams pick in one order in round 1 (i.e. 1 to 30) then reverse order in every subsequent round (i.e. 30 to 1). In that case the spread of wins is only 36 to 46, and that’s only because team 29 (for some reason) drops to 36; every other team is 39 or higher. If you used the opposite order in every subsequent round, the 14 other players on the team will come close to making up for the benefits of the better players taken in round one.
I’m sure you could do better than this; have a more systematic way of matching wins and so on. If we really wanted to assume that teams knew how good their players were, we would also change how many minutes they played; Bargnani would not get 2300 minutes any more. But using a fairly simple system you can pretty close to total parity; the spread from 36 to 46 is a difference of less than 2 points per game from average in either direction. If you threw in some luck and injuries across seasons, I would guess this would be not significantly different from every team being average. So, if teams were better evaluators of talent and willing to spend for it, parity is at least theoretically possible.