Ted Leonsis is the new owner of the Washington Wizards (a deal was reported this past week to be in place). And on Saturday he made the following statement about Stumbling on Wins:
David Berri and Martin Schmidt have written a very interesting book called Stumbling on Wins: Two Economists Expose the Pitfalls on the Road to Victory in Professional Sports.
I just concluded reading it. I recommend it highly to the passionate, opinionated fan. The stats speak loudly and clearly.
This is a wonderful counter cultural look at the world of sports by economists and math majors. Buy it today.
If he really did find it a good read, it should benefit the woeful Wizards.
I find this an interesting read on the debate between usage and efficiency.
http://www.countthebasket.com/blog/2008/03/06/diminishing-returns-for-scoring-usage-vs-efficiency/
That’s very interesting. Let’s see what Washington does this summer, eh?
Dr. Berri, please address the usage vs. efficiency research of Eli Witus (Harvard Law Grad, currently works for the Rockets) linked above by Alvin. I am very curious to get your response.
I don’t expect it to matter much for the Wizards. Prof. Berri linked to a story reporting that a Raptors owner read the Wages of Wins and they went & gave $50+ million each to Hedo Turkoglu and Andrea Bargnani.
YESSS. WIZ have a new owner! ESPN has a rumor that the Wizards are considering trading their lottery pick for Michael Beasley, I really hope this is not true. I think he has the ability, he just hasn’t done anything yet.
Alvin and Josh Smith,
How much training do either have in econometrics? If you have some, I think you could each comment on the Eli Witus study.
Looks pretty sound, though the half season sample is slightly smaller than one would hope for. When Ryan Parker (currently a statistics grad student and working for the Blazers) duplicated the study with more data at Basketball-Geek.com he got a somewhat smaller trade-off, but still significant. But, I don’t have a Ph.D. and you do so I was hoping you might have some insight beyond what I am capable of. There certainly could be issues with the study, I’m not suggesting it’s flawless, I just don’t see any flaws.
Josh,
You didn’t answer my question. What is your background in econometrics/statistics? It is not a sound study, but you probably needed to have some exposure to the basics of regression analysis to see the most obvious problem.
Two undergraduate classes, Econometrics and Advanced Econometrics. Also several statistics classes. No post-graduate work.
Excellent. Now look at the model. Is this the way you would contruct a regression model?
I can’t believe this is still actually a debate.
Someone wake me up when Mark Cuban tells Erick Dampier (TS% of close to 65% over the last 4 years) to score 30 points a night and then let’s examine his TS%.
IS//
You mean the same Dampier who increased his shot attempts and boosted his TS% in 2003-04 and shot worse after decreasing his shot attempts the year after? I’m sure some guys are able to take more shots (mostly guards) and retain their shooting % better than others, but the question is, how significant is the effect for most realistic (i.e. not asking what would happen if Dampier takes 20+ shots a game) situations?
Obviously you cannot have ALL non-scorers (a point made by dberri many times earlier) , but for most NBA guards and forwards that’s a non-starter since most of them are pretty skilled ball handlers and shot creators if they made into the NBA as a perimeter player.
I’m being entirely earnest here, yes, it does seem like a very reasonable way to set up a regression model to test a specific hypothesis, in this case that lineups containing several high usage players might outperform their expected efficiency and lineups containing several low usage players might underperform. It is very possible I am missing something, and this study has a had a profound impact on how I understand the usage-efficiency trade off, so please elaborate on what you feel are the problems with this approach.
Hi, Professor Berri. This is Joe Sill, the guy who runs hoopnumbers.com. Congratulations on the attention your book is receiving.
I was interested to see that you had run an analysis on some of my RAPM results, according to your comments in this thread from a couple of months ago:
http://dberri.wordpress.com/2010/03/07/john-hollinger-dean-oliver-and-some-other-people-comment-on-plusminus/
You claim the following:
“I did look at the correlation between 09-10 RAPM and 08-09 numbers. The results are almost the same as we see from BasketballValue (explanatory power of 7.7%).”
It is indeed true that if you regress the *full season* player ratings from 08-09 on the *partial season* player ratings from 09-10-ratings based on data through February 25th-you get an R-squared of 7.7%, which is only slightly larger than the 7% figure you cite for the standard APM ratings from basketballvalue.com.
However, in your footnote on page 183 of “Stumbling on Wins”, you say that the 7% figure for standard APM comes from an analysis of 08-09 player ratings vs. 07-08 player ratings, where in both cases the ratings are based on a full season of data, presumably.
Given that the 07-08 full-season RAPM player ratings are available on my site, why did you decide that an analysis which looked at partial-season 09-10 RAPM ratings vs. full-season 08-09 player ratings was the best way to compare the results to the results you got from the 07-08/08-09 basketballvalue APM study, which was based on full-season ratings for both seasons?
Okay, maybe looking at these models might help. This is on a different topic. But these models have the same flaw as what I am seeing in the study we are discussing. Perhaps by reading the following three paragraphs you can see what I am seeing.
It has been argued that scoring is not that important in the NBA. But this is clearly incorrect. And one can see this with the following “cutting-edge” model. I collected data on team winning percentage and scoring in the NBA across the past four seasons. I then regressed winning percentage on points per game. The results revealed that points scored are statistically significant. Each additional point is worth 0.014 wins (standard error of 0.003) and the model explain 15% of wins. This model clearly demonstrates that scoring is crucial to wins in the NBA.
One might argue that the sample was too small. So I extended the sample back ten years. Again, scoring is statistically significant. And each additional point is worth 0.012 wins (standard error of 0.0016).
All of this is quite “damning” to the notion that scoring is not important. I estimated two “cutting edge” models and they both indicated that scoring is statistically significant. These models are clearly quite sound and well-constructed. And if you disagree, you are not practicing good science.
I think what I have here is a good impression of the type of studies I see on-line. I ran a regression and I included a number of catch phrases in describing my results. To make this complete, just imagine the four or five people have just chimed in telling everyone that my models are quite good.
Despite these catch phrases and support from my readers, though, my models are clearly flawed. What is the obvious flaw in my regressions?
Well, I think in those models one would say that there are some relevant variables omitted… Those models are not really testing what we are interested because of poor specification. I assume this is your point.
However, I’m not sure what important variables you would consider to be omitted from Eli’s specification, besides, perhaps, some measure of opponent quality.
Well, opponent quality is one important issue. But it not the only factor I think the researcher need to control for in this analysis (we take a different approach in Stumbling on Wins that does control for a number of factors).
From what I can tell, he only employed one independent variable. The whole idea behind regression analysis is that we wish to know the link between Y and X, holding several other factors constant. We can’t hold factors constant, though, if the other X variables are not in the model. When you fails to include other variables (with few exceptions) you are really not specifying your model correctly. And therefore, the results — despite the rhetoric I am reading — are difficult to believe.
It should be troubling to people that no one noticed that this model only had one independent variable. So nothing was held constant in this analysis. Again, this really is a fundamental issue in regression analysis. And the fact everyone missed this lends credence to my hypothesis. The on-line people do not do a very good job of evaluating statistical models. Given the training many of these people have received (and their general lack of peer-reviewed research experience), this is not a surprising shortcoming.
Are you referring to the R-squared being so low? I am actually currently taking an econometrics course so I’m really not sure
Joe Sill,
If you like, I can re-do my analysis.
dberri//
I know you don’t believe econometrics can be taught through blogs, but I love reading your comments of econometrics. Like many others interested on these stuff, I’ve had some very rudimentary stats courses in college and even some econometrics, but it surprises me that I learn more about some basic concepts about econometrics from this blog than I did back then.
Also I’ve just finished reading SoTD and wonder why Canadian stations haven’t called you to question about your goalie study? It seems like a very pertinent issue, especially with all the big NHL games going on now.
On a related note, I’ve just found a blog post criticizing that goalie study, although the main point made in the book remains intact.
http://sabermetricresearch.blogspot.com/2010/03/stumbling-on-wins-is-there-really.html
Well, that’s up to you, I guess.
I infer from the fact that you declined to answer my question that you are conceding that you did not employ the ideal methodology for your comparison ;).
Note that you’d also want to control for the set of players being tracked from one year to the next. In your “Stumbling on Wins” footnote, you say examined 239 players when looking at basketballvalue’s APM. You’d want to use the ratings of the same 239 players from my site. I believe the minimum-number-of-minutes-played cutoff is different on my site than on Aaron’s, so if you simply took everyone who was rated in both years then you’d probably get a different set of players from my site than from Aaron’s. I’m pretty sure the players rated on my site are a superset of the players rated on basketballvalue, although if that doesn’t turn out to be true than you might not be able to control for this precisely. It’s good to control for this, if possible, because otherwise the extra players rated from my site which were not rated on basketballvalue will all be low-minutes players whose ratings are likely to be more poorly estimated since the amount of data is small. So it’s not an apples-to-apples comparison if you’re looking at the year-to-year rating variability of such low-minutes players in my case but not in the case of basketballvalue.
By the way, in your most recent post, you were wondering how things worked at the MIT Sports Analytics Conference regarding “discussants” and whether there was an opportunity for alternative viewpoints to be expressed regarding the research being presented. I presented a paper there this March. There was no single “discussant” designated for the papers, but the final 7-10 minutes of the 35 minute presentations were reserved for questions, and there was ample opportunity for skeptical views to be heard.
Whenever the usage/efficiency debate comes up here, I always think back to this post:
http://dberri.wordpress.com/2009/01/11/devin-harris-takes-many-shots-and-the-nets-improve-slightly/
And specifically:
“And yet, his Wins Produced has clearly risen. When we look carefully at Table Two we can see the reason. The primary difference is shot attempts. Whether we look at shots from the field or shots from the line, Harris is simply launching more basketballs towards the hoop. And because Harris is an above average scorer – both from the line and the field (and despite a field goal percentage below 50%, Harris is above average from the field) – the more shots he takes the more wins he produces.”
I agree that this issue is complicated. However, this somewhat disproves the idea that Wins Produced doesn’t give credit for shot creation.
Joe,
I will look at your data (not sure when, though).
Discussants at our meetings have the paper ahead of time and have prepared comments. We also open the floor to comments. But having someone read the paper ahead of time who is given the job of discussing the paper makes a huge difference.
Yes, I’m sure it’s helpful to have someone who’s read the paper carefully discuss it at the conference.
I’m not sure whether you realize this or not, but at the MIT conference, the 4 papers presented during the “research paper” track had already been reviewed by 3 referees prior to being accepted for presentation. Those 4 papers were selected from (if I recall correctly) about 40 submissions. Although I don’t know for sure (since the referees were anonymous) I assume that at least some of the referees were present at the conference and had the opportunity to comment from a position of deeper familiarity with the work.
Professor, how would you go about running a proper test to look at the relationship between usage and efficiency (I know either you ran one, or at least accept a test that was performed by someone else, as you have cited it). While I see your problem with the test performed by Eli, I like the idea of looking at lineups and seeing the difference (if any) between their efficiency and expected efficency
This whole debate seems to boil down to two sides
Side 1…Scorers are important no matter how little else they offer on the court
Side 2…if a guy can do alot of things on the court but doesn’t like to shoot he can shoot more if need be and easily replace the scorer who does little else on the court
I tend to be on side 2.
Dr. Berri,
I appreciate greatly appreciate your feedback and the time you took giving it. I would really like to hear what control variables you would include, if you were to execute a study in the same basic framework (i.e. lineup level analysis) as Eli.
Simon,
I agree that in many situations the relationship between scoring and efficiency can be irrelevant because of the way teams are built.
However, that’s not the same as saying there is no relationship or that there aren’t specific identifiable situations where it is an issue.
When virtually every player, coach, scout, and GM on earth agrees about something I think it’s worth taking a special look at what they think and why. People like this often have decades of experience tinkering with lineups, learning by trial and error, etc.. trying to optimize results.
That doesn’t make them perfect or right about everything, but it does make them worthy of listening to even if they learned a different way.
I don’t think statisticians should start with the assumption that they are correct because their models are showing something different.
Statisticians and their modals are not perfect either. Sometimes it takes time to mathematically capture what others already know.
In my case, if I see a low usage scorer with a WP48 of .25 and compare him to high usage scorer with a WP48 of .18, I feel confident the former is contributing more.
However, if the numbers are .25 and .24, I’ll take the higher usage guy almost every day of every week.
At another level, I might consider it gray.
I can’t prove I am right, but I know I am because there are contributions that reasonably efficient high usage scorers make that the currently available stats don’t capture.
To me it’s up to stats guys to eventually prove it and not tell me I’m wrong because they can’t yet.
Josh,
One issue one should have with line-up studies is how little time any one line-up seems to spend together across a season. Looking at the 82games.com data, it appears most of these line-ups are together for less than 500 minutes.
There is also the issue Dean Oliver raised some time ago. Players are not trying to win 3 minutes in the second quarter. They are trying to win the entire game. Players will vary their effort level across that game, which means looking at specific segments is going to run into difficulties. This is one reason why plus-minus data has such problems.
That is why I prefer to look at aggregate data. In Stumbling on Wins we talk about how a player’s shooting efficiency varies with shot attempts. And we talk about how one player’s shot attempts are impacted by their teammates’ attempts. Each of these studies control for a variety of factors. And neither seem to support the notion that shooting efficiency is impacted tremendously by shot attempts or that getting shot attempts in the NBA is difficult (for NBA players).
I.Stallion, It’s worth noting that we manage to get to the same debates over new test subjects, time and time again. Allow me to take another stab at it through the example you offered: “Someone wake me up when Mark Cuban tells Erick Dampier (TS% of close to 65% over the last 4 years) to score 30 points a night and then let’s examine his TS%.” And just for the record – I can’t believe this is still a debate, either – go figure!
Let me make a bold statement by saying you’re implying his Ts% will go down If Cuban does instruct him to do as you ordered :p
The average center stands (by definition) at 0.100 wp48. That however fails to capture the true value of center in the NBA. 3rd string Centers and other scrubs soak up some of those “precious” Center mins. I don’t have the Wp48 numbers (I can only estimate that using Dberri’s forumula) but I can tell you this last season all Centers in the NBA posted a WS48 of 9.5. However if you examine the the avg. mark of Centers who played at least 20 mins per game you get roughly – 13.
I would put Dampier in that group which is essentially the pool of NBA’s centers whom function as starters and primary backups.
Dampier posted a mark of 14.85 which is a better mark than this group, but not by a huge margin.
You argue (or at least, imply) that Dampier has limited value. Well guess what, so does WoW.
In his last 5 seasons Dampier had not eclipsed the 25.2 MPG mark. For his career he stands at 25.1.
This is probably has more to do with physical issues than with “Usage” issues, but nevertheless, Many players in the league produced more wins for their teams and thus would be rated more positively by the WoW metric. Dberri never suggested this was a per min contest. On a per min. basis Dampier has indeed been historically productive (and consistent). I don’t see the quarrel here, really: he’s been above avg. even if you compare him to his “peers” (the 20+ MPG group) in respect to FG%, Oreb, Dreb, To’s and BLKs while being below average in: PTS, AST, STL and PF (the latter possibly aiding to his limited mins, as well). The fact is that even this select group (20+ MPG) “only” scores 18.5 PP48. Dampier scores 12.36. Off course he uses 6.1 FGA to get there.
You would probably argue (correct me if if i’m wrong) that that is precisely your point!
“Dampier only takes the easiest shots, etc. putting strain on his teammates to “create” offense”.
A couple of things are flawed in this line of reasoning. First of all it’s not like all his FGA’s are dunks – far from it. 17% of his FGA’s last season were jumpshots. 17% were dunks (considerably less for instance, than the mark of the league’s highest “usage” Center – Howard) and 57% more were close shots. the rest were “tips”.
Moreover even when being compared to this “select group” and not to the average, overall NBA center – Dampier is right about average in %assisted, indicating that he didn’t particularly “relied” on his teammates to set him up i.e – “create” for him.
My expertise in econometrics is not even worth discussing – I have none. As such I simply cannot takes sides judging the validity of Witus’s “study”.
I do know that even if his conclusions were firm (Dberri insists that this study is flawed and I tend to think he knows what he’s talking about, but even if this were redone and produce similar albeit valid results) they don’t amount to much:
Dberri has claimed (as would I) that a team consisted solely of “Dampiers” would be hampered. He has never advocated directly or indirectly for the construction of such a team.
The effect that Witus’s study pertains to is only notable at the exteremes – espeically when fielding a lineup that consists of “low usage” players in the 90-95% Usg combined. I don’t think it’s challenging to compile a lineup that wouldn’t suffer in any way that isn’t marginal from this “drop-off”.More importantly, the High “usage” (high PER) players in this league aren’t “made up” the same “way” and it’s a mistake thinking they are.
When we Bind together “high usage” players (especially wings) and put them in a collective bin marked “shot creators” we make a crucial error.
With players recording usage rates in the low-mid 30′s it’s extremely easy to simply assume (without evidence) that by taking all those shots they’re somehow actually helping their teams. Assuming they simply save the team (namely other low usage teammates, especially late in the clock) from taking even “poorer” shots. There’s no evidence to support that kind of thinking.
In the NBA around 37% of the shots are taken within the first 10 seconds of the offense. That’s where offenses do their most damage. the last 3-4 seconds of the shotclock are indeed a less valuable situation for offenses in terms of the EV of PPS. Wouldn’t we expect these so called “high usage” players to take more of those shots? (more than the league’s average, that is) or at least MAKE more of them?
Neither is happening collectively. Some “star” players do in fact take more shots late in the clock (Nowitzki, Joe johnson for example). Others (Durant for instance, take less). Most though, take about the average (yes Kobe, that means you). As for making them? Some “stars” are good or even “great” at performing this (Kobe is “good – Pierce is “great”, for instance) while others are average or just woeful (both Wade, and Carmelo for example, “manage” to take less shots of the aforementioned “clock is winding down” type shots while converting considerably less than the league avg. That would be the league avg. of “those” shots, not shots in general). So all in all we can come to this conclusion: some scorers who do manage to be efficient (Lebron stands out in this respect) andor contribute in other EQUALLY important aspects of the games are in fact, “star players” and the WoW metrics acknowledges them as such. Others are less productive due to either contributing less in “other” departments or being less efficient scorers – or both.
You do not “need” a Joe Johnson in your offense to save the world from the likes of Dampier. You cannot and should not put Lebron in the same “Bin” as you do J.Johnson. Decision makers, as judging by Johnson’s previous (and upcoming) max deal seem to do.
Not All “high usage” is a “blessing”, and not all “low usage” is a curse.
I’ve only touched briefly on some of the issues at hand, I’ll be more than happy to hear your reply (and others) so I could be more specific, and not use the entire WWW hosting capabilities to illustrate my arguments :)
P.S – A question – If Dampier would have been tasked (for some reason) to take the 6 extra shots per 48 that would “make” him average in that respect, you do realize that if he’d only convert them at 50% (which is a very significant drop in efficiency from his current, and career numbers)
WoW would rate him as productive as he is now, while you would probably rate him better? I for one, seriously doubt his efficiency would drop below that (I even suspect it’ll be somewhat better, but that’s just guesswork). You would accuse me of “admitting” I’m wrong – If he takes more shots – his efficiency would drop, ergo – increased usage leads to a drop in efficiency; I cannot say this enough times – yes, Dampier can’t score 30 pts a night without decreasing his efficiency but… so what? He can’t do that anymore than Rose can lead the League in steals, but no one here is harping about that. As for increasing his usage to an average level, I doubt the drop in efficiency would be more than a couple of % pts, which is what I would call marginal. He would still be rated by WoW as an extremely productive player (Albeit per min, and with recognizing that the way the game is played and given the “short supply of tall men” – Dampier isn’t really that far above the average for players that are actually employed with regularity i.e – the 20 MPG group).
He has a very distinct role in the NBA – being a center. He’s not the most offensively gifted center there is, that’s for sure. However to get to an average level (even compared to the 20+ mpg group) he “only” needs to takes 6 more FGA p48. and only needs to score 6 more pts. not get to 30.
If you’re above average in more ways (and quantity) than not – you’re productive. Dampier has managed to do this throughout his career, in spite of being a low scoring Center. and that is the story WoW tells, repeatedly and consistently. Why are you so “gung ho” about Dampier becoming an above average scorer? Why not being on Derrick rose’s back for stealing the ball at the rates of an NBA center (which is simply abysmal)? or on Nowitzki’s for not blocking more shots? all these things contribute to wins in a direct (and equal, fashion) – yet as Dberri demonstrated time and time again, When you’re above average in scoring (PPG), it’s award time and more money for you?
I’m Eric Dampier, Hear me roar! JK.
“My expertise in econometrics is not even worth discussing – I have none”
So than don’t bother trying to criticize Witus’ model, as you did. If you don’t think his findings would go against what Dberri said, than you clearly you don’t understand it. But, Dberri did point out flaws with the model, making the point moot.
After that I stopped reading your rambling, because you still weren’t addressing what IS said and it was all rather long-winded and uninteresting
palamida: Oddly enough on Friday I did a stats compare between Melo (who many consider awesome) and Andre Iguodala (who some consider overrated)
I basically concluded was better than or as good as Melo on all things save two. Melo shoots 8 more times a game and gets to the FT line 2 more times a game.
So real world example of your Dampier just taking 6 more shots a game would probably increase perceived value.
Alvin, I may not have expertise in the field, but I do have a brain, and a set of eyes to go with it. The truth is my comment was directed at IS’s previous comment, his recent comment was added while I was typing mine and I had not read it prior to posting mine. So it’s hardly a mystery why the two appear unrelated.
As for criticizing Witus’s study – (here the brain and eyes part really come in handy) – I wasn’t. Learn to read and then criticize!
I said: I can’t attest to the validity of his study. (that stems from my lack of expertise). Then I argued that even if those results were replicated in a valid study (adding to say that I value Dberri’s expertise and suspect that he wouldn’t point this flaw out ,so many times if this weren’t an obvious flaw, not to mention no one of the self-proclaimed “scientists” here or over at APBR has refuted, or even attempted to refute Dberri’s claim that the study is flawed, further leading me to believe it is in fact flawed). Still with me, Alvin? pay attention now, you might get it the second time:
Even if the results were replicated in a valid manner they would have very little meaning.
that was the argument I was making, using Stallion’s Dampier example (and other examples and points of reference).
I don’t believe I need to have “credentials” argue that, or do I?
and please, Don’t “bother” replying if you don’t understand the argument. From past experience I probably should have foreseen that saying blasphemous things like : “My expertise in econometrics is not even worth discussing – I have none” – would invite Ad Hominem attacks.
Can’t say I’m surprised. Nevertheless I stand by my logical analysis. If you do not agree with the logic, or the arguments presented – state your objections; enlighten us all. “Bothered” or not, I did not in fact, criticize Witus’s work. I didn’t even hint at criticism. I merely argued the small scope, and marginality of it’s conclusions even *IF* it’s conclusions are valid.
Dre, Iggy is actually an interesting analogy. Not only do many consider him overrated, all year long there were talks and rumors about how he’s own his way out (If only there were takers for his hefty contract). In reality If you attempt to incorporate individual defense into the WoW metric (Work along the lines of what Ty does over at the Courtside Analyst) you can really see Iggy’s full nature and value – an elite wing defender. I would take Iggy over Melo in a heartbeat.
One little stat just for fun: in the NBA as previously discussed the worst shots come late in the shotclock. Melo converted those shots last season at 38.5 Efg%. Th mind-boggling thing is that that’s actually a great mark for him, very close to his career high (in 06/07) of 39.1%. Last few season saw seasons as low 32.1%. All of those btw while taking considerably less than the league average in that clock span (relative to his own shot attempts), so if the clock is winding down, and you’re desperate for some “shot creation” I wouldn’t be overjoyed with the “ability” to toss the ball to Melo, the elite “shot creator” and just “let him do his thing”.
@Palamida: Iggy is rated very well by the WoW metric. I believe he is top 5 in SF and probably top 3 in SG depending on where you place him for this last year.
My comment on Iggy vs Melo is that Iggy and Melo are both bad shooters (less than team ave), but Iggy is better at taking fewer shots.
@Italian Stallion:
In regards to statisticians not assuming they are right, well you are absolutely right! Listening to “common knowledge” is an excellent starting point.
In that regard though, you can use correlation rather well. If there is no correlation between common knowledge and its results(e.g. superstars make money for the home team in ticket revenue) then you can discount it.
If there is correlation between common knowledge and its results then you can begin to examine it, but should not take it as fact(e.g. Robert Horry means you will win a championship)
But Melo would also (presumably) not get to the line as much if he were to take less shots. What really makes Igoudala better is that he basically does everything else better (steals, rebounds, and passes the ball better, while also turning the ball over less, and fouling less)
@Alvin: Absolutely! As a G-F Iggy is awesome at everything EXCEPT shooting. As a F Melo is pretty much average at everything and is not good at shooting! But in comparing Iggy and Melo as “bad” shooters, Iggy wins because he is “smart” enough to take fewer shots.
Dre and Alvin,
I elected not to state the obvious and by your responses I’m getting the feeling my words were taken as if I don’t “agree” with the obvious.
That’s my bad. Yes, the WoW metric rates Iggy favorably. When I mentioned his “true ability” I meant that if you incorparte ind. defense in the manner of counterpart production you find his production is even higher. I didn’t mean that it isn’t high as it is. And yes, Iggy is above avg. in just about everything BUT shooting and that’s where he’s value stems from. It’s further enhanced by the fact that not only does he “fill the statsheet ” he also adds to it by limiting his opponents from filling their own sheets. that was my meaning.
One final note, Alvin, you’re correct when you mention the FT’s issue. The numbers (from 82.games) I referenced to (the shot clock distribution) are FGA numbers that do not take into account FT’s. Sadly I have no such data, and as such I can only assume that those numbers, meaning PPS and not Efg% would improve Melo’s blight, somewhat. To what extent, I do not know. I do suspect though, that most of his FTA come earlier in the clock when he’s attacking the basket andor in transition offense and not later, when defenses are usually set, the lanes are clogged and defenses are basically inviting the guy wih the ball in his hands to take a mid-range jumpshot. If that is the case, I doubt the overall picture will change.
There are many ways to skin a cat. Your apparent aversion to single-predictor regression models is merely one opinion. And opinions backed up by misplaced arrogance are still just opinions.
By the way, you’re not even a real statistician, you’re just an economist, so don’t you dare lecture us about understanding regression models.
Joe
I downloaded the RAPM numbers for 2007-08 and 2008-09. In your data set there are 272 players who played both seasons. The correlation coefficient between the RAPM for each season is 0.33. So about 11% of what a player did this season is explained by what he did last year. This is better than APM. But not much better.
stu//
You must be feeling really sorry for all the university students taking their econometrics courses taught by economists who dare to lecture about understanding regression models. No wonder the school system is broken… ;)
I do, actually. Economists often believe they are full-fledged statisticians, which is quaint.
There are many ways to skin a cat, Stu. Your apparent aversion to non-single-predictor regression models is merely one opinion. And opinions backed up by misplaced arrogance are still just opinions.
By the way, you’re not even a real statistician, you’re just a random internet dude, so don’t you dare lecture an actual well published professor about understanding regression models.
Random internet dudes often believe they are full-fledged statisticians, which is quaint.
/ apparently the professor struck a nerve.
Yes it is pretty odd to see a guy on the internet telling a professor of an accredited university who teaches the subject of econometrics “not to lecture” them about the regression analysis. :|
I’m sure stu’s next step would be sending out letters to economics departments in the country, asking them to pull out all those unqualified economists from econometrics courses. That should improve the quality of economics education. Who says internet forum can’t be helpful?
Pingback: Is John Wall the Obvious Number One Pick for the Wizards? « The Wages of Wins Journal
Pingback: Yi Has Been “Successful”? « Roblog
Pingback: The Rebounding Myth « Arturo's Silly Little Stats
Pingback: Argument Stoppers: Rebounds are Overrated rev.1 « Arturo's Silly Little Stats