Monday, September 24, 2007

A rebuttal to Wins Produced (not by me)

In evaluating players, I've relied on Wins Produced and Win Score a few times. While I've never really gotten around to deciding how good a metric I thought it was, I definitely viewed it favorably. It seemed to reflect my view that most commonly used tools for evaluating players over-emphasized scoring to the detriment of other parts of the game (rebounding, steals, etc.). Plus, quite frankly, Dave Berri puts forward a number of interesting observations about the game of basketball based on this metric, and those observations often struck me as being very sound.

That said, via TrueHoop comes the tidbit that some statisticians think his work isn't so hot. I have no false belief that my statistical knowledge is good enough to figure out on my own who has the better of the argument (I remember my aborted attempt to do a statistical analysis last month that still needs to be revisited), but I am looking forward to seeing the negative analysis being put forward by his peers. At the very least, reading the forthcoming paper seems like it will increase my knowledge of the world of basketball statistics and the major schools of thought currently being investigated. If this sort of thing interests you, I recommend clicking on the link above and following the linked post to all the relevant articles.

3 comments:

Louis said...

Having done some statistical research on basketball myself I can say with some sense of clarity that the numbers in basketball, like any other statistic, can be made to say whatever you want them to say. Almost twenty years ago I tried a combination of points, steals, offensive rebounds, defensive rebounds, assists, minutes played, and salary and found that the best player per dollar on the 76ers was Scottie Brooks. He was, according to my rankings, significantly better for the team than Charles Barkley. My statistics were clear and equally applied but I did not write the team with my important findings because I interpreted my statistics to mean absolutely nothing.

Sam Cohen said...

I think you're being unfair to statistics. If you're using statistics appropriately and honestly, they give you valuable information about activities in the aggregate. Minimally, they give you another tool to help your evaluations. Scottie Brooks very well could have been the best "value" on the Sixers per dollar, especially if salary variation between the top and bottom player is greater than the per minute production between those players. Just because we care more about the best players than the best players per dollar (less true now because of the salary cap) doesn't mean what you came up with was necessarily wrong or unhelpful.

Louis said...

Though my information may have been true, and maybe even helpful to a front office that was only concerned with the dollars, the fact is that my statistics that showed Brooks to be the best value are almost certainly different than his wins score relative to the other players. My statistics are true, their statistics are true, but then taken together how do you know which way to steer a team’s personnel decisions? Maybe there should be a wins-per-dollar statistic that combines them and allows GMs to balance how many wins they need before the trade deadline with how much room they have under the salary cap.