Here's an excerpt from a recent post by David Berri, one of the Wages of Wins guys, on his blog.
Anyone who believes in the conventional, adhoc methods of valuing basketball players has to answer for the injustice done this year to Kevin Garnett:
I analyzed the 2005-06 season and the player who led the league in Wins Produced was Kevin Garnett. Garnett finished the season with 26 Wins Produced. This marks the fourth consecutive season Garnett has been the most productive player in the NBA. Yet the sportswriters didn’t show KG any love when it comes to the MVP voting. Not a single sportswriter named him on an MVP ballot. He also wasn’t named to the All-NBA First Team, Second Team, or Third Team.
The problem for KG is not his game, but his teammates. In 2002-03 Garnett produced 31.5 wins. The rest of his team produced 15.1. The next season Garnett produced 30.5 wins while the rest of the T-Wolves produced 25.3 victories. The increased productivity of his teammates allowed the writers to notice Garnett and give him the league MVP award.
In 2004-05, as we note in the book, Garnett was essentially the same player who won the MVP. He produced 30 wins, but his teammates only produced 14.8 victories. This past season, just to make life even less fun, his teammates only produced 9.4 wins. So over the past four seasons, Garnett’s teammates have averaged 16.1 wins per year. Although Garnett is averaging close to 30 wins per season, it is not enough to overcome the performance of his co-workers.
This past season no player in the league had less productive teammates. Only two of his fellow T-Wolves posted an above average wins production per-minute played – Wally Szczerbiak and Eddie Griffin – and one of these was traded away during the season. And Griffin was only barely above average.
And that is the tragedy of Kevin Garnett. Year after year he is the most productive player in the league. And year after year he plays with many players who are not only not average, but quite a bit below average.
What are the conventional, ad hoc methods of valuing basketball players?
The NBA's own player evaluation statistic, Efficiency, says Garnett was first.
Posted by: chris m | May 29, 2006 at 12:34 PM
Here's wages of wins comments about the NBA efficiency model:
http://www.wagesofwins.com/WOWCh7.htm
According to them, Antoine Walker was not that good in 04-05 because he made a low percentage of his shots, which they believe is the only thing that matters about scoring.
Surprisingly, this year, Walker went to a good team, the Heat, took less shots, and set career highs in eFG% and TS%:
http://basketball-reference.com/players/w/walkean02.html
Posted by: chris m | May 29, 2006 at 01:04 PM
As pointed out above, the Win Shares analysis doesn't actually do any work in the Garnett MVP case, since he comes out as the best player almost no matter what statistical measure you choose.
This might mean that MVP voters don't get statistics, or that they use flawed norms for evaluating players, but it could also mean that they don't think that the MVP is an award for the best player. In fact, there's a lot of evidence that MVP voters think that the award should go not to the best player simpliciter but to the best player on the best team.
But I wanted to say something about the Win Shares method in general.
On the one hand, I agree that the Wages of Wins stuff is interesting and represents a serious contribution to the statistical analysis of basketball performance. On the other hand, I think it needs to be acknowledged that some of its counter-intuitive results don't show our ad-hoc judgments to be wrong but, rather, show the Win Shares metric to be in need of further tweaking. The most obvious example of this is given by Allen Iverson, but there are others.
The project here is to reform our evaluative practices by introducing evaluative norms that give us a better picture of reality. It's no flaw of the new norms if they push us to change some of our prior judgments, but it is a problem if the judgments they challenge are particularly well entrenched.
When we get that sort of conflict, we need to bring other evidence to bear to resolve the dispute. In the case of Iverson, that evidence seems to show pretty conclusively that the Win Shares formula gets it wrong.
Some of the commenters on other threads have taken this to show that Win Shares is inherently flawed. That's going to far. The question is, where does the formula go wrong, and how can it be fixed?
Posted by: zwichenzug | May 29, 2006 at 03:18 PM
Very interesting post. I would put it this way: that the real value in statistical algorithms is the relative weighting they give us, not the absolute ratings. I happen to believe that Wins Scores are a very useful evaluative tool. Do I believe that Iverson was actually the 91st player (or whatever his precise Win Score ranking was) last year? Or that Rodman was a better player than Jordan in 1997? Not really. But I do believe that Wins Score can help us identify those instances where our intutitive judgements are out of touch with reality. So Iverson is not nearly the player we think he is; and people like Rodman and Garnett are (odd as it seems to say this) over-rated.
Posted by: Malcolm Gladwell | May 29, 2006 at 03:30 PM
Maybe Kevin Garnett's play isn't conducive to having good teammates. Would win production reveal this case?
Posted by: Lucifer's Apologist | May 29, 2006 at 04:02 PM
A possible case for Nash:
Supporters of Nash insist that he makes his teammates better, and point to the fact that other Suns all put up career-high numbers this year playing with Nash. I mean, just look at Tim Thomas in the playoffs.
Perhaps this claim is not as subjective as we think, and maybe it too can be substantiated by Win Score: by comparing the Win Score of Nash's teammates this season, with their previous seasons without Nash?
(although personally, I think it has more to do with D'Antoni...)
Posted by: cho | May 29, 2006 at 08:59 PM
One could also easily surmise that with better teammates, KGs numbers would be down, reducing his wins produced. I am not quite sure how the wins produced number works, but assuming Garnett had better teammates, it is entirely possible that his teammates would eat into his wins produced as they would be instrumental for some of them. That said, I have no idea how KG didn't make it to one of the all-nba teams.
Posted by: Deepak | May 29, 2006 at 09:33 PM
I probably underrate Garnett as much as anyone. For all the stats he produces and all the love-gushing his fans engage in, the bottom line for me is winning. And the bottom line is his teams don't do much in the post season. That is, when his teams even make the post season.
I wish we could see the win percentage of Garnett's teammates in three stages: pre-Garnett, with Garnett, and post-Garnett. As another poster commented, maybe as good as KG is, his game isn't really condusive to allowing teammates to produce good win percentages on a consistent basis.
Also, as another poster commented, maybe KG's exhobitant salary combined with the paucity of drafted talent combine to ensure Minnesota will never be able to surround him with able talent.
McHale is directly to blame for the dearth of 1st round picks (lost in the scandal) and the bad 1st round picks. However, maybe there is something in KGs game that reduces the talents of his teammates. As bad as Cassell was last year, he was stellar in LA this year. Is that a result of the KG effect?
Posted by: Overworm | May 30, 2006 at 09:53 AM
Hmmmm. My ad hoc impression is that Garnett's game makes teammates better, but I'd be open to a statistical tool if anybody's got one.
Re: Cassell -- he was great in his first year in Minnesota, bad in his second. Garnett's numbers were about the same both years. I think it's tough to see Garnett as a causal factor there. More likely, I think, is that Cassell wanted out of town, and played like it. Or, that he just had a bad year. It happens.
Re: Duncan, Garnett, and early career value. There's something to be said for the idea that teams get a better return on their investment in young stars, but that explanation doesn't seem persuasive in this case. Duncan was drafted by a model franchise that had had a down year and then got lucky in the lottery. Garnett, on the other hand, was drafted by Minnesota. 'nuff said.
Back to Win Shares - maybe the right way to think of this metric is as just another number that we use when making what ultimately reduce to qualitative judgments. Useful and informative, but not necessarily definitive.
Posted by: zwichenzug | May 30, 2006 at 10:27 AM
The list of Wolves players who have thrived when they left the Wolves (and Garnett) is a short one: Chauncey Billups. And even Billups first found his NBA game with Garnett by his side. Sure, Cassell had a nice year this year, but Cassell's career is a study in good first impressions and bad exits, and he is an entity all unto himself that probably defies any meaningful correlation with Garnett or anything else besides his contract status.
Based on that (and simply watching Garnett play), it's difficult to argue that Garnett has been depressing the value of his teammates. The reality is that he's a fantastic player who has, with the exception of the Cassell/Sprewell year, hasn't had great teammates. A player who's a primary ballhandler (LeBron, for example) can overcome that to a degree, but Garnett depends on people to get him the ball, and then depends on them to knock down shots when he's inevitably double or triple-teamed. History shows how well that has gone.
It's amazing to me how much grief Garnett gets for not being a "winner". If the Wolves had had even a slightly below average GM over the last 11 years, we wouldn't be having this discussion. Please, for the love of God, can anybody explain why McHale still has a job?
It's to the point, as a Garnett and Wolves fan, that I hope the guy gets dealt to a contender, as long as Kobe's not involved. I'd like the guy to be appreciated for all that he is instead of beaten down for all that he is not. And I'd like to see him win a title.
It won't happen in Minnesota.
Posted by: Doug Hennessee | May 30, 2006 at 12:03 PM
A couple comments:
First, I have trouble feeling too much sympathy for Garnett, because for most of us in our jobs, if we perform exceptionally, but our team fails, we face more dire consequences that being extremely well compensated but not being publically regarded as the best in our fields.
Second, in Gladwell's chat with Bill Simmons (now behind Insider), Galdwell modelstly compared his situation at the New Yorker to the Detroit Pistons, in that he looks good because he is surrounded by people who create a culture of success.
Isn't this one of the main jobs of a star player like Garnett? Isn't he the main shaper of the culture of the Timberwolves? If his teammates consistently underperform, this says something about that culture, and Garnett's role in shaping it.
I don't think objective statistical analysis in basketball is going to get us all the way there. The fact that the system ranks a player whose team has missed the playoffs the last two years as far and away the best player ought to be a clue that the system is off.
It would be nice if this were presented as a possibly useful complementary tool to existing player evaluations, rather than as a replacement for them. I think this is how sabremetricians get a bad name, too -- come out with "bold" proclamations that contradict conventional wisdom (Derek Jeter is a below average defensive SS, Allen Iverson is not an elite player; Kevin Garnett is by far the best player in the NBA).
I will have to check out the book, but a dose of humility would be nice.
Posted by: JohnMcG | May 30, 2006 at 02:28 PM
To put another way, how many of the remaining playoff teams would be better off with Garnett than their best frontcourt player?
Miami -- Shaq? Probably not.
Detroit -- Rasheed? Close.
Dallas -- Nowitzki? No.
Phoenix -- Marion? Not sure.
Posted by: JohnMcG | May 30, 2006 at 02:32 PM
First, Garnett/Rasheed isn't close, it's Garnett by a Minnesota mile. Second, 'best frontcourt player' is deceptive. Shaq and Garnett are both tall, but other than that there's no comparison. You'd really be talkinga bout Garnett for Antoine Walker. Garnett/Marion is Garnett in a walk. Which means that Garnett /Nowitzki is the only exchange that's close.
I also thought that the discussion of Garnett had resonance with the remarks in the Simmons article, but took a different lesson. Namely, that Garnett would have flourished if he had had a better support system. That he's done so well playing for such an awful franchise is a testament to his talent and dedication.
The difference, I think, is that to me the idea that the star player is responsible for setting the tone of the franchise is a little off. Sure, a star player can help (or hurt), but there's only so much you can do when you don't write the checks.
Posted by: zwichenzug | May 30, 2006 at 03:29 PM
I don't want to turn this discussion into a debate on the merits of KG, versus other players, but I guess the originating article set us up for it.
First, I have to agree that any system that says Garnett is the best player in the NBA is faulty from the start. That is simply not true. Not if by "best" we actually mean "the player who has the largest impact on winning games." Shaq is that player.
Shaq took the Magic to the NBA finals. Shaq led the Lakers to three championships. Shaq, in all liklihood, is going to lead the Heat to the NBA finals this season. That's makes three for three. He's played for three franchises, and they've all had the opportunity to play for a championship.
Maybe KG is more like Kobe, in that he generates a lot of stats, makes a lot of highlight reels, but when it's all said and done, they are usually a month into their summer vacation when the NBA finals roll around.
Results, people, results are what count. KG doesn't have them. Kobe, in the two years since he had Shaq shipped out of town, does not have them. Give me results. As Yoda said, "there is no try, there is only do or do not." KG has done not.
And for the record, I think the KG Rasheed comparison is very fair. Rasheed's career numbers could look just like KG's if he played on bad teams his entire career, as KG has. Let's compare.
Here are KG's career averages: 20.4 pts, 11.2 rb, 4.5 ast, 1.7 blk, 1.4 stl. Those are the numbers that people look at when declaring KG a great player. Those numbers and the fact that he's a very tall guy who has some of the skills (re: mobility) of a smaller player.
Let's look at Rasheed's stats from a year where he played with a team that got ousted in the 1st round of the playoffs (like most of KG's teams), the 2001 Portland Trailblazers: 19.2 pts, 7.8 pts, 2.8 ast, 1.2 stl, 1.8 blk. All while averaging 32% from 3pt land. He did all this while playing with better passing guards, a better passing center, which depressed his assists total.
Those averages are remarkably similar, yet few people claim Rasheed is nearly as good (or, gasp, better) than Garnett. Why not? Because KG's style of play is "exciting"?
I love KG, but if I had a choice between Rasheed and KG at their current salaries, I'd take Wallace every year. Rasheed could be as good as KG on the court, but his salary makes is easier to fill in the other pieces.
Posted by: Overworm | May 30, 2006 at 04:25 PM
Garnett is not a star player. He's the greatest compliment to ever play the game but hardly a guy to build around.
He doesn't deliver in pressure games. Garnett is the only max-contract player who at the end of a game, down by one, will pass the ball to Mark Madsen on a pick and pop 20ft from the basket.
This really happened and it speaks volumes about his inability & lack of desire to close out games.
Posted by: Tony | May 30, 2006 at 04:28 PM
I love when articles such as these come out where people I respect show a player I love some respect. There were many interesting comments here but I wanted to touch on a few. Surely the below average GM and ownership of the Wolves has hurt this team from scandal and lost draft picks to bad drafts and simply the fostering of a culture of failure has had a huge influence over my favorite NBA team. To say that Garnett is at fault for this culture is to not know too much about Garnett the man. (check out some of the bio info on Garnett before pointing to his character as a problem)
Statistics, as we know can be flipped, turned, and manipulated to illustrate many views, using the wins produced stat in the ways that Berri does puffs up the importance of this one stat while artificially eliminating the effects of other variables. Like others have said, this stat is perhaps best used in conjunction with other qualitative and quantitative measures.
Posted by: Ben Edwards | May 30, 2006 at 04:38 PM
I am not out to trash or defame Kevin Garnett's character. Eveything I know about him leads me to believe he is an exceptional young man with great passion for his craft.
Because of this, it is tempting to agree with the conclusion of this post -- that Garnett is the best player in the NBA, but has had poor teammates and deserves our sympathy.
I am looking at the objective results, and they do not look like the results of somebody who should be considered far and away the best player in the NBA. They simply aren't.
What other player who was reasonably considered the best player in the league missed the playoffs in consecutive seasons? I understand that the T-Wolves have been managed poorly, but is it so historically inept that it causes the franchise to flounder in spite of having the best player in the NBA by far?
Again, I'm not out to get Garnett so much, as I am what seems like a bad metric. If this metric were truly to gain currency as the measuring tool of performance, used to determine awards and salaries, then it would drive players' behavior. And as other have noted, the result would be great players like Garnett and Bryant passing up shots so that they would not be "inefficient" in their scoring. And nobody wins there.
Hey, who wants to go watch an Oakland A's game?
Posted by: JohnMcG | May 30, 2006 at 04:55 PM
Based on my knowledge of mathematics, such a formula is fallible by nature--because of the inordinate amount of variable in NBA play--referee calls, the quality of teammates. Put Garnett on the Pistons or the Mavericks and his numbers are certain to drop. While the formula does it's best to take these into account, it cannot account for the variables. As such, more qualitative measurements are likely better assessment tools.
Posted by: Tim Spivey | May 30, 2006 at 05:40 PM
I agree, Steve. I think KG would make an awesome 2nd banana. He's a great talent, but he's not cut out to be THE MAN on a team. There's no shame in that, and there's no slight intended by me in saying that.
Posted by: Overworm | May 30, 2006 at 08:59 PM
It would be interesting to see how the NBA's 50 greatest players rank in terms of Win scores. Who would come out on top?
Posted by: PJ | May 31, 2006 at 01:44 AM
It is really great to see all this interest in our work. Malcolm asked me to respond to some of the comments people have made, and I will do my best to address some questions.
One issue raised was how well one can forecast the future with Wins Produced. Young Hoon Lee and I have worked on a study looking at this issue. In the Wages of Wins I comment briefly on this research: “... Lee and Berri (2004) examined the relationship between the number of wins a team achieved and the productivity of its players in the previous season. This work indicates that between 65% and 75% of current wins can be explained by what a team’s players did in the prior season. In other words, knowing past productivity allows one to predict better than one could if all you knew was the current salary players were scheduled to be paid.”
So if you know what a team's players did last year, you have some idea how the team will do in the future. I comment on this in more detail at our blog, The Wages of Wins Journal (dberri.wordpress.com). At our blog I also have tried to offer few more answers to some of the other questions people have raised. If you get a chance, take a look at some of these posts. If you still have questions, I will try and address these as best I can.
Posted by: dberri | May 31, 2006 at 02:03 AM
I have not read the book, but isn't it impossible to have a team with players that are all ranked high on this statistical scale? You can't have a team with all superstars. There are always going to be one or two players that stand out on a team. Alot of this has to do with salary, but a great deal of it has to do with that all players can't score 35 points a game. If Garnett gets 20 rebounds in a game, he is basically taking those rebounds away from oppents and his own team. So in a sense other players on his team are sacraficing their stats because another player increasing theirs.
Posted by: Jason | May 31, 2006 at 09:11 AM
To pose a similar question to one posed earlier, but this time using it to gauge the quality of Garnett's teammates
Here's the rest of the wolves starting five for the last quarter of the season
SG Rashad McCants
SF Ricky Davis
PG Marcus Banks
C Mark Blount
Would you want any one of these players to replace any current starter in the NBA playoffs?
I think not. KGs teammates are currently not helpful and the organization is awful. The Lakers have a much better starting five than the Wolves and this is why they won more games than the Wolves, not because Kobe takes big shots.
Also, see game 7 against the Kings in 2004 playoffs to see KG deliver in big games.
Posted by: Nate | May 31, 2006 at 11:01 AM
YES, I'm A MINNESOTAN - I love it - give me more and more and more and more and more...
Posted by: Jamey Johnson | May 31, 2006 at 12:51 PM
Jason,
You make a great point about diminshing returns. As we note in the book, as a player plays with more productive teammates, his productivity will decline somewhat. This point is important when people try and contruct Olympic teams. There is only one ball. When you put five "stars" on the court it is important to ask the question: If this player can't score, because after all there is only one ball, what else can he do to help a team win?
We spend a part of the book discussing the Law of Diminishing Returns. Not only does this law apply to the NBA theoretically, but we also present empirical evidence supporting this notion.
Posted by: dberri | May 31, 2006 at 01:58 PM