Editors Note: Today’s guest column is authored by Jason Eshleman. For regular readers of the WoW Journal, this is the “Jason” who offers so many wonderful comments. In addition to posting frequently at the WoW Journal, Jason is a research associate in Anthropology at UC Davis. Jason co-founded Trace Genetics, which is described on their website as “specialists in genetic identity DNA analysis with expertise in DNA ancestry, forensics, ancient DNA analysis, molecular diagnostics, and population genetics.” When Jason is not doing stuff that is far more complicated that measuring the value of an NBA player, he’s living life as a long suffering fan of the Golden State Warriors.
On January 3, 1991 one of Bill Cartwright’s famously sharp elbows connected with Akeem Olajuwon’s face. [Olajuwon would add the H to his name shortly after returning from the injury.] The Dream’s right orbit was fractured. While the Rockets would hang on to defeat the Bulls that night, Olajuwon would miss most of the next two months. This was not considered good news in Houston.
For good reason, Olajuwon was considered one of the game’s best centers. Olajuwon led Houston in scoring, rebounds, blocks and steals. He was more than an effective player. He was their star, their go-to guy, their do-it-all dominating force. And as he went down, so too did their hopes for a successful season.
When Olajuwon found himself on the injured reserve, the Rockets had won 17 games against 13 losses. They were on pace to finish with 46 win. While better than the 41 the team had won the year before, it was a far cry from the NBA elite. And among the elite is where Olajuwon expected to finish. Yet despite this ambition, Houston had, for most of Olajuwon’s career, been mired as a low to mid-40s win team. Clearly the do-it-all All-Star center was not the problem. The finger pointing was usually directed at his supporting cast. Despite the stellar performance from inside the post, Houston was only moderately better than middle of the NBA pack. Without him, their chances for success seemed slim-to-none.
Or so it appeared. After Akeem went down, the Rockets proceeded to win 15 of their next 25 games, a record that, extended over an entire season, equates to a 49 win season. Their best player stopped playing. The team got better. The rest of the team stepped up.
Or did they?
The short answer is both yes and no and depends greatly on the definition of “step up.”
As a whole, the Rockets team statistics did not vary much once Olajuwon went down. For the 25 games missed, the team shot 46.8% from the field, viritually indistinguishable from the 46.7% that would pace them over the entire 82. Total field goal attempts per game remained almost entirely unchanged, though with Akeem’s 17 shots per game erased, the shooting load had to be redistributed. More than a few have argued that when a player has to shoot more, his field goal percentage will suffer, that the additional shots will likely be of a low quality marginal variety that require the charms of fate to fall. Yet while all four remaining starters did see their FGA go up once Olajuwon went down, only forward Buck Johnson saw his FG% decline, ever so marginally from 47.7% on the season to 46.7% over the 25 game stretch. Vernon Maxwell actually appeared to gain accuracy with increased attempts, shooting 4% better while taking 8% more shots, a disproportional amount of these from beyond the arc. The remaining players shot more, played more and, for the most part, played like they had when Olajuwon was in the lineup, playing somewhat more minutes and shooting somewhat more often, but with more or less the same rate of return. Maxwell stepped up, the rest of the team continued to do their jobs and thus finding someone else to take (and make) Akeem’s shots seemed to be easy.
It may not be intuitive that a 20+ ppg scorer who shot more than 50% from the field could be so easily replaced, but on closer inspection, it is not nearly as surprising. While Olajuwon hit 50.8% of his shots from the floor (he did not make a 3-point attempt, thus his FG% and effective FG% were equal), the rest of the team was not too far off this pace. The rest of the team had an effective FG% of just under 49%. In other words, the rate of return on his teammates’ shots, while lower than Olajuwon’s, was not that much lower. As such, divvying up the shots, plus an increased efficiency from only one player -Maxwell’s eFG% was 45.4% in games with the Dream, 51.4% without- could offset Olajuwon’s scoring load and the benefit the Rockets received from this scoring. [Maxwell would seem an unlikely candidate to step up, as one might expect that an inside presence like Olajuwon would help Mad Max get free on the perimeter when the defense collapsed on the Dream. But instead, without the inside presence, Maxwell shot more, shot from outside and shot better. Go figure.]
But of course the game is more than just scoring points. What of the other end of the court? What of the rest of the game? How does one lose 14 rebounds a game and not suffer something? While it has been argued that shooting more could cause some players to shoot less efficiently, so too some have argued that a significant portion of many a player’s rebounds are ‘taken’ from teammates. This suggests that if you remove the superior rebounder from the game, his teammates will start to grab some of those freed rebounds, negating some not-insubstantial part of the loss.
Perhaps such returns have been witnessed elsewhere, but no such fortune befell Olajuwon’s regular front-court mates. The 14 rebounds Olajuwon regularly pulled down did not magically fall into the waiting hands of his teammates once the Dream was no longer there. Regular frontcourt mate Otis Thorpe saw only a very modest increase in his rebound rate; SF Johnson and both guards all saw a modest decline, though in all cases variation could well be explained by normal variation within a smaller sample. Here too it looked like Olajuwon had little effect on the statistical efficiency of his teammates. The rebounds did not fall their way once Olajuwon went down. It was not possible for the remaining four starters to redistribute the load. His presence had not helped his teammates totals, nor did it appear that it had hurt them when it came to their rebound rates.
So it seemed that the story of the Rockets success was not most of the shooters finding their target, nor the starters reaching beyond themselves and playing above their heads. It was not by and large the rest of the team “stepping up.” We should then be surprised that the Rockets, minus an extremely effective player-and make no mistake, Olajuwon was one of the best in the game by just about any measure-with little evidence of the remaining starters playing substantially better, should suffer defeat after defeat. But they did not. The Rockets got better.
And indeed this result would have been a surprising result were it not for Olajuwon’s replacement. The Rockets did have other centers on their roster. Nominally, Dave Feitl, a 7-foot tall journeyman out UTEP started the first two Dreamless contests, logging 16 and 18 minutes respectively. Feitl could not, and did not, replace Olajuwon’s contributions to the cause. The Dave Feitl era would end quickly though, as game three in Denver would see a new starter in the middle, a 6’8″ rebounding machine in the person of Larry “Mr. Mean” Smith. Smith, it seems, did step up, and in a very big way.
Larry Smith did not shoot as well as Olajuwon, did not score as much as Olajuwon, not did he have the propensity to get steals or block shots. But he could rebound. Mr. Mean had always been an effective rebounder. Over his 13 year career, he would average just under 10 boards a game while logging just a shade under 26 minutes a game. As the PF off the bench, Mr. Mean was collecting a rebound every 2.9 minutes in 18 frantic minutes under the hoop, almost as often Olajuwon’s rebound every 2.7 minutes. “Almost” as often…
“Almost”, though, would not have been enough when you are replacing one of the game’s greats. But as the starting center, Smith did step up, surpassing, pulling down a rebound ever 2.4 minutes. And consequently, 35 minutes of Mr. Mean meant that the Rockets could gather rebounds as well without Olajuwon as they did with him. And as a consequence, the team held together in Olajuwon’s absence.
On February 28, 1991, Olajuwon again suited up again, for a loss to the visiting Clippers, a team that would finish the season with less than 20 wins. This would be an anomaly however, and the Rockets finished the season with 52 wins, better than they had performed in several seasons. Overall, they were a better team with Olajuwon-he really was that good-and it may well be that the Larry Smith experiment would not have continued to be as successful over the long haul. It is hard to replace a player who gives you 20 points on more than 50% shooting to go with nearly 14 rebounds and 4 blocks a game. However, if you need to do it, having someone like Larry Smith step up isn’t a bad option.
– Jason Eshleman
Evan
December 12, 2007
Great column Jason.
dustin
December 12, 2007
Good research. It’s always nice to address some of the criticisms with examples of what has happened in the real world.
Panda Bear
December 12, 2007
Great column. I hope Jason becomes a regular columnist here!
Kent
December 12, 2007
I always thought the “Jason” posting in the comments was “Jason Chandler.” Is he one of the people that uses zoological nomenclature?
Kent
December 12, 2007
Jason, what do you make of this? — http://thelede.blogs.nytimes.com/2007/12/10/watsons-black-dna-ultimate-irony/?hp
Kent
December 12, 2007
(That last comment was directed at Jason Jason Eshleman not Jason Chandler.)
Paulo
December 12, 2007
No tables? I feel so dirty inside when I read here and I don’t see at least one table. But other than that little thing, great post. It reminds me of how Dikembe Mutombo was able to approximate Yao Ming’s production last year. I hope Jason can do this more often. :)
Harold Almonte
December 12, 2007
Oh my god! What a bad try! Do you know how many Reb/game grabbed Houston that year? 46.1. Do you knoe how many in 22 games without H.O.? 46.0. They lost 0.1 R/g.
Like FGA are re-distributed, and efficiency doesn’t change too much with that. His rebounds are not substituted by another one similar rebounder, are re-distributed among everybody, and everybody’s individual Rebr% don’t change too much. Diminishing return?
Try again.
Harold Almonte
December 12, 2007
The only scoring efficiency that changes with different FGAs and not much, are those scorers who are dependant of somebody does half the shot creation job for them (i.e. Iguodala and Korver this year)
Jack Mott
December 12, 2007
Harold, I cannot parse what you are trying to say. Please be more clear.
Dave, great article, as a Houston fan this was very interesting to me
Ron J.
December 12, 2007
Harold Almonte, what are you saying in your first post? It sounds as if you are arguing with someone who agrees with you.
Harold Almonte
December 12, 2007
I read again, and we agree. This example shows you can make up teammates stats with effort, but not the grade of difficulty/easyness they perform (depending on if it is shooting or rebounding). That is, the value of replacemnet of a player is not only his direct replacement differential, you need to account some marginals other teammates replace. But, this carry us to nothing new about weighting stats and its winning factor, just a demonstration about player’s individual skills.
Harold Almonte
December 12, 2007
I think a higher usager player doesn’t make teammates a lot better or worse just because his presence on the floor, if the most of his help is no a direct help, like a pass, a deflection, a double team, etc.
Jason
December 12, 2007
(I’ll avoid turning this into a genetics blog, beyond saying that Watson seems to be pretty good at sticking his foot in his mouth and classifying someone as X percent anything is a far, far, far more complex issue than can be reported in a single statistic. In reality, we’re all 100% African if you go back far enough. But I’ll leave it at that. Anything else is far, far too much like real work.)
Animal
December 12, 2007
MC5 posted this earlier and I would be really interested in Dberri or Jason’s though on this.
Click to access nessis.pdf
I also wish one of them would participate in the APBRmetrics board so they could discuss basketball statistics with people who know as much as them. A person named “Flint” on the APBRmetrics board is the WP advocate, and while he does contribute a lot to the board, I think that it would benefit much more from either Dberri or Jason
dustin
December 12, 2007
I’m not sure, but I’m pretty sure berri addressed the majority of that paper when it was in powerpoint form. I recall one of his defenses was with the team adjustment for other metrics including residuals. I am not statstically inclined and can’t really say much about what that means.
Animal
December 12, 2007
I believe this was the first time this much of the paper has been shown. I am pretty sure there was a much less-complex paper shown, but none that had the 28 pages that this paper had. If I recall correctly, the last paper that was seen was a 12-page and not nearly as detailed as this one.
dberri
December 12, 2007
Animal,
It is essentially the same as the powerpoints. And it has the same problems. Go back and read the thread on the Shane Battier post. I have nothing to add to what I said then.
MC5
December 12, 2007
Just to be clear, I did not post that link simply to light an unnecessary fire. I am very interested to hear some rebuttals, and I figure this might be one of the only places where a substantial rebuttal might be posted, since the APBRmetrics board seems almost unanimously in agreement with their opinions of Wins Produced (which is a view that I share, but I’m always willing to re-evaluate my position).
Animal
December 12, 2007
I am with MC5 on this, regarding my wish to just here well though-out rebuttals to this because I know that will not occur on APBRmetrics board. Just to clarify, Dberri, your main problem with the presentation by Rosenbaum and Lewin is that the defensive-adjustments are not the same and they do not let their defensive adjustments for other metrics be known?
dberri
December 12, 2007
Let me repeat what I said on the Battier post.
You cannot use residuals to evaluate a model. Rosenbaum claims that the residual from a team efficiency model is the same as the “team adjustment” used in Wins Produced. It is not. The team adjustment is not a residual. And you cannot, and I repeat you cannot, use a residual to evaluate a model. Team efficeincy regressed on points scored does not result in a model that explains very much of team efficiency. To say that with the residual the model can explain efficiency is meaningless. Any collection of independent variables plus the residual would explain a dependent variable. So Rosenbaum’s basic approach is incorrect.
And one more point (which I also made before), Rosenbaum shrugs off the inability of his plus-minus model to forecast. But since that is how he is evaluating all models (even if his evaluation is very, very flawed) it is odd that his model does not come close to passing his one standard.
dustin
December 12, 2007
obligatory link to the shane battier post. Last half of it addresses the paper/slides.
https://dberri.wordpress.com/2007/11/01/what-the-box-score-data-says-about-shane-battier/
Jason
December 12, 2007
I’m not terribly impressed with the paper at this point. It reads like a combination of a draft littered with ideas, some good, some not, obscured by a not well veiled attack at a particular method and a particular person.
I’ve said before that models perform within the parameters of observations that went into the creation of the model. Deviating from this tends to result in ridiculousness and such, critiques like the hypothetical “have the best shooter take every shot” isn’t much of a criticism. That strategy deviates from the realm of observations considered in formulating the model and as such, I don’t think critiquing the model because critics claim it suggests a strategy so far outside of the realm of the observed is valid. It shows little appreciation for what models are and how they are to be utilized in interpreting data and suggesting strategy from this. It’s tough for me to take much of their analysis of WP seriously since their appendix indicates to me that they do not actually understand the model nor why one uses a model.
I’m still trying to figure out their logic in evaluating measures by how they compare to adjusted +/- when adjusted +/- seems to have a poor predictive value. I am certain I’m missing something, but it may be that it’s just not spelled out. I do not believe peer review to be a solution to all problems, but I’d rather this article were reviewed as I suspect that whatever valid bits would be clarified and the ridiculousness stripped out in a review process.
If someone cares to explain this better than Lewin and Rosenbaum did, I’d really like to know what it is that they’re thinking other than that they don’t like Dave.
Animal
December 12, 2007
Well I don’t know anything about residuals and whether or not they can be used to explain future wins, so I will just trust you because you know what you are talking about.
My question is though, considering AWS has the same team adjustment as WP, does it mean anything that it outperformed WP in predicting future wins? Or is the way that test was run even with the same team-adjustment still invalid?
(Just remember I really don’t know much about running tests in stat/econ so sorry if these are bad questions or if your answer above answered this and I didn’t realize. I am just attempting to get a better understanding of what is going on here.
Animal
December 12, 2007
Also, I don’t think this was an attack on Dave personally. While I have no idea what was said personally between Dave and Dan, Dan does say that he think Dave has contributed a great deal to basketball statistics and has caused great discussion and deeper thinking by anyone who studies basketball statistics. Maybe he has a personal vendetta against WP, but at least publicly he doesn’t have a problem with Dave (anymore at least).
dberri
December 12, 2007
Animal,
From what I can tell, he used the residual approach in testing all his models. So all of his tests are incorrect.
Jason (as usual) is correct. This paper needs to undergo peer review. Once it does that, it will either be rejected (and then forgotten) or accepted. If it is accepted, I will write a detailed response.
dberri
December 12, 2007
Animal,
You need to read all that Dan has said about my work and the work of my co-authors. Dan has often made this personal (which it should not be).
And I would add, the last we heard from Dan in this forum, he was promoting his own work with anonymous comments (again, look at the Battier post). This is hardly the behavior we expect from scholars.
The Franchise
December 12, 2007
So, with the right residuals, height and weight of the players would be an accurate measuring tool, then?
Animal
December 12, 2007
Yea, I know in the past Dan has been very critical of you. He has sometimes made this way more personal than need be, but I am referring to more recent comments in the past month or two that he has made about you on the APBRmetrics board. While of course I have heard him criticize WP, as expected, I have not heard the same harsh words that in the past about you.
Here is one quote today: “I also want to point out that Berri should get some credit for the way in which his work has motivated deeper thinking about how to evaluate players. I sincerely mean that.”
This is one of a few that have been posted I would guess in the last month or two. I think he is attempting to make peace personally, although maybe it is too late at this point, I don’t know. But publicly, he is at least trying.
Kent
December 12, 2007
The feud between Dan and Dave goes way back– http://youtube.com/watch?v=yKJkfE1M9wA
dberri
December 12, 2007
Franchise,
You could use popcorn sales and it would work just as well.
Animal,
My co-authors and I do research for a living. We are quite accustomed to being criticized. Dan has gone well beyond what is normal criticizing in the past.
But that is not very important. This paper is simply not very good. It has obvious flaws, which I have pointed out. Unfortunately, if you don’t understand econometrics, you might have a tough time seeing these issues.
Jason
December 12, 2007
Kent, quite amusing.
(But still no love for Larry Smith)
Jimm
December 12, 2007
Most surprising to me was Houston’s ability to adjust defensively without the big man back there. Makes you wonder whether people overvaluate the blocked shot (which I have actually weighed less in my own EFF metric) in favor of good team defense and rotations.
Also, I’d be interested if Houston’s schedule was any relatively easier or difficult during that period, as well as proportion of home/road games.
Kent
December 12, 2007
I want to second Panda Bear’s hope that Jason becomes a regular guest columnist for this site.
Jimm
December 12, 2007
From my own observations, Dan makes a lot of claims about his metric that really aren’t supported by the evidence, such as that he accurately accounts for the strengths of respective units (for instance, comparing the second units of the Lakers and Cavaliers respectively, and then noting the effect on Kobe and Bron’s +/- accordingly).
His claim that he believes his metric would have the most predictive power of any competing metric is also curious since he later explains there isn’t enough data to prove such a claim either way (and he has also never predicted anything to my knowledge).
Where does that confidence come from without at least some kind of positive feedback and verification?
Jimm
December 12, 2007
I ask these questions, because I raise doubts about claims that adjusted +/- accurately deals with varying strengths of depth and benches on other forums, and am always told by fans of adjusted +/- that these concerns have been addressed adequately, even though I’ve read some of the more general descriptive papers and all I see are claims that this is adjusted for (and, even if adjusted for, HOW is it adjusted for, and does this open up a similar can of worms as the PER/EFF/WIN debate where many are passionate and almost religious about what should be the proper weighing procedure)?
Still seems to me that some subjective choices are made to this adjustment and these are not publicly debated or even known by casual folks who state confidently that the adjustment is well and accurate.
Jimm
December 12, 2007
Then again, following DBerri’s claim of the close correlation of EFF and PER, maybe it doesn’t matter how they adjust for production in Adjusted +/-, in terms of what players you’re playing with and against, as well as what players you’re not playing with and who they’re playing against (and who they’re not playing with and against and so on)…:)
Animal
December 12, 2007
Berri, thanks for letting me know the issues. It is just disappointing that the relationship between you and Dan is so horrible that none of this can be discussed between the two of you, hindering the improvement of APBRmetrics.
I am not saying it is your fault that this is the way it is, I am just commenting. Don’t take these remarks to be offensive to you. I understand that you feel like Dan has disrespected you and thus you don’t want to speak with him.
Kent
December 12, 2007
Animal,
I think what Rosenbaum is trying to do is to create as a predictive benchmark the basketball analog of what Marcel does for baseball.
“lots of people are producing forecasts for the 2007 season, and one of the first things every decent projection system will do is regress a player’s performance to the mean. In fact, there is one system that does nothing other than regress each player’s performance to the major league average as a basis for its 2007 projection. It’s called Marcel, because it’s so simple that even a monkey can do it. (Marcel, from Friends. Get it?)”
http://www.hardballtimes.com/main/article/but-i-regress/
From the Marcel baseline we can evaluate models.
Kent
December 12, 2007
If I was Rosenbaum I would have done this as my Marcel baseline …
Take a team’s average points scored minus points allowed for a season and divide by all players based on how many points they scored per 48 minutes relative to the team’s points scored divided by 5. Each player then gets valued as such.
(1) If no players switch teams over the next year your projection for team performance next year will just be the actual team performance last year. Of course your arbitrary scheme for allocating team performance doesn’t get exposed unless there are roster switches.
(2) As soon as there are roster switches we can test how good our model is at predicting. If 2 players switch teams how are the team’s respective performances the next year. Does win score or Marcel or adjusted +/- do best?
dberri
December 12, 2007
Animal,
I am not taking offense. But your take on this is incorrect. A few weeks ago I did discuss this with Rosenbaum’s alter ego. At that time I told him that his residual approach is incorrect.
Unfortunately for those looking for a longer discussion, there isn’t more to the story. You simply cannot do what he is doing.
Kent
December 12, 2007
Animal,
Some of the flaws in Rosenbaum’s approach are revealed in comments on this thread– https://dberri.wordpress.com/2007/11/01/what-the-box-score-data-says-about-shane-battier/
Panda Bear
December 12, 2007
Moneyball for the NBA– http://www.blogmaverick.com/2004/04/01/moneyball-for-the-nba/ … This is old but someone sent to me today.
Panda Bear
December 12, 2007
Another cool post by Mark Cuban– http://www.blogmaverick.com/2005/01/04/stats-that-make-you-go-hmmmm/ It’s from a few years ago.
Owen
December 12, 2007
Kent – That was pure comic genius, literally rotflmao….
dberri
December 12, 2007
Kent,
Forgot to note your youtube video. Yes, I agree with Owen. Comic genius. Kind of puts everything in perspective. As I recall, neither Dave or Dan gets the gold.
Westy
December 12, 2007
I guess I find it a little surprising that the paper can be so quickly dismissed when folks like Dean Oliver strongly imply that something worth considering is there. He notes, “Read it carefully. I can’t say that I “got it” in Boston and need to read the paper to really understand what is going on. I’d think that everyone here would have to as well.” Obviously it’s not in final form, but it certainly seems like a start.
I don’t know, but I have to think that if a CalTech grad such as he who’s been doing this for a career as long as he has thinks it’s worth considering, maybe we should.
Could we get Justin Wolfers (obviously has an interest in basketball) or Steven Levitt (has discussed Moneyball extensively before) to mediate and sort this out in a day or two??
Okapi
December 12, 2007
Westy,
In the spirit of Robin Hanson I think prediction/betting markets should settle the dispute.
Animal
December 12, 2007
As an unbias person who has yet to decide who he believes in this argument, it is frustrating to see both sides defend themselves on their seperate “home turf” but have no discussion. Rosenbaum has offered a response on APBRmetrics to crtiques of his paper here.
http://sonicscentral.com/apbrmetrics/viewtopic.php?t=1589&sid=da528d87e6f04f57d784645d8870eb2f
So I have two great statistical minds telling me two completely different things, and there is no discussion being made so I have no idea what to believe. I can’t form my opinion, I don’t know anything about econometrics. The last thing I thought this argument would boil down to is proper econometrics. I wish there was a third party that would come in and help, because nothing is getting solved here
dberri
December 12, 2007
Animal,
I understand your frustration. There is a proper third party. And that is a peer-reviewed refereed journal. Dan can talk about this paper at the APBR forum forever. But that is not a proper blind reviewed forum. My sense is that most people posting there do not understand econometrics well enough to evaluate his argument. Plus, it is not blind review. As JC Bradbury told Dan when he was posting anonymously here — submit your paper to an economics journal and let other economists evaluate your work properly. That is how this should be done. In fact, that is what should have happened a long time ago. I have a real problem with releasing working papers to a crowd and then declaring, before any academic has seen your work, that you have discovered something. That is not how academic research should be done (or in most places, is done).
Animal
December 12, 2007
Thats a fair assessment, especially because most of the people on the APBRmetrics board are anti-WP, so the likelihood is he wouldn’t get an un-bias opinon anyways, unless Dean stated his opinion. I am sure he will see this, and hopefully he does submit it to an economics journal.
Pete23
December 12, 2007
Westy write, “Could we get Justin Wolfers (obviously has an interest in basketball) or Steven Levitt (has discussed Moneyball extensively before) to mediate and sort this out in a day or two??”
Westy, I agree exactly. Rosenbaum can have neutral economists evaluate his work by submitting it to a journal.
Pete23
December 12, 2007
In the meantime everyone should make a prediction when Jason Kidd gets trade to Mavs about what impact will be on each team. It’s only one data point, but at least we’ll have an out of sample prediction. I’ll bet on dberri’s prediction. Harold A, want to take the other side and go with Rosenbaum?
The other Pete
December 13, 2007
Where are you hearing that Kidd got traded to the Mavericks?
The other Pete
December 13, 2007
I see it on Yahoo! Sports now. Kidd rumoured to be going to Mavs. It doesn’t say who the Nets will get, though. This is bad for the Nets. They need to turn good before moving to the new arena in Brooklyn in 2009.
Brian
December 13, 2007
Jason –
When you say
“I’m still trying to figure out their logic in evaluating measures by how they compare to adjusted +/- when adjusted +/- seems to have a poor predictive value.”
You must have missed this paragraph in the paper:
“Because adjusted plus/minus statistics can also be used to rate players, it is sometime argued that this somehow invalidates using adjusted plus/minus statistics as an evaluation method. An analogous argument would be that if someone used team wins to evaluate players, somehow that would invalidate team wins as method of evaluating player metrics. Adjusted plus/minus statistics simply measure how the team does when a given player is in the game and that is precisely what player metrics are supposed to measure. And they do so without using any box score statistics, so they are not systematically biased towards any box score based metric.”
Individual player metrics are not going to naturally be the best means of predicting future team success. Nor, really, are any of these statistical models good at predicting team wins in future years – indeed, it would be remarkably strange to evaluate PER on the basis of how well it predicts future team wins, since that is not its aim. The paper shows that PER is the best method for providing *some* insight on the quality of play of individual players.
Of course, none of that is going to stop Berri et al’s arrogant (and, I would charge, not so subtly racist) attempt to claim to know the “worth” of basketball players better than people who are actually involved in the sport.
Animal
December 13, 2007
If you want credibility when you post, don’t try to attack someone by calling them a racist with really no grounds for it. To expect a response after you make that claim is just ridiculous. If you want to try and argue basketball, fine, go ahead and do it. There is just really no reason for a remark like that.