Statistical Analysis

STAT: Berri and Winston drop tables

Photo by Julie Goodrow

You might have seen my recent Truehoop posting on the Wolves’ existential crisis. As I mentioned in the story, I gathered more information than could fit into an article any normal non-Wolves fan could possibly be expected to read. But you people, you are different. So over the next few days, I’ll be giving you guys a look at some of this excess in greater depth. As I said, David Berri of Wages of Wins and Wayne Winston were extremely generous with their time and efforts in helping me out. So below you’ll find some of the info they passed along to me, in table form.

Children of Production

First is Berri. Using his Wins Produced and Wins Per 48 metrics, Berri ran an analysis of how much players might be expected to play given their age, position and levels of production. As I discussed in the story, Berri found that their were certain discrepancies between which Wolves were producing and which were playing. Those results (through the games of December 11) are shown below:

Name Games Minutes ActualMGM Predicted
DifferenceMGM Differencein Total


WinsProduced WP48
Kevin Love 24 834 34.8 39.5 -4.7 -113.4 8.3 0.477
Luke Ridnour 19 539 28.4 27.8 0.6 11.4 1.5 0.134
Anthony Tolliver 19 356 18.7 24.6 -5.9 -111.4 0.4 0.049
Corey Brewer 24 595 24.8 19.1 5.7 135.9 0.3 0.021
Lazar Hayward 15 112 7.5 13.8 -6.4 -95.5 0.0 -0.003
Maurice Ager 4 29 7.3 17.0 -9.8 -39.0 0.0 -0.064
Sundiata Gaines 7 64 9.1 20.0 -10.9 -76.2 -0.2 -0.171
Michael Beasley 22 732 33.3 26.2 7.1 155.4 -0.3 -0.017
Wesley Johnson 24 691 28.8 21.0 7.8 186.1 -0.3 -0.022
Kosta Koufos 20 140 7.0 13.5 -6.5 -130.6 -0.5 -0.180
Sebastian Telfair 24 521 21.7 22.1 -0.4 -8.5 -0.5 -0.049
Wayne Ellington 19 352 18.5 19.5 -1.0 -18.1 -0.5 -0.074
Nikola Pekovic 15 214 14.3 10.8 3.5 52.2 -0.7 -0.164
Darko Milicic 23 605 26.3 24.1 2.2 51.5 -1.0 -0.081

But this is what I find kind of interesting:

These results appear to reveal a missing variable in the model (so the minutes model isn’t perfect).  We know from past published research that high draft picks receive more minutes than their performance suggests.  Beasley, Johnson, and Brewer are all high draft picks.  So I think if we considered where a player was drafted, the allocation of minutes by Rambis would line up with what a “typical” NBA coach would do.  In other words, typical NBA coaches are influenced by scoring and draft status of a player.

So Berri concludes that, aside from Love’s playing time (and we all know that in the last few weeks, Love’s minutes have been inching closer to 40/game), the Rambis isn’t doing anything too out of the ordinary. This is not exactly good news:

A bigger problem, though, is that his teammates are not very good.  An average player posts a WP48 of 0.100.  Of Love’s teammates, only Ridnour is posting a WP48 that is above average.  And ten of Love’s teammates are producing in the negative range. So the T-Wolves have one amazing player and a large collection of players that are not helping.  Given this roster, there really isn’t much Rambis can do to change the outcomes we observe.

Minus Men

Winston’s metrics work considerably differently. Rather than developing a model of production (based on points, rebounds, shooting efficiency etc) Winston measures how well teams perform when given players and lineups are on the floor. What separates this from your run-of-the-mill efficiency metrics, though, is the fact that Winston adjusts for the quality of an opponent. In other words, a team’s starters are theoretically not punished by the fact that they routinely play against their opponents best players. Let me show you numbers from some of the Wolves’ most frequently used lineups (these are also calculated through the games of the 11th). Winston tells me that the essential number is the one on the far right, which is the given unit’s adjusted +-.

What we find is pretty much what Winston said we would find. The team does not do well when Love, Darko and Beasley play together. They are even worse than that when Wayne Ellington or Kosta Koufos plays. They are pretty good when Anthony Tolliver and Corey Brewer are injected in to the mix. (By the way, there are lots more of these but I didn’t want to hurt your eyes).

MIN   106.85 minutes
Beasley       Johnson       Love          Milicic       Ridnour -13.19

MIN   57.01 minutes
Brewer        Johnson       Love          Milicic       Ridnour 16.20

MIN   40.83 minutes
Beasley       Brewer        Johnson       Love          Ridnour  2.28

MIN  34.90 minutes
Beasley       Brewer        Love          Milicic       Ridnour   -10.76

MIN     27.91 minutes
Beasley       Brewer        Johnson       Love          Telfair 30.69

MIN   18.55 minutes
Beasley       Johnson       Milicic       Telfair       Tolliver  33.01

MIN      14.08 minutes
Brewer        Johnson       Love          Ridnour       Tolliver 16.66

Look at the effect of replacing Beasley with Brewer and/or replacing Love or Darko with Tolliver. Winston reports that when Love and Darko play together the Wolves are -9.03. When Brewer and Johnson play together, on the other hand, the Wolves are +9.06. Tolliver with Beasley: +8.56. Tolliver with Johnson: +7.39. Beasley with Love: -6.5. B-Eazy with Darko: -8.7. Sobering stuff. Now you know.

Share this because Rubio would pass this along:
Tagged , , , , , , , ,

0 thoughts on “STAT: Berri and Winston drop tables

    1. Not sure I understand what you mean by “pointless.” I actually think the Winston stats in particular are pretty meaningful and in many cases line up with what we actually see on the floor.

  1. First of all, this is awesome so keep it coming. Those of us who are matematically illiterate appreciate you breaking it down to a grade-school level.

    Second, the “wins produced” column in the first table is horrifying. Horrifying! Please post a disclaimer before putting such graphic content on your blog.

  2. Ugg, I may have to rethink how horrified I am about Ridnour playing big minutes and transfer that angst to a few other wolves.

Leave a Reply