Scoring totals and team wins. Or team wins and scoring totals. As observed a few days ago, it’s these two factors that primarily determine the media’s choice for Rookie of the Year. And two years ago I noted it is team wins and scoring totals that primarily determine the media’s choice for Most Valuable Player.
Given this research, it was pretty easy to predict who the media would consider for MVP. At the top of the scoring list is Dwyane Wade. But his team only won 43 games, and that’s usually not enough to top the MVP list. Next on the scoring list are LeBron James and Kobe Bryant. People expected the contest between these two players to be close. But LeBron did finish with a higher scoring average and his team had a slightly better record. So we should not be surprised that LeBron James was named the 2009 MVP.
But what would happen if we looked past a player’s scoring and team wins? After all, scoring totals are not a perfect measure of a player’s impact on team success. And when voters consider team wins they are not completely separating a player from his teammates. In sum, the approach taken by the media is less than perfect.
The Amazing Chris Paul
One problem with discussing the MVP award is that the term “Most Valuable” is undefined. Certainly it is possible that one could define “Most Valuable” as the primary offensive option on a winning team. Of course, one suspect that such an explicit definition would not be accepted by everyone.
A better approach (well, at least a different approach) is to focus on identifying the “Most Productive” player. In other words, let’s look at Wins Produced. Such an approach will consider much more than just offensive contribution, and furthermore, do more to separate a player from his teammates.
For example, consider Chris Paul. Chris Paul’s 29.4 Wins Produced clearly led the New Orleans Hornets (and the NBA). The Hornets, though, only won 49 games this year. And when we look at efficiency differential (offensive efficiency minus defensive efficiency) and Wins Produced we see a team performance consistent with only 45 wins. In sum, the Hornets were not much beyond average in 2008-09.
This outcome, though, should not diminish the season of Chris Paul. In the 21st century only one player – Kevin Garnett – has managed to produce more than 29 wins in a single season (KG did it three times with Minnesota from 2002-03 to 2004-05). And one has to go back to the days of Magic Johnson to find a point guard that was this productive.
Unfortunately, Paul’s teammates didn’t help much. After Paul the only above average players on the roster were Tyson Chandler [Wins Produced per 48 minutes (WP48) of 0.117] and James Posey [0.110 WP48]. Average is 0.100, so it appears there was little else on this team besides Paul. We can see this clearly when we consider the combined productivity of everyone else on the roster. Once we move past Paul’s 29.4 Wins Produced, we see a roster that combined to produce only 15.5 wins. This means that if Paul was replaced by an average point guard – think of someone like Rafer Alston [0.106 WP48], Derrick Rose [0.104 WP48], or D.J. Augustin [0.094] – the Hornets’ Wins Produced would have only been 21.8 this past season. In sum, moving from Paul [0.470 WP48] to an average point guard [0.100 WP48] would have cost this team about 23 victories
LeBron and Kobe
What happens if we analyze LeBron James and the Cleveland Cavaliers in the same fashion?
Cleveland’s Wins Produced was 64.6, with 27.1 of these attributed to LeBron. Replacing King James with an average small forward – think Danny Granger [0.105 WP48] or John Salmons [0.093] – would cause the Wins Produced of this team to fall to 43.9. This would still be good enough to make the playoffs in the East, but obviously removes Cleveland from title contention.
Although Cleveland would clearly not be as good without LeBron, the Cavaliers do have other above average players. For example, Ben Wallace [0.198 WP48], Anderson Varejao [0.155 WP48], Delonte West [0.144 WP48], Maurice Williams [0.138 WP48], Zydrunas Ilgauskas [0.133 WP48], and Wally Szcerbiak [0.120] all helped LeBron win this award. In other words, if LeBron was blessed with Paul’s teammates, it is unlikely King James would have been name MVP.
What about Kobe? The Lakers finished with 61.2 Wins Produced. Of these, 13.9 could be connected to Kobe. Such a mark ranks in the top 15 in the league, but clearly falls far short of the marks posted by Paul and James. And if you replaced Kobe with an average shooting guard – think Kevin Martin [0.096 WP48] or Jason Terry [0.092 WP48] – the Lakers would have still posted 53.4 Wins Produced. Again, such a mark would probably not be good enough to contend for a title this season. But it would be good enough to make the playoffs (and maybe advance to the second round).
It certainly appears the Lakers are about more than Kobe. Pau Gasol [0.269 WP48], Andrew Bynum [0.198 WP48], Trevor Ariza [0.192 WP48], and Lamar Odom [0.181 WP48] combined to produced 39.5 wins in 2008-09. And before anyone thinks that Kobe was the reason these players were so productive, let me note that Gasol, Ariza, and Odom have all posted above average numbers on teams that did not include Kobe.
So here is the story. The Hornets without Paul are essentially not much better than the LA Clippers [17.7 Wins Produced], Sacramento Kings [17.9 Wins Produced], or Washington Wizards [20.9 Win Produced]. Cleveland and the Lakers, though, might still be playoff teams without LeBron and Kobe. And I think this suggests that Paul – a player that nearly 30% of MVP voters did not rank in the top five in the league — should have received at least a bit more consideration for MVP.
Let me close by noting that this is not going to be my last post on the topic of MVP. My next column is going to focus on the MVP on each team in the NBA. And I hope to follow that column with an examination of Kobe and Dwyane Wade (quick preview… Flash tops Kobe in overall productivity and the difference is rather large).
– DJ
The WoW Journal Comments Policy
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
Wages of Wins Journal 
brgulker
May 8, 2009
Thanks, Dr. Berri. I’d been looking forward to your analysis on this since the award was given out.
I was surprised to see how many wins LBJ did actually produce and how few KB produced.
What do you think of the phrase that gets tossed around so much: “best player on the best team…”
If we assume that is at least part of what the media considers this year, they got it right, didn’t they?
Andrew Alexander
May 8, 2009
I think that LeBron James should be credited for raising the play of his teammates.
They all benefit from LeBron immensely, from shooting wide open three pointers to getting easy layups.
I know that you have looked at Mo Williams pre and post LeBron, but could you look at the other players as well? And then compare them to Chris Paul’s teammates?
mrparker
May 8, 2009
cue up all the assist inflation arguments…id like to see what paul’s wp48 would be if we took away 2 or 3 assists a game just to give everyone an idea of how good he would still be.
Michael
May 8, 2009
So Lebron finished with a WP48 of (I estimate) 0.426, and Paul one of 0.470? And Lebron’s team had 66 wins to Paul’s 49? And Lebron finished on top in PER (best non Michael Jordan season ever), Win Shares (best non Michael Jordan season ever), and Statistical Plus Minus?
I agree that Chris Paul deserved a lot more attention, but his team sucks.That said, I’m more than comfortable with the media’s choice this year, Lebron’s been historically good.
Ray
May 8, 2009
I have a question regarding the WP48. In Dr. Berri’s mid season analysis of every player http://www.wagesofwins.com/AllPlayerMid0809.html there are certainly some interesting people in the top 10. I know WP48 is a measure of productivity, not necessarily talent. However, with that being said, if Dr. Berri is starting an NBA team, is he going to take Marcus Camby over Kobe or Wade?
I’m well aware that Camby is a really good player, and I’m not arguing that his WP48 doesn’t reflect his work, but Dr. Berri would still take Kobe first though, right (not factoring age in, hypothetically)? Same goes for Biedrins. I know he always shoots 60% from the field, which is good for WP48, but do you take him over Brandon Roy (22nd)?
I know WP indicates how many wins you account for for your team, and Camby gets more wins than Kobe. So, if Dr. Berri wants to win, he’d pick Camby over Kobe, correct? And I’m not one of these guys who only looks at scoring; I value rebounders and passers and defensive players and shooters, so I definitely am I fan of tough rebounders and efficient players like Camby and Biedrins. My second pick would probably be a Center like those two, but for my first pick, I’m going to take Kobe (I’d actually take LeBron, but since he’s 3rd, he doesn’t really apply to my question.) Does this make me a bad (fake) GM? Ignoring the data and going with my gut? I know that’s the problem with a lot of GMs and even sportswriters, but I feel like for this one instance, Kobe and DWade should be guys that transcend the data. (And for the record, I’m still someone who’d look at the data and take Camby over someone like Tony Parker, happily.)
brgulker
May 8, 2009
Isn’t one if CP3’s biggest contributions to productivity numbers his turnover to steal ratio? He gets a lot of steals without turning the ball over very much – I would think that weighs heavier in dberri’s metrics than assists. Or am I wrong about that?
Michael
May 8, 2009
Your right. That and his high shooting percentage and rebounding rate for a point guard. I would think turnovers are the only thing holding Lebron back from being the best player according to this metric as well as all the others.
Phil
May 8, 2009
Paul’s been better this year than last year, in every conceivable way. Lebron (and Wade) have too, but still, it’s odd to see him drop so far.
Also, Kevin Martin is average? What happened?
Evan
May 8, 2009
DWade doesn’t need to transcend the date. DWade is very productive.
Ray
May 8, 2009
Camby is still ahead of him (although Wade’s second half may have slipped him ahead.) I’m mainly focused on Bryant. He’s only 14th.
Will
May 8, 2009
Ray,
I have been following the wages of wins for nearly two years, and yes, I would pick Camby over Kobe if I were starting a new team and wanted my team to produce as many wins as possible (although things like age and injury history do matter). I’m glad you’re willing to consider empirical evidence in your decision making process – I wish more of my friends were willing to do the same – but with Kobe is seems as though you have already formed your conclusion and are evaluating the WP48 model through through that belief, rather than the other way around.
Ray
May 8, 2009
Thanks for the comment Will. Here’s the thing. Nobody loves stats more than me. I’ve looked at 90% of every box score for every game this season. So, I’m really into the Wp48. I’m not saying it’s ineffective. I’m all for anything that rewards Camby, Battier, Varejao and others with positive numbers that aren’t usually represented. But I’ve watched plenty enough basketball to see the things Kobe can do (and I’m not a fan of his, nor the Lakers.)
For me, efficiency is terrific; I love it. But the game is about putting the ball in the basket. Kobe’s reliable (often explosive) scoring, to me, is more dangerous than a Camby defensive rebound. Kobe is drawing double-teams and contorting the defense, while Camby’s offense is a secondary role. Again, I love what Camby-typer players do, I just think when you have the offensive compacity of LeBron, Wade, Bryant, and possibly even Melo and Durant, that your presence is greater than the fundamental numbers that Camby and Biendrins rack up.
That being said, I don’t really use WP48 to evaluate stars. I use it more for role players, guys who are trying to do the little things, and which of those players do it best. By watching a lot of Cavs games circa 2007, I saw how terrible Hughes and Pav were, even though the announcers would find some way to talk about them positively. WP48 shows us that Hughes produces 0.0 wins as a starter, and Pav didn’t fare much better. That’s why I like WP48.
brgulker
May 8, 2009
The argument’s really a non-starter, isn’t it?
My reasoning, and please correct me where I’m wrong:
1) GM’s don’t get to ‘pick’ their teams like we do in fantasy basketball. Not only do they not get to start from scratch with a clean slate and pick anyone they want, they have to negotiate contracts, draft picks, etc. etc.
2) Wins produced or win score, as helpful as they are, don’t always indicate who’s going to win championships, no matter how accurate they are in terms of evaluating regular seasons. If you’re a GM, you’ve got at least two responsibilities: 1. have successful regular seasons, so you can get butts in the seats paying ticket prices and buying concessions; and, 2. win championships. My point? Crazy things can happen in the playoffs. Seven-game series minimize aberrations, but they don’t eliminate them. Hence, a team like GS can knock of a team like Dallas, even though Dallas is the more productive team.
In short, “Who would you pick first in a pick-up game” is too theoretical to be relevant, because a situation like that never actually happens in the real world of basketball.
Rob O'Malley
May 8, 2009
So you like to use it to confirm what you already believe and you ignore its use when it goes against your popular belief? You’re basically not using it at all in that case. At that point Wp48 doesn’t factor into you’re evaluation. It’s just for selective confirmation, which you can use anything for.
Michael
May 8, 2009
Honestly at their ages you would be mad to start a team with either of them!
That said I would personally pick Kobe over Camby for the following reasons. First as a non-scoring defensive rebounder I think Camby has been very overrated by this system. He’s basically defenseless (just ask Charley Rosen about that lol,) is extremely injury prone (last year is the only season he has even come close to playing every game,) and on top of that next year he will be 34 years old to Kobes 31! Then, given the fact that Kobe plays more minutes, and more games, there really isn’t much difference between them in actual wins! For example even if Camby maintained his mid season pace of 0.437WP48 (which is doubtful,) given the number of minutes he played by seasons end he would have produced approximately 17.5 wins to Kobe’s 14. Seriously, all things considered it would seem a little foolish to take Camby ahead of Kobe in starting a team. But there are a number of WOW purists who would disagree!
brgulker
May 8, 2009
So you like to use it to confirm what you already believe and you ignore its use when it goes against your popular belief? You’re basically not using it at all in that case. At that point Wp48 doesn’t factor into you’re evaluation. It’s just for selective confirmation, which you can use anything for.
No, at least I don’t think so.
What I am saying is that if I were a GM, I couldn’t just pick any group of 12 players I wanted to pick. Furthermore, I wouldn’t be just thinking about winning a lot of games in the regular season.
So, the question “Who would you pick first” might be fun to entertain, but it’s purely theoretical and has no application to real NBA GM’s.
Ray
May 8, 2009
brgulker, “who’d you pick first?” is basically the only way you can compare the two. Because of contracts and fans and everything involved, I can’t sit here and say would you trade Kobe for Camby. I’m more or less saying, is Marcus Camby better than Kobe Bryant? I use “who would you pick first” as a platform for engaging the debate. I know you can’t really pick for teams in the NBA.
Rob, I know that’s what it seems like I’m doing, but I keep trying to stress that I’m not. I know that Camby and all of those other guys in the top 25 that the casual fan would scoff at, that those are really good players. I agree with WP a great deal, I’m just not sure that it (or any statistical measure) calculates the impact Kobe has on opposing defenses. Kobe draws double teams, which is freeing up other areas on the court. Since there are only a few players in the league who draw legimate double teams, then WP48 is hits the mark routinely. I’m not ignoring WP. Bryant is still a cery high 14th on the list. I just feel that he creates more pressure on the defense that he doesn’t get credited for statsitically. Not to mention a few guys above him on the list, David Lee and Biendris for example, are really bad defenders. But I completely understand that WP doesn’t calculate that.
brgulker
May 8, 2009
@ Michael:
Yeah, Kobe’s younger, but he’s also got a lot more miles on the odometer — came straight out of high school, has played a lot of playoff games, has played in important summer events (Olympics, World Championships), and has played a lot of minutes in all of those.
So yes, he’s younger in terms of years. But I’d take an 01 Honda Civic with 80,000 miles before I’d take an 04 Honda Civic with 130,000 miles.
I realize it’s a flawed analogy, but it’s something to think about.
Ray
May 8, 2009
And the whole point is ignoring ages and injuries. Camby and Bryant are both well into their NBA careers, so you wouldn’t REALLY start your team with them. But based on their production right now is why I raise the question. Who is more valuable, basically? Without the age and injuries of course.
Golden Graham
May 8, 2009
Your analogy would make more sense if the 01 Honda Civic only had 80,000 miles on it because it was in the garage 6 months every year getting repairs.
brgulker
May 8, 2009
Heh, good one.
Michael
May 8, 2009
Ray ignoring age and injuries you may as well debate starting your team with Bob Pettit or Oscar Robertson.
brgulker
May 8, 2009
I’d take Wilt.
Michael
May 8, 2009
lol Wilts been dead for 10 years, I chose the other two because they are actually at least still alive!
Michael
May 8, 2009
Also, my point wasn’t really centred around Kobe being a few years younger. The fact is given the number of minutes and games each of them plays there doesn’t actually seem to be that much difference in the number of wow wins they produced this year. then taking into account Camby’s history of injuries (compared to Kobe’s durability) and the fact that he is a defensive liability and an offensive afterthought, means I would take Kobe if I had to. The point is that stats (no matter how much you like them) aren’t really an excuse to stop thinking about the game of basketball.
Ray
May 8, 2009
Michael, you’re missing the point. WP48 doesn’t care about your age, contract, or what the fans think about you. I’m asking, who is more valuable, right now? Forget about the starting a team stuff, I was just trying to use a platform for discussion. All it is, who is more valuable.
brgulker
May 8, 2009
I just had a thought that may or may not be helpful:
Production is objective. Value is subjective.
Is that helpful or worthless?
Michael
May 8, 2009
Ray did I mention age in the last post? Or the players contracts? Or what the fans think? (unless you want to characterise me as a fan.) First you wanted to know who people would start their team with, now you ask who is more valuable. My answer to the latter is the same as the former. Even if Camby maintained his stellar first half right the way through to the end of the season their total wins produced are not much different. Camby is a great defensive rebounder, Kobe is a great offensive player. Of the two I think it would take a significant disconnect from reality into statistics land to say that Camby was more valuable. Of course you are free to disagree.
brgulker
May 8, 2009
Assuming it were possible:
If I were starting a team, I would take LBJ, hands down, no questions asked. He’s the best player on the planet, and he does everything well. And he’s not yet reached his ceiling.
Ray
May 8, 2009
Michael, I asked who would you take first, I don’t know, like you’d ask your buddies in a bar. People were misinterpreting what I meant by that, so I just simplified it to who is more valuable. The contracts and fans comments were just in case the next poster was going to comment, “well Camby only makes __ and Kobe makes __, so I’d take Camby.” I asked from a pure production standpoint, not any variables that circumvented my question. And if you’re saying Kobe is more valuable, I’m not sure what we’re arguing, because that’s my stance. Camby is great for what he is, but Kobe’s role is more vaulable.
Michael
May 8, 2009
Exactly I agree. Now if we were arguing about something like Kobe or Lebron that would be a different conversation because Lebron is awesome.
TRad
May 8, 2009
I wonder about positional scarcity. My point is maybe by choosing Bryant or Wade first and C/PF second we get a more productive team than by choosing Camby/Biedrins first and some SG second.
Ray
May 8, 2009
Michael, I can give you some LeBron bromance like you’ve never seen. I’d take LeBron first in .001 seconds. He is incredible.
Rob O'Malley
May 8, 2009
@ brgulker, I was actually responding to ray. you managed to post 2 minutes earlier than me so I can see how it was confusing. I actually agree with what you said.
chan man
May 8, 2009
In hypothetical cases of a fake GM for a pick-up game, no I wouldn’t pick Camby over KB; I’d pick DHoward over Camby (for Center), and DWade over KB (for 2Guard). It’s important to consider position, i.e. let’s compare apples to apples. It’s been noted numerous time on this journal (either by the Prof or commenters, or both) that you couldn’t have a team of 5 centers (5 Camby’s) nor a team of 5 point-guards (say 5 AI – is he a 1 or a 2?!).
Yes it’s amazing some of the things KB can do – which I wouldn’t doubt could also be done in similar fashion by CP3, LBJ, DWade, or even Iguodala, BRoy, Starbury (in his prime) and AI (in his prime). That doesn’t equates to winning basketball as a team sport. (I play bball myself and know how bad it is with a team of all scorers and no one’s interested with rebounding.)
Italian Stallion
May 9, 2009
Lebron was given MVP by such a wide margin partly because he was the best player, partly because he was viewed as having an inferior supporting cast to Kobe, but mostly because they gave it to Kobe last year. Now it was Lebron’s turn (pure politics).
If they gave it to Lebron last year, Kobe would have gotten it this year or at a minimum it would have been much closer.
It’s also certainly not just about scoring.
Otherwise, it’s real tough to explain Nash getting TWO of them. Nash got two of them because the perception (and reality for that matter) was that the Suns were WAAAAAY better when he played than when he didn’t. Those years they viewed “valuable” as the literal definition of “valuable” as opposed to “best player”.
The weighting of each of the factors that determines who will be MVP VARIES from year to year.
It’s a kind of popularity contest.
The selectors will use varying justifications to make whoever they want MVP, but obviously he has to have good stats, be considered a great player, and play for a good team.
Soon, they’ll be looking for an excuse to give it to Wade.
Italian Stallion
May 9, 2009
I’m sorry to be so disrepectful, but in this case I just can’t help it.
Anyone that would select Camby over Kobe doesn’t know a damn thing about basketball!
I’m not even a Kobe fan. I happen to agree with the consensus here that’s he overrated. I would also point to lot of the same stats to make my case. But this is the preposterous side of basketball statistics at this stage in its development.
Brilliant people build brilliant models that try to capture the skill, talent, and value of various players, but neither the brilliant people nor the models are perfect.
When you see an idiotic result like this, you should at least be asking yourself “is there something wrong with the model” not starting with the assumption “why is everyone on planet earth with eyes and functioning brain wrong not to understand what my model says”.
I’ve argued that this model has two problems.
1. It tends to have the opposite problem of PER. Where PER tends to overrate inefficient high usage scorers, this model tends to overrate very efficient low usuage scorers and underrate high usage scorers of at least adequate efficiency.
As long as efficiency is adequate, high usage scoring is valuable because those low usuage highly efficient guys CANNOT score enough to win games and maintain their efficiency. The high usuage scorers actually contribute to the efficiency of the low usage guys.
Anyone that thinks I’m wrong about this has obviously never played basketball.
Change Kobe’s contract to reward him for efficiency and I guarantee you he becomes the most efficient scorer in the league despite being a SG. The problem is that his scoring would go way down, guys like Ariza etc.. would have to take more bad shots and get worse, and the team would be worse.
2. Rebounds are extremely valuable for producing wins at the team level, but not at the individual level. Rebounds are one of the most fungible stats. If you remove a top rebounder from a lineup, it’s pretty easy to look down your bench and find someone that will get most of them back. Then, the rest of team gets a portion of the remainder so you only lose a little. Other things like scoring, shot blocking, assists etc… are much less fungible.
For an example, look at David Lee.
Lee is a leading rebounder in the league and considered the only good one on the Knicks. Yet when he’s off the court the Knicks are just as good a rebounding team. To ignore things like that is simply folly. David Lee’s true value is in his moderate usage but very efficient scoring. When he’s out of the game, the Knicks can’t replace that.
Sure if you have Dennis Rodman, that’s going to be tough to replace, but rebounding is VASTLY overrated on a individual level despite being tremendously valuable at the team level.
Michael
May 9, 2009
Stallion with the exception of Will I think everyone has picked Kobe in this little hypothetical scenario.
Also, whilst adding the disclaimer that I basically agree with the points you make, can you tell me where you got the opinion that PER overrates some high usage scorers? Which scorers?
And on your point about rebounding, would you agree that a strong offensive rebounder such as Kevin Love or Emeka Okafor is more valuable than a strong defensive rebounder like Marcus Camby or Troy Murphy?
Evan
May 9, 2009
Change Kobe’s contract to reward him for efficiency and I guarantee you he becomes the most efficient scorer in the league despite being a SG.
Like that time where he took 1 shot in the game because he was criticized for shooting too much?
Michael
May 9, 2009
Evan when you mentioned this in February you said he refused to take any shots. As far as I can tell there are only 14 games in Kobe’s whole career where he took either 1 or 0 shots. Of those there were only 2 games where he played more than 10 minutes. Also with the exception of a 2004 game where he played only 1 minute, all of these games were in 1996 or 97.
Since this is the second time you’ve mentioned it do you want to clarify what game you are referring to? Do you mean the second half of the playoff game against the Suns in 2006?
Ruben Durham
May 10, 2009
“Anyone that thinks I’m wrong about this has obviously never played basketball.”
Stallion, I feel your argument and agree with it very deeply, but unfortunately it’s not a very original critique in these parts – and worse yet, it’s a very easy one for Dr. Berri to sidestep. You gotta come at these guys with something they haven’t heard yet to try to get through to them.
Now I’m not sure when the burden of proof shifted from the late-comer economist with the flawed model to the people who make their living evaluating basketball players, but in WoW world that’s exactly the way it is. So you have to present evidence that contradicts DJB’s model (hopefully to the point of total invalidation).
Either prove that applying what correlates to wins at the team level to individual players is a grossly inappropriate leap in logic, or just give up and let them go on living in fantasyland. It’s a waste of time to try to come at them with that angle, because they simply won’t be deterred by it.
Evan
May 10, 2009
Michael
Sacramento, 2004. Look it up, ldo.
mrparker
May 10, 2009
The same people who would take Kobe before Camby are the same people who would have taken Kobe before Garnett last year and are the same people who continuously lose bets to people like me who call them out on their bs assertions.
Michael
May 10, 2009
wow 2004….thats a very long time to hold a grudge!
Btw he actually took 13 shots in that game.
PJ
May 10, 2009
Food for thought: of the top 20 players by WP48 at the midpoint of the 08-09 season, there are 8 centers, 4 point guards, 3 power forwards, 3 small forwards, and 2 shooting guards.
Now, one could draw many different conclusions from that (admittedly small) set of data, but I thought I’d throw that out there.
For what it’s worth, at that point in the season, Camby and Kobe were both the second-best players at their respective positions.
Note: I had to guess on a few of the positions (Duncan, Gasol, Bosh) as I’m not sure what position Berri assigns them.
Daniel
May 10, 2009
Kobe was not #2 at his position! Ginobili has been a better per-minute player for YEARS, including the first half of this season. Ginibili was at .297 and Kobe was at .237– Ginobili was 25% more productive per-minute for the first half of the season.
There’s a reason the Spurs got bumped in 5 by the Mavs. If you subtract Kobe from the Lakers or Wade from the Heat, their teams would have difficulty in the playoffs too… though the rest of the Lakers might be good enough that they could win the West this year with an average replacement anyways…
Joe
May 10, 2009
PJ,
My initial guess would be that this is a result of the “short supply of tall people.” If you (re)read the book, specifically the MJ chapter, you will get a much more thorough explanation than I could ever offer. I don’t have the book here or I’d quote that part probably.
The basic premise: When you look at the underlying population, larger sample sizes(6 foot people) tend to yield smaller deviations from the mean. When you look at smaller sample sizes(7 foot people), deviations tend to be larger.
So basically taller positions are less competitive due to the underlying populations.
Because of this, the elite are likely to be tall-heavy and it “makes sense” to me.
Also, “tall players” probably play less minutes(just speculating) which would further this idea since more of this small sample has to be used and the players getting these minutes would tend to be poor players one would think.
Michael
May 10, 2009
Its actually more likely that centers dominate the top 20 in this metric because they are more likely to take high percentage shots, get rebounds, and not turn the ball over. In other words, they are more efficient than other positions which have to put the ball on the floor and create shots from a greater distance. They are also the players who would get most of the credit for defense in this system since when a shot is missed they are credited with the defensive rebound rather than the forward or guard who forced the difficult shot.
Joe
May 10, 2009
Michael,
There is a positional adjustment made to offset this. Please read the archives if you are going to critique the model at the very very least.
How to calculate wins produced:
http://www.wagesofwins.com/CalculatingWinsProduced.html
simulator
May 10, 2009
At the start of the season, not a single person in the media predicted that Cleveland would win anymore than 50~ games this season.
Berri’s own prediction was that Cleveland would win 45~46 games (and he used 3 year average of players).
Given that the team won way more than what his model has predicted, I’m surprised that Berri still insists on the ‘every player production is independent of the other’ model, which is obviously flawed.
Michael
May 10, 2009
Joe, I realise that there is a positional adjustment involved in calculating wp. There are also adjustments for team mates blocked shots, team mates assists, and team defense. So many adjustments, and yet the basic point that (as PJ pointed out) center’s dominate the rankings still stands. Funnily enough that fact is even mentioned on the page you so dismissively linked me to, so I am not quite sure why you consider mentioning that to be a ‘critique’ of the model. If you are actually referring to my brief point about the way credit is awarded on defense, then you should understand that that is not a critique of the statistical methodology used to create wp, as much as of the underlying assumptions it is based on. I also find it odd that you would implore me to ‘read the archives..at the very very least’ as if you have a vested interest in quelling dissent towards this model, and as if any hint of such would obviously merely come from a failure to research the process involved in its calculation. Very odd.
Jason E.
May 10, 2009
Evan, I tried to “look it up” but I can find no game against Sac in 2004 where he only took 1 FGA, or against any other team. If you are so certain of this, rather than calling people “idio.”s and demanding that they confirm your claims, perhaps you can simply provide a reference, perhaps a date, anything more robust than your claim that it happened.
Michael
May 10, 2009
Hi Jason, Evan was referring to a game on the 11 of april 2004 against the Kings where Kobe only took 1 shot in the first half. He finished with 8 points on 23% shooting, it wasn’t a particularly good game for any of the Lakers as nobody finished with more than 15 points and they lost badly.
I think ldo was net talk for ‘like duh obviously, but I could be wrong maybe he was calling me an idiot!
http://www.basketball-reference.com/boxscores/200404110SAC.html
Ray
May 10, 2009
Michael, I couldn’t quite pick up what you meant 3 comments ago (my fault, not yours) when you said how Centers get a higher WP because they rebound, take high percentage shots, and don’t turn the ball over. Were you criticizing the model for that, or just pointing out the favor towards centers? And regarding your question about rebounding, I favor offensive rebounding over defensive. Creating an extra possession for your team is so important. Defensive rebounding is almost nonchalant, while offensive rebounding requires more effort and hustle.
Joe
May 10, 2009
Michael,
You do realize that SGs, PGs, SFs, PFs, and Cs all average .100 as a WP48, so your argument is invalid. Centers don’t dominate, rather the elite centers dominate more than the elite SGs.
Why is that? As I state above, read the book.
Michael
May 10, 2009
Hi Ray, I just think that if you have a metric which focuses on efficiency then centers will naturally congregate at the top of the rankings because they are the most efficient players. I don’t think you could criticise the model for that, dominant center’s are definitely important to team success (Jordan’s Bulls’ are basically the only teams to win titles without one,) and given the rush of league front offices to pay multi-million dollar contracts to basically anybody 7ft or above regardless of skill I think it is obvious that they play an important position. That said I think you could point to the treatment of rebounding (a defensive rebound [0.033] is worth the same as an offensive rebound [o.032] and both are equal to a made two point fg [0.032]) would lead to the more prolific rebounders scoring higher than the more prolific scorers, especially when you factor in the increased propensity for latter to turn the ball over [-0.032] against the formers increased likelihood to block shots [0.019.]
Joe, I don’t need to refer to the book to know the reason they all average 0.100. You said it yourself, it is because of the adjustments ;-)
Man of Steel
May 10, 2009
Italian Stallion has a good point (but a little dogmatic) about usage. Footballoutsiders.com makes an adjustment for this, differentiating between players who are very efficient at low usage rates and those who perform at average or slightly above average efficiency at high usage rates. Basically, the premise is this: being average, or slightly above average over many minutes (or carries/passes for their puposes) is valuable, and is worth more than simply “average.” I believe they do this by comparing high-usage players to league-wide average replacement level. Here is a concrete example: Kobe (or whoever) is a little above average with respect to efficiency in a lot of minutes. When Kobe is not on the floor, though, the Lakers don’t get to replace him with an average player. In all likelihood, the team will have to replace a player such as Kobe Bryant or Kevin Durant with a player who performs at a level far below average. Thus, “average” is not zero, nor is average the base from which one evaluates performance. A high usage player is not worth “average” to his team, he is worth (“average” – replacement level) X 40 min. per game. Thus, having players who can shoot efficiently at a high volume is important (e.g., it is important to have someone like DWade at SG instead of Allan Houston), and having high-efficiency/low-usage players is important as well (e.g., having Joel Pryzbilla instead of Eddy Curry at C). Both are important.
Michael
May 10, 2009
I think Player X versus Average is a stronger argument than Player Y versus replacement because although the Player Y argument works for the team internally, it doesn’t take into account external factors such as the strength of the opposition player which is important. For example if you had an average shooting guard, you could argue that his value is average minus his replacement (player y argument) but in reality unless he is better than his average opposition, the team will not be helped (and may even be hurt) at his position. I think usage arguments are important because (and I think Stallions arguments are similar) a high usage player if efficient enough to draw defensive focus from the other team can significantly increase the number of open looks etc for his team mates. His value therefore is his own efficiency combined with the increase in efficiency he brings to his team mates. I know the argument against this is that for example a player like Kobe doesn’t improve his role playing team mates because they produced at the same rate on other teams, but that fails to take into account that most role players are role players wherever they go. For example a guy like Trevor Ariza can go from Orlando to LA and maintain production, but that doesn’t mean the fact that Kobe draws the defense hasn’t helped him, as in Orlando he had Dwight Howard or Rashard Lewis fulfilling the same role. I do think both arguments have pluses and therefore it is important to look at metrics which measure each side however overall in judging the individual player the average argument (at least to me) is slightly superior.
Ruben Durham
May 10, 2009
You’re talking about a VORP like stat, which already exists with basketball reference’s Winshares and BBPro’s WARP. But neither of those take into account the extra on-court benefit of usage, which is creating high-% chances for teammates by drawing in defenders. Obviously a regression at the team level will not pick up on this because somebody on the team has to shoot and always does – either that or it commits a turnover. So it treats efficiency as the only thing that matters. But with a 24 second clock, you cant just play hot potato waiting for the defense to come out to you, you have initiate action and everything else is created from that – especially those high-% shots by the Cambys and Biedrinses of the world. But because teams do not increase/decrease usage (its always 100%) and his regressions are on teams, Dr. Berri assumes players can also increase their usage % indefinitely with no loss in efficiency. Now we just need to prove that this is not true. The conventional wisdom is clear, and DJB has not offered any real proof that the conventional wisdom is wrong, but the statgeeks on the other side have not offered any proof that it is correct either.
Ruben Durham
May 10, 2009
Sorry, that was is response to Man of Steel, but I agree with Michael also.
John
May 10, 2009
I agree with what (I think) Ruben is saying, the burden of proof falls to both sides. There are plenty of instances of conventional wisdom, even conventional wisdom that makes perfect logical sense, being dead wrong. Same goes for models that contradict conventional wisdom.
It would be nice if more people on both sides of the debate, but particularly the people arguing against Wins Produced, could provide something more than anecdotal arguments. “It’s obvious” or “you’ve never played basketball” aren’t going to convince anyone who isn’t already a believer. Those who have made evidence-based arguments, like the diminishing returns to rebound study, have at the very least done a better job of getting through to me, an agnostic in the debate.
mrparker
May 10, 2009
I didn’t know playing basketball was prerequisite for team building ability. No wonder Michael Jordan has been such a great basketball executive.
Michael
May 10, 2009
John I think the problem is that people who like to see ‘evidence-based’ arguments generally just mean ‘statistically based’ arguments. When people try to provide these arguments however it is usually dismissed as either a failure to understand, a biased attack based on a vested financial interest in the status quo, or irrelevant in any case because it hasn’t been peer reviewed. (At least that is how I have perceived it.)
Also I think Ruben was saying that the burden of proof lies with the people who created this system (i.e “exceptional claims require exceptional evidence.”) I think that possibly the prima facie case for wp was accepted by the majority of early long term posters on this site, which in their eyes shifted the professor’s position from burden of proof to benefit of assumption, a position which leads later visitors to the site to an encounter a situation which as mentioned by Ruben, seems unusual/hostile given the nature of some of the more extraordinary claims made here.
ilikeflowers
May 10, 2009
The only evidence required for someone’s model/opinions to be taken more seriously than the standard internet dude comment is the ability of said model/opinions to predict wins from per-minute player ratings close-to or better than w48. There is no burden-of-proof, it’s a simple comparison.
Ruben Durham
May 10, 2009
Not really, ilikeflowers. Coaches tend to use the same players in set roles no matter what team theyre on, so theres no way of knowing how a team of all low-usage players would do by predicting the future. A team like that would never happen in real life. Maybe because coaches and GMs know a team like that would never survive? (no matter what their ‘wins produced’ says going into the season) Just a thought.
Evan
May 11, 2009
To those of you were upset by what I said —
ldo does in fact equal “lol duh obviously.” You should really spend more time on the internets. Or not. More seriously though, if someone would be offended by “ldo” then I am truly sorry. You have more issues than I can help you with.
Second, geez, I thought I was supposed to be the stathead who doesn’t watch enough basketball. Kobe got criticized for shooting too much, and as a result he took 1 shot in the 1st half (or whatever, if i mix up the results a little, it has been 5 years). It was a ridiculous display of petulance as he refused wide open shots in favor of passing. I also seem to recall him giving an interview at half where he openly implied that he was going to not shoot and see where that led the team. I’m pretty sure this was the Sac game.
Vorn29
May 11, 2009
I think Tony Parker should be a candidate for MVP. Without him this years Spurs were a 10win team
Mo
May 13, 2009
Seems like sooooo many people don’t realize what CP3 is doing or how good the little guy is.
Man of Steel
May 16, 2009
Thanks Michael and Ruben for your feedback. It appears that I still have a bit of a reading curve to work my way through on other sites than this one (such as those Ruben mentions). I do have something like VORP in mind, but I think my language was not precise enough and Michael misunderstood what I meant to say (but failed to say clearly). When I used the terms “replacement” and “replacement value” in my earlier post, I did not intend to refer to the actual player’s replacement – the real-life player who comes off the Laker’s bench to replace Kobe Bryant in my example. Instead, I was intending to point out that the value of a player depends on his value relative to the average replacement player – the average bench SG in the NBA. So if Kobe (or perhaps someone not quite as good would be a better example) is slightly above average per minute, compared to a replacement-level player far below average, over a large number of minutes, then the value of player A (Kobe or whoever) is not (his efficiency level per minute – average efficiency level per minute). Rather, player A’s value is more accurately reflected by an equation such as that used to calculate VORP (player A’s efficiency level per minute – replacement level efficiency per minute) X minutes.
Contrary to conventional wisdom, this sort of idea would suggest that the value of efficient high-usage players is not that they “make their teammates better by drawing away the defense,” but instead their value is found in their ability to produce at an average or above-average level of efficiency at a high level of usage, contrary to most role players and replacement level players.
Also, Ruben’s point about teams taking the same amount of shots is probably partially accurate, but I think that the real point of his observations is that one cannot compare high usage players and low usage players like apples and oranges. I recall Dr. Berri pointing out somewhere that there is no evidence that players decline in efficiency as they take more shots or increase efficiency as they take fewer shots. Unfortunately, we don’t have the data sample to be able to confirm this. Players don’t go from extreme low usage patterns (like Joel Pryzbilla) to extreme high usage (like Dwayne Wade). Players may fill slightly different roles on different teams, we don’t have a data set to analyze for players who went from taking 3 shots a game to taking 20 shots a game. Dr. Berri’s point thus needs qualification. We don’t have evidence of significant change in efficiency at changing usage rates within a player’s assigned role. Even if Joel Pryzbilla shoots 6 times a game in 40 minutes per game next year instead of 3 shots a game in 20 minutes (or whatever) he has not actually changed his base level usage pattern. Similarly Stephon Marbury has gone from taking 20+ shots in 40 minutes per game for a long time to taking 5 shots in 10 for the Celtics. His actual usage is the same (1 shot every 2 minutes). Although both players have changed their raw usage rates, they still play the same roles, just recast on different teams.