Win Score: Statistical models for player evaluation

[GSW2K4]

I was looking through some draft prospects today and came across this win-score when looking at JR Giddens.

I looked into it more and it seems like an intriguing stat though I'm not sure what to make of it. Essentially it is designed to evaluate the player's value by looking at their contribution to team wins.

Here's the description:
http://dberri.wordpress.com/2006/05/21/ ... rformance/

So I took a look at this past year's stats and something that jumped out at me was where this stat puts Rajon Rondo in relation to other guards:

http://www.draftexpress.com/stats.php?s ... G&qual=all

It seems as though it would be difficult to judge what Rondo contributes to wins if surrounded by KG, Allen, Pierce, etc. Though the economists who designed this stat said we shouldn't try to disprove its worth by pointing out individual players (Marion had a score than Nash one year), I find this to be an interesting dilemma.

What good is it to calculate the number of wins a player contributed to a team if you can't adjust for the fact that the team as a whole dominated its competition?

What could be done to fix this?

[cwas2882]

Rondo jumped out at you? Sessions jumped out at me. 6th on the list.

[GSW2K4]

Sessions only played 17 games, of which they won 2... so I just ignored that due to small sample size.

All his numbers are misleading -- +/-, eff, per...

[drza]

Here's the description:
http://dberri.wordpress.com/2006/05/21/ ... rformance/

So I took a look at this past year's stats and something that jumped out at me was where this stat puts Rajon Rondo in relation to other guards:

http://www.draftexpress.com/stats.php?s ... G&qual=all

It seems as though it would be difficult to judge what Rondo contributes to wins if surrounded by KG, Allen, Pierce, etc. Though the economists who designed this stat said we shouldn't try to disprove its worth by pointing out individual players (Marion had a score than Nash one year), I find this to be an interesting dilemma.

What good is it to calculate the number of wins a player contributed to a team if you can't adjust for the fact that the team as a whole dominated its competition?

What could be done to fix this?

I think the point of the stat is that you can quantify the individual contributions of each player independently, and if the measure is operating correctly then the combined scores of all of the individuals should add up to roughly the number of team wins. For example, check out his review of the Nuggets this year:

http://www.wagesofwins.com/Denver0708.html

Camby had 21 wins produced, Iverson and Melo around 9, on down to Von Wafer who had negative wins produced. Add those numbers up, and the stat would have predicted that the Nuggets would have had 51 wins total based upon the individual contributions. The nuggets actually had 50 wins, which was close to the estimate (He only has 3 team write-ups done currently, with the Nets (27.6 wins predicted, 34 wins actually) and the Hawks (36.3 wins predicted, 37 wins actually) as the other 2. The other 27 are forthcoming).

So it's not that the measure erroneously doesn't adjust for the strength of teammates...just the opposite, in fact. The measure is meant to quantify exactly how strong each individual is, to objectively be able to compare a player's contributions to team wins without being biased by how good or bad the rest of the team is. And by comparing the actual to the prediction, you can get a quick sanity check for whether the measure makes sense or not.

[shrink]

Nice post drza. As evidence, Garnett used to be the best in the NBA in win-shares every season for a mediocre team, simply because his teammates were so ineffective.

[Buck You]

Sessions only played 17 games, of which they won 2... so I just ignored that due to small sample size.

All his numbers are misleading -- +/-, eff, per...

I don't think his numbers were misleading. He is a very good player.