Shawn Ryan is a 24 year old student currently living in Austin, TX and pursuing dual degrees in Computer Science and Economics. He found the Wages of Wins at a time when he had become newly infatuated with the field of behavioral economics, and it has greatly influenced his educational goals. He has been a fan of basketball from a young age, but has tended to have different favorite teams over time, including the Suns, Cavaliers, Blazers, Hawks, and Rockets among others. He loathes NBA play-by-play commentary, and often has to resort to turning off the sound for the sake of his mental welfare. He is glad to have the opportunity to contribute to this blog, because for some reason, his fiancée tends to fall asleep at unfortunate times while he discusses his very interesting ideas about the NBA with her.
This is part one of a two-part post on the Houston Rockets. In this first part, Shawn will review the various moves the Rockets made in building their 2009-10 roster. Part two will be a look ahead to the future.
The 2009-10 season was a season of flux for the Houston Rockets. The team’s two most recognizable faces, Yao Ming and Tracy McGrady were non-factors for the Rockets; Yao was sidelined for the entire season, while McGrady played only 46 minutes for the Rockets before being shipped off to New York. Additionally, the Rockets swapped Ron Artest with the Lakers for Trevor Ariza. At the trade deadline, the Rockets entered a complicated 3-team trade with the Sacramento Kings and New York Knicks, the highlights of which are detailed below:
Houston Rockets
to Knicks: Tracy McGrady
to Kings: Carl Landry
New York Knicks
to Rockets: Jordan Hill
Option to swap 2011 1st round picks (top-1 protected)
2012 First Round Pick (top-5 protected)
Sacramento Kings
to Rockets: Kevin Martin
So did the Rockets come out ahead in these trades? To answer this question, let us first take a look at how many wins all of the players involved in these transactions produced in and out of Houston. Figure One shows the 2009-10 production of the major players involved in the trades mentioned above.
As a point of interest, the players involved in this trade played nearly the same number of minutes in Houston as they did with the various other teams involved in these transactions. But in Houston these players produced more than five additional wins. Most of this difference, though, can be attributed to Carl Landry’s post-trade decline. So the quick snapshot in Figure One might be skewed. That being the case, let’s look at the production of these players over the last three years (see Figure 2). 
From these numbers, we can see that the Rockets again appear to do quite well in these transactions. Over the last three seasons, the players the Rockets acquired outplayed the players lost when we look at performance per 48 minutes. Furthermore, the players acquired are 5.66 years younger. And finally, the Rockets also acquired two first round draft picks form the Knicks. Unless the Knicks can land LeBron James or Dwyane Wade (while keeping David Lee), it is quite possible these picks from the Knicks will be towards the top of the draft. And since these are the Knicks, a more likely scenario is the Knicks add Rudy Gay, let David Lee walk, and send very good picks to Houston.
Beyond these draft picks, the play of Jordan Hill has also been a pleasant surprise. If Hill maintains the level of production that he showed in Houston after the trade (from Fig. 1: Wins Produced per 48 minutes or WP48 of .181 in 384 minutes), then he will be able to largely fill the hole left by Landry’s absence (of course this is dependent on the number of minutes he is given). And while it is more likely that the production over his career will resemble his performance in Houston than New York (i.e. rookies tend to improve over the season) the sample of minutes is on the small side; so it can not be guaranteed that this is what we’ll see in the future.
On the other hand, neither Ariza nor Martin have played as well as expected this season. Martin’s drop-off in production might be attributed to his playing through injuries for the second year in a row. If he is able to re-attain something in the vicinity of the WP48 he achieved in the 2006-07 and 2007-08 seasons (.154 and .182 respectively) then he will be a very welcome addition.
Trevor Ariza, though, doesn’t seem to have the same excuse. In 2008-09, he achieved a career high of 7.58 Wins Produced with aWP48 of .182 (in just under 2000 minutes). This season, he only produced 4.37 wins with a WP48 of .080 (while having played in excess of 2600 minutes). Fig. 3 compares Ariza’s 2008-09 production with that of 2009-10.
As you can see, Ariza increased his field goal attempts per 48 minutes significantly (by 27.4%) while his shooting efficiency dropped precipitously.
As shown in Fig. 4, Ariza’s shot selection in 2009-10 shifted a great deal in favor of jump shots and away from inside shots. Since Ariza is a below average jump shooter, and a very talented inside finisher, this shift in shot selection was very ill advised. And this change in shot seletion can plausibly be considered cause for his decreased shooting efficiency (though it should be noted that shooting percentages are relatively volatile from year to year, so one should be careful when attributing meaning to changes in shooting percentages, as sheer randomness plays a larger role than most fans, and even NBA “experts” realize).
This increased focus on taking relatively poor shots also coincided with a decrease in steals and rebounds and an increase in turnovers. Consequently, per 48 minute Ariza’s Net Possessions declined 2.94.
Still, Ariza’s points per 48 minutes increased by two, so given the bias of the majority of the NBA’s decision makers to overvalue scoring totals and undervalue stats that increase Net Possessions, it is likely that he has only increased his market value (which was not very high, in spite of his terrific production, at the end of the 2009-10 season). So perhaps there is a market for Ariza’s services.
If a market for Ariza’s services doesn’t emerge, the Rockets would be well served if Coach Rick Adelman and his assistants did their best to influence Ariza to shift back to his prior focus on possessions and efficient shots. If this happens, maybe we’ll see Ariza once again break his single-season wins production record. And perhaps then we’ll see the Rockets back in the playoffs. In the second part of this article, we will look at that very prospect.
– Shawn Ryan
The WoW Journal Comments Policy
Wages of Wins Journal 
ilikeflowers
June 7, 2010
Uh oh, another Trevor Ariza usage thread!
Shawn Ryan
June 7, 2010
ilikeflowers:
=O
dberri
June 7, 2010
Palamida had a great response last time. I am ready to re-post this comment if Palamida doesn’t want to type it all out again :)
Alvy
June 7, 2010
Ariza strikes again!
Filipe Furtado
June 7, 2010
Shawn , is it possible to look at Ariza’s numbers after the Kevin Martin trade? He was heavily feauture on offense early this year and mostly used (and behaved) like a “star” (in the conventional sense), I think this made his offensive efficiency drop and made him go away from all the little things he did well in 2009. From the few Rockets game I saw late in the year he looked closer to a role player but still setting for jumpers too often.
Mok
June 7, 2010
A better look at Ariza in Houston would be to compare how he fared with and without Martin, ie before and after the Martin trade.
ilikeflowers
June 7, 2010
Another good post, BTW.
Italian Stallion
June 7, 2010
I agree that there may be some value in looking at Ariza before and after the Martin trade.
Once Martin was added, there was no reason for Ariza to retain the higher usage level and take as many jumpers. Martin is the much better offensive weapon.
It might also be worth looking at his Ariza’s shot distribution in what could be a more relevant way.
Let me explain.
Every single player wants as many dunks and layups as possible. The fact that they don’t get all dunks and layups is not a solely matter of their decision making. It’s also a matter of availability.
As you add usage, you typically don’t get to add an equal proportion of dunks and layups as you had at lower usage levels.
(If you did have more dunks and layups available you would already be taking them).
Therefore any addition of usage is likely to be tilted towards taking more of the tougher jump shots. That automatically makes it look you are taking more poor shots ( a matter of decision as represented in the table above) when you really have no way to increase your usage without taking more of those low percentage shots.
So I wholeheartedly agree that if Ariza stopped taking as many jump shots he could increase his efficiency again.
However, to do so he would also have to reduce his usage because he won’t be able to substitute all dunks and layups for those jump shots.
If he could, he would already have done it because he and most NBA players and coaches are not total morons.
Shawn Ryan
June 7, 2010
Thanks ilikeflowers
Tindall
June 7, 2010
dberri – palamidia and this blog entry (which was very good, by the way) argue for usage, not against it (although perhaps my understanding of usage differs from yours). Palamida even admitted that players like Andris Biedrins could not dramatically increase their scoring while maintaining their shooting efficiency because Biedrins’ scoring comes primarily close to the basket on putbacks and open passes. The fact is, if you ask a player to take on responsibilities they do not excel at, they will do more poorly. In fact, don’t you say as much in Wages of Wins?
On page 125, you and your co-authors write that “knowing the value of each player is only the starting point of analysis. The next step is determining why a player is productive or unproductive. In our view, this is where coaching should begin … Although we have offered some insights into why players are productive, ultimately this question can only be answered by additional scrutiny into the construction of a team and the roles a player plays on the floor.”
Mok – Houston acquired Martin at the All Star break. Check out Ariza’s splits here: http://www.basketball-reference.com/players/a/arizatr01/splits/2010/
Upon acquiring Kevin Martin, Ariza shouldered less of the load on offense (19.1 usg% which is closer to his career average – it was 22ish at the beginning of the season IIRC) and played fewer minutes (33 mpg, down from 38mpg). His scoring efficiency increased (44% FG, 41% 3pt%, up from 38% FG%, 30% 3pt%) , as did his steals and rebounding (offensive boards, in particular). What does this mean? You be the judge.
Frankly, I’ve been playing devil’s advocate a lot on this blog because I disagree with a lot of friends and sports journalists on these issues, but I’m often incapable of completely dismantling some their arguments. I know how I would respond to certain arguments on Kobe’s behalf – but how would some of you? It’s my hope that by asking a few questions and challenging some of you experts, I will move a bit closer to the truth.
Oh, and Itallian Stallion deserves recognition for predicting Ariza’s downfall when the trade occurred.
dberri
June 7, 2010
Tindall,
I guess I have to repeat what I said every other time we mentioned the Rockets. A number of players on the Rockets increased their usage. But only Ariza saw a significant decline in productivity. If Ariza supports the usage argument, don’t all the other player lead us to reject this argument? You really are not doing objective analysis when all you do is focus on the results that confirm what you already believe and ignore everything that doesn’t.
And again I would note… I have looked at 30 years of player data. Changes in shot attempts do not lead to a very large change in shooting efficiency. So whatever people think they are seeing with respect to Ariza is not something we generally see in the data.
palamida
June 8, 2010
The thread that has been mentioned here is the lengthy “Aaron Brooks is not the Mip” thread. For anyone interested in further reading and analysis on this subject – namely – usage vs. efficiency. I recommend looking it up.
I don’t have any intention of re-posting the lengthy comments Myself (and others) have made on this subject in the past.
I’d just like to add one more issue that wasn’t mentioned in that thread:
Synergy have a strict policy regarding the public use of their data. (which is why I personally never bring up those stats). To all of you who have access to it I recommend the following:
Take the time to watch ALL of Ariza’s plays labeled “P&R Ball handler”.
What in the world was Adelman thinking?
Without presenting the numbers , and in truth the numbers only tell half of the story – for the whole story – I guess you have to see to believe…
There are many methods in which a certain player can “increase” his usage – Houston\Adleman\Ariza chose a very poor method for Ariza (btw this went on long after Martin was acquired).
so if you can – take 15 mins. to watch it – it’s… painful.
For many other aspects of this discussion (on both sides of the “fence”) I would recommend reading the comments (in full!) of the “Aaron Brooks” thread.
Paul
June 8, 2010
Dberri,
You might want to consider what sort of factors lead to changes in shot attempts.
Rather than doing this, you seem, as in your per-play analysis of quarterbacks, to treat coaches’ decisions as random draws from some underlying distribution.
This means your “30 years of player data” approach falls well short of proving what you seem to think it does.
Tindall
June 8, 2010
The usage argument is that certain players are able to increase their usg% without a decline in offensive efficiency until a very high usg% while other players experiment a decline in efficiency at lower levels of usage, correct?
The reason players like the Michael Jordans and the Shaqille O’neals can maintain efficiency at a high usg% because of their offensive skillset, which allows for a variety of efficient ways to score. A player with a limited offensive skillset who is asked to take more shots will suffer, as we saw with Ariza.
When you write “ultimately this question can only be answered by additional scrutiny into the construction of a team and the roles a player plays on the floor” you seem to understand that some players are limited in certain areas and consequently it would be a bad idea to force those players into areas they do not excel. This post’s analysis appears to confirm that.
Thus, I do not understand your point when you write “A number of players on the Rockets increased their usage. But only Ariza saw a significant decline in productivity. If Ariza supports the usage argument, don’t all the other player lead us to reject this argument?” because the usage argument expects that players of different skillsets will thrive in different roles. We would expect a player with a diversified offensive skillset – one who draws fouls, is an excellent jumpshooter, and can slash to the basket – to maintain efficiency with increased usage. We would expect a player whose offensive game is limited to putbacks to suffer if he/she was forced to take significantly more shots because that player would be forced to take more jumpers, handle the ball, and deal with double teams. The usage argument explains why Brooks, an excellent jumpshooter from all distances, can take more shots without seeing a decrease in offensive efficiency. It also explains why Landry, a great mid-range jumpshooting big with awesome hands and lethal under the basket, can take more shots. And so on.
“You really are not doing objective analysis when all you do is focus on the results that confirm what you already believe and ignore everything that doesn’t. ”
I’m not entirely sure what I believe, and I’m certainly not sure what you believe. On one hand, you agree that Ariza’s efficiency decreased because he took shots he could not make, which supports the usage argument that when a limited player takes on a larger role of the offense that player’s efficiency will suffer as he is forced to take poorer shots. On the other hand, you deny usage plays a significant role in offensive efficiency. I’m waiting to read a definitive rebuttal to the “usage argument.”
“Changes in shot attempts do not lead to a very large change in shooting efficiency. ”
Perhaps because coaches recognize the strengths and weaknesses of their players and use them appropriately i.e, not give Trevor Ariza 18 shot attempts – mainly jumpers – per game. That’s just one possible explanation, but perhaps you could offer a ‘why’ explanation that offers further insight.
Jamaal
June 8, 2010
+1 to Tindall.
Man of Steele
June 8, 2010
Thanks Shawn. I really appreciate your write-up, especially the analysis of Trevor Ariza’s changing shot selection/distribution.
As a note, I think you’re right that the trade numbers are a little skewed. Carl Landry had a bad stretch, but he’s been a good player for 3 years, so I would think that he will be more likely to revert to the .170-.200 range in future.
Shawn Ryan
June 8, 2010
I think that the effects that usage advocates claim, are incorporated into the Wins Produced model. Basically what the model tells us is that wins come from the confluence of three factors: offense, defense, and possessions.
A player trying to increase his Wins Produced score would try to get more rebounds and steals, avoid turning the ball over, do his part to help team defense (though WP doesn’t punish individual players for failing to do this as much as would be ideal, but it does incorporate it via team defense), and take high percentage shots (i.e. shots that are high percentage for him/her, so yes, this factor is dependent on the players offensive skill set). There are of course other factors but those listed above are the most consequential.
The problem with Usage, from my perspective is a lack of statistical rigor, objective perspective, and systematic projection (and I readily admit that my exposure to the idea is mostly limited to the commenters on this forum, if any of you are able to provide white papers, peer reviewed articles, or any more systematic, broad base, empirical and objective workup on the subject, then I would be happy to read it and revise my stance as necessary, but without any of these things, I see no need).
It’s one thing to say “Ariza won’t do as well with the Rockets as he did with the Lakers because his usage will go up and he is not a skilled offensive player.” If no forces other than randomness come into play, then you have a 50% chance of being right because he will either improve or decline, to varying degrees. It’s quite another thing to create a model, like Wins Produced, that says “Here are the ratios of how these myriad factors affect wins. By plugging league numbers into the model, we will be around 2% of predicting the number of wins that teams have achieved (in aggregate, for specific teams it jumps around a little more, but generally the outliers in a given season seem to be off by less than 5%), and what’s more, we can tie those wins to individual players.” The second claim is several orders of magnitude more ambitious, and more generally useful.
All that I can see usage being, given my current level of understanding (and I have read all of the comments on the Aaron Brooks MIP article, btw) is a subjective explanation of a small subset (when compared to Wins Produced) of the factors that affect outcomes in the NBA, that completely lacks statistical rigor.
(note: I’d meant to expound on this part more, but have to go, I may finish my thoughts here when I get back) And that’s absolutely fine. Offense is generally more fun to watch in basketball than rebounding. As fans, we can focus on scoring because that’s what pleases us. But it is not really appropriate to compare Wins Produced and Usage as competing models because the scope and rigor of the former is so much greater than that of the latter.
Further, I don’t think that Usage is something that should be incorporated into the articles of this form to a much greater degree than which I’ve done in this article. It may be interesting to note how specific players took more shots while seeing decreases in shooting efficiency, but until Usage achieves a higher level of rigor, such notes should always be succeeded by a caveat.
I don’t know if this comment will stir the hornets nest, so to speak. It is certainly not my intention to do so, but merely to explain my reasoning for not incorporating Usage into the system that I use for trying to understand the NBA. I’m afraid that at this point, the only arguments that I see as having any chance of swaying my view are those that I requested above: white papers, peer reviewed articles, or any more systematic, broad base, empirical and objective workup on the subject. I will certainly read all replies, but I hope that those replies defending usage are accompanied by supporting academic work.
Dre
June 8, 2010
“The usage argument expects that players of different skillsets will thrive in different roles”
“The reason players like the Michael Jordans and the Shaqille O’neals can maintain efficiency at a high usg% because of their offensive skillset, which allows for a variety of efficient ways to score.”
This argument seems very subjective, at best. Dr. Berri is pointing out that the aggregate data shows a simple model of more shot attempts does not typically lead to a dramatic decrease.
This does not explain Ariza, of course but even the best model won’t be 100%. In fact it’s funny as I have seen the argument “Players like Ariza and Biedris clearly can’t take more shots, they don’t have the skillset, players like Jordan and Shaq can!”
Doing more in depth analysis of what happened to a player as Palamida did is very useful. Relying on a subjective term to explain anomalous behavior and pretending it trumps aggregate research does not.
Alvy
June 8, 2010
Tindall,
I believe palamida’s post in the Aaron Brooks thread is the “‘why’ explanation that offers further insight” that you may be looking for.
Marparker
June 8, 2010
IMO the usage argument is really the three pointer argument. Changes in 3pt frequency have drastically effected several players wp48 consistency. Ariza took 5.6 3 pointers per 36 minutes last year. He is a career 32% three point shooter. If a player is taking the majority of their shots from 3 then they are completely taking themselves out of the offensive rebounding game. So, if he is a pretty good rebounder/slasher and a bad three point shooter of course his production will decrease if he starts shooting more 3 pointers. Delonte West is an example of the opposite. His production goes down whenever he’s not shooting enough 3’s. Josh Smith is another example of a guy who’s 3 point attempts greatly effected his production.
Shawn Ryan
June 8, 2010
Man of Steele
Thank you for the kind words.
I also expect Landry’s numbers to rebound as long as he continues to get a chance to play. Subjectively, it seems that it often takes a while for a player to return to prior levels of production after a trade. I think I vaguely recall the subject of the effect of roster consistency on production shifts being discussed by Dr. Berri, but I don’t remember when it was, or the details of the discussion.
todd2
June 8, 2010
What do the Rockets need to address during the off-season?
Mo
June 8, 2010
Usage is a red herring, the issue is role in the offense. When Ariza played for the Lakers, his role in the offense was to take open 3s and slash to the basket when his man cheated off of him. He was the 3rd or 4th option in that offense. However, when he went to Houston, he became the top banana in the offense and the shots that are available when the defense is keying on Bryant and Gasol aren’t available when the defense is worried about Brooks and Scola. Aaron Brooks’ role is essentially the same, but at a higher usage, which is why his WP48 is essentially the same.
The only thing the Rockets can do to get Ariza his old shot selection is to get other players good enough for the defenses to focus on and sag off of Ariza, preferably in the triangle offense.
Italian Stallion
June 8, 2010
D.Berri,
“I guess I have to repeat what I said every other time we mentioned the Rockets. A number of players on the Rockets increased their usage. But only Ariza saw a significant decline in productivity. If Ariza supports the usage argument, don’t all the other player lead us to reject this argument? ”
You keep repeating this and we keep repeating the same response.
Each player has a unique skill set and scoring range. At any given time a player may be under utilized, properly utilized, or over utilized depending on the other scoring options on the team (and other factors).
So understanding and predicting the impact of an increase in usage is not something you can plug into a one size fits all formula the way mathematically oriented people like to do.
It requires an analysis of the individual player and what he will be asked to do if his role is increased or reduced.
Shawn Ryan
June 8, 2010
todd2
That is precisely the subject of my next post =D
dberri
June 8, 2010
Yes IS, you do keep saying this. As others have noted (and I have noted repeatedly), an anecdote is not a replacement for systematic analysis.
I have yet to see systematic analysis that supports the usage story. Your comments, though, suggest strongly that you don’t think this is possible. So essentially our differences are as follows: I think you have to have systematic analysis before you reach a conclusion (and an anecdote is not systematic). You argue that you get to believe whatever you want regardless of how the evidence is constructed. Not sure we are ever going to resolve this difference.
3pointersmatter
June 8, 2010
If a guy shoots too many 3’s his production will go down. This is not some magic skill set observation. Nowhere in the Bible of Basketball does it say a first option must jack 3’s even if he can’t make them.
Italian Stallion
June 8, 2010
I should add that we typically don’t see wide variations in efficiency as usage changes because coaches are typically smart enough to define roles for their players that match their skill set.
If the player improves relative to other players on the team, his role will expand and vice versa.
Also, most teams are built without glaring scoring weaknesses.
3pointersmatter
June 8, 2010
Some times if you watch games you will see guy shooting from behind a line. I’m not sure if you guys who don’t watch basketball and only look boxscores know that this line is the three point line. Thus a player is rewarded 3 points for making only one shot. When a guy who shoots the ball from behind this line is not good at making those shots it will drastically effect his wp48. See:Josh Smith
3pointersmatter
June 8, 2010
Sorry but I am being copy cat of the the guy who already copy catted the guy who is seriously annoying the rest of us who come here to enjoy learning multiple views on certain topics.
Shawn Ryan
June 8, 2010
Italian Stallion:
“So understanding and predicting the impact of an increase in usage is not something you can plug into a one size fits all formula the way mathematically oriented people like to do.
It requires an analysis of the individual player and what he will be asked to do if his role is increased or reduced.”
That, exactly, is what makes Usage unacceptable for the purposes for which we use Wins Produced. Post-hoc rationalization is built into the Usage “model”. It is not useful for explaining wins in the NBA in a scientifically rigorous fashion, therefore, as you find, it is rejected by the more scientifically minded people of this forum.
As I said above, if you are trying to come up with a story for why something one thing or another happens, then Usage is fine. It is not, however, scientifically rigorous, so we largely reject it.
3pointersmatter
June 8, 2010
What I am really saying is how much more important my thoughts are than the rest of you idiots who think/know differently from me. I will not give you a chance to scroll past my comments that you don’t agree with. Instead I will either kill the thread or you will be forced to address me even if you address my posting style and not the substance of my comments.
Italian Stallion
June 8, 2010
Dre,
“Doing more in depth analysis of what happened to a player as Palamida did is very useful. Relying on a subjective term to explain anomalous behavior and pretending it trumps aggregate research does not.”
I agree with you.
I suspect that many NBA clubs have the time and resources to evaluate the skill set of their players well enough to demonstrate things like this quite clearly.
The rest of us watch tons of games, look at some stats, and occasionally bring insights to the table about players based on that more limited analysis.
However, as I’ve suggested in the past, not everyone is blessed equally in the visual skill/observation department and sometimes the insight is so darn obvious to any serious fan, there really shouldn’t be a debate.
3pointersmatter
June 8, 2010
PAY ATTENTION TO ME.
Even though I’ve stated everything useful that I have to add to the discussion in my username
marparker
June 8, 2010
How do you quantify a skillset when it is constantly being improved by players who have endless time to work on their game?
Shawn Ryan
June 8, 2010
Heh, looks like Dr. Berri and I were responding to IS at the same time. My comment is somewhat redundant.
Italian Stallion
June 8, 2010
D.Berri
“I have yet to see systematic analysis that supports the usage story.”
Neither have I.
However, if you go to the basketball court this afternoon and take shots from everywhere on the court and track your results I suspect you might conclude that you are better at some things than others and should stick with what you are good at instead of expanding your usage.
“So essentially our differences are as follows: I think you have to have systematic analysis before you reach a conclusion (and an anecdote is not systematic). You argue that you get to believe whatever you want regardless of how the evidence is constructed. Not sure we are ever going to resolve this difference.”
I argue that should not suspend commons sense and intelligence simply because you don’t have the data yet to prove what you already know to be true.
Shawn Ryan
June 8, 2010
“The rest of us watch tons of games, look at some stats, and occasionally bring insights to the table about players based on that more limited analysis.”
This is the kind of analysis that leads to bad decisions. There are many cognitive biases that severely limit the efficacy of subjective observation. And as for “occasionally looking at stats”, stats can easily be misinterpreted, which is why you need a vigorous and systematic analysis, such as the Wins Produced model, to be able to make solid claims about causality.
Shawn Ryan
June 8, 2010
Italian Stalion
“you don’t have the data yet to prove what you already know to be true.”
The definition of post-hoc rationalization. I don’t understand why you are on this forum. Surely your time would be better spent in a forum where the people care less about logical fallacy, and scientific rigor. Your arguments are fine as a fan, but are anti-scientific.
Dr. Berri is an economist. If you are not interested in following the accepted foundations of logic that that , and so many other scientific fields are predicated on, then why not go to a non-science based forum? It’s absolutely fine to do so. But you’ll never convince someone who holds scientific rigor as a predicate for the acceptance of a conclusion to accept the non-rigorous conclusions that you’ve made.
ilikeflowers
June 8, 2010
One thing that probably can be done systematically is to assign players a position based upon their shot distribution which can be approximated by looking at %3pointers taken, %2pointers taken, %fouled (all per minute), and height and weight – this will likely spit out 10 to 15 positions. Then one can look at what happens to productivity when a player changes from one of these hybrid-positions to a different one and see if there’s something predictable going on. It would also be interesting to see if productivity stability is greater when using these extra positions. If it is then it’s likely measuring something real and better than the standard positions are, if not then it’s just unnecessary complexity (or the shot distribution/body type model is a poor predictor of ‘true’ position).
ilikeflowers
June 8, 2010
Not to sound like a scratched CD, but I get the sense that there are two meanings of usage that are getting tangled up sometimes. One is simply how much a player is being used (minutes) the other is exactly how a player is being used (position or role).
chibi
June 8, 2010
do you guys feel that box score categories need to be re-conceptualized so as to present a more precise measurement of player productivity?
what does one make of a player who alters 250 shots, and deflects 150 passes, but registers 0 blocks and 0 steals?
ilikeflowers
June 8, 2010
chibi,
That is the same as asking, would one like one’s model to be more accurate. The answer is always yes. The lack of perfect data, doesn’t impact the current accuracy or usefulness. At some point you reach diminishing returns in terms of adding more data, we may or may not be significantly into that territory already with current box score data.
Are there any such extraordinary players who even come close to such an incredible stat line where a pair of (presumably) strongly linked stats have effectively no relationship? Regardless, there will always be outliers, the only issues are minimizing the occurrences, the magnitude, and the number of scenarios under which they can occur.
Arturo
June 8, 2010
Chibi,
If we say that possession and points should be the main drivers for value in basketball then there are opportunities in the Box Score. Charges are one example (the offensive player gets debited but the defensive player does not get credited ). Blocks are another (does a blocked shot by Dwight Howard that goes out of bound get credited as a block and a turnover on Howard as it should).
I feel there are some players/skills who get undervalued or overvalued (say Glenn Davis and his propensity for drawing charges). but there is still enough information there to draw informed conclusions .
IS,
As for Ariza and usage, I think what we are seeing is a function of how the data is compiled. WP48 data is compiled in aggregate (multiple games) regardless of how the individual player is used. Swing Players or multiple position player are going to be vulnerable to skew because of this (because they’re playing with the wrong personnel in the wrong position etc.). As we progress perhaps a better method would be to calculate the numbers on a per minute basis (Player A played 20 minutes at PG and 10 at SG) and make adjustments accordingly. Theoretically something like adjusted WP48 should be possible. I would however clarify that I don’t think we are gaining that much more information as to player value compared to the effort of generating the numbers (after all this is a hobby for most of us and not a career).
John Giagnorio
June 8, 2010
I find it very hard to believe that anyone altering a bunch of shots and deflecting a bunch of passes is recording 0 blocks or 0 steals. Maybe in 1 game, but over any meaningful sample size it would seem the guy blocking the most shots is also altering the most (you block some % of what you alter) and the guy recording the most steals is recording the most near steals (same reasoning). However, I also can’t imagine there being very reliable data to test this out.
ilikeflowers
June 8, 2010
I think that a much better indicator of the limitations of the box score due to a shot changer is a player who is so adept at changing shots, that the other team significantly alters their shot selection, so as to minimize said player’s shot changing opportunities. A player who is so good at something that they produce the absence of it in the box score is a difficult thing to credit using the box score. These types of players are presumably rare however and the effect could be small so we really don’t know how this works in reality. The impact would likely be captured at the team level (unusual shot distribution skews and – presumably – reduced pps for the opposing team) even though it is primarily caused by one player.
jbrett
June 8, 2010
ilikeflowers,
I like the direction you’re going there. Bill James addressed a similar issue with baseball’s defensive stats–namely, how can Johnny Bench be acclaimed as one of the very best defensive catchers of all time, and have stats that say he was terrible? As I recall, his theory began with the idea that Bench’s skills altered the behavior of his opponents, and built from there on ways to try to measure the change.
I have no idea how that might translate to this discussion–but I’ll bet someone here does. Feel free to do a bunch of hard thinking and number-crunching, and get back to us dilettantes with some results. And thanks in advance :-)
ilikeflowers
June 8, 2010
jbrett,
I can’t take any credit for the concept of course. From the NFL world, every once in a while a true ‘Shutdown’ corner emerges in the league and he then promptly disappears from many of the statistical categories because the opposing team figures that they’re much better off picking on the other players in the secondary. Nnamdi Asomugha is currently this guy in the NFL, with ‘Prime Time’ being the previous incarnation. Of course it’s much easier to refocus your efforts in the NFL since there are more players and interactions.
jbrett
June 8, 2010
Great example with Asomugha; I’m a big Raider fan, so I would have gone there myself were I not traumatized. The Big Red Machine is a memory that DOESN’T make me want to hit myself in the forehead with a brick.
ilikeflowers
June 8, 2010
Yeah, Asomugha is to Da Raiders as KG was to Minnesota. Maybe things are looking up for you though, Jason Campbell is an average QB, which is a tremendous upgrade from what you used to have.
jbrett
June 8, 2010
Yay, average! Now give me a coach who’ll make me shout, “Yay, sane!” and an owner who makes me go, “Yay, not senile!” and maybe I can stay home on Sundays in the fall.
Sam Cohen
June 9, 2010
Arturo- to the best of my knowledge, a blocked shot does not create a change of possession. The change of possession only occurs after a player on the defensive team grabs the ball– and I believe that the player is then credited with a rebound for doing so (and an offensive player is also credited with an offensive rebound if he grabs the blocked shot). So charging Dwight Howard with a turnover wouldn’t make sense (unless you also start charging a turnover for everytime a defender deflects the ball out of bounds).
Italian Stallion
June 9, 2010
Shawn,
“There are many cognitive biases that severely limit the efficacy of subjective observation.”
I agree. However, there are also loads of flawed models out there put together by people with advanced degrees and stratospheric IQs.
What you have to understand is that just as not all models are equal, not all people making observations are equal either.
At the same time, some things are so extreme there is very little chance for mistake even if you are only using observation.
IMO, it will take time for the scientific and mathematics community to fully explain some of the things that have been learned by the basketball community after decades of trial & error in games and thus eliminate the remaining errors.
The reason I am here is because I strongly believe in the ability of mathematics to eventually answer those questions.
I just don’t think we are there yet.
So I think is a monumentally foolish idea to suspend common sense in some of those situations just because we don’t have the data or ability to prove something.
Shawn Ryan
June 9, 2010
Italian Stallion:
Thanks for the response.
I’ll grant that “there are also loads of flawed models out there put together by people with advanced degrees and stratospheric IQs.” In fact, it appears that the field of econometrics is rife with models that take absurd assumptions in order to achieve some counter-intuitive result that will get them published in a top journal.
I don’t think that Win Produced is of this ilk however. For one thing, it’s based on a game with finite rules which makes it vastly easier to model vs. real world macroeconomic subjects.
Also, Wins Produced has been consistent in its ability to predict wins within a certain error, and has a reasonably small error. That small error tells us that in aggregate, including something akin to Usage, will have diminishing returns. What those additions will do, once systematized, is to decrease the amplitude to which the model over/underrates a relatively small number of players who don’t fit easily into the model.
So, say you’re a GM in the NBA. You have the Wins Produced model, and a list of possible acquisitions. What you might want to do is first use Wins Produced to rate the acquisitions. Work up WP48 per expected salary, and throw that into the mix. Then, think about how each of the players would fit into your system, being critical, and trying to find any areas in which this specific player is over/underrated by WP48, and then, crucially, to come up with a weight to apply to that factor that fits rationally into the Wins Produced model (otherwise you might suffer the “1/n strategy” bias, giving equal weight to unequal factors, see: http://en.wikipedia.org/wiki/Na%C3%AFve_diversification ). And finally, if you’re like Billy Bean, perhaps throw out any players that have factors, such as mental health issues, that might cause the your acquisition to produce less than he/she has in the past. (note: I’m using Wins Produced here because I think it is really the only game in town as far as a far reaching model of wins in the NBA. If, hypothetically, there were another of equal or better quality, you could substitute)
So on an individual decision level, you do have to use [I don’t like the phrase “common sense”, so I’ll just use “judgement”]. If Magic Johnson was available for your acquisition in his prime, but had suffered a freak injury that had forced him into a leg amputation, you wouldn’t still acquire him just because he had an outstanding WP48 the previous year. You have to take into account factors that are not encompassed by your model that might invalidate it’s usage in a particular instance.
Usage, could potentially be part of that process, but it would be difficult to model in a systematic manner. And at best, you are going to be refining 1 of 3 major aspects of the Wins Produced model. At the moment, I wouldn’t incorporate usage in that decision making process were I the one making decisions as there seems to be more evidence contradicting the notion that it makes a large difference on an aggregate scale, and likewise does not contain a systematic way to separate those players to which it does apply, and those to which it doesn’t (i.e. players that have large offensive toolkits, and those with small ones, or players that will suffer efficiency penalties when increasing usage vs. those that will not). And you may say that it is common sense that Ariza has a smaller toolkit than Kobe, and it seems to stand to reason, but how does one go about using this incite to make a decision. I would contend that if you only ever acquired players with “large toolkits” you would acquire a lot of players who produce poorly. If you were to acquire players that had only high WP48s, using roughly the simplified decision process I listed above, then you could expect to fall within 5% of where Wins Produced projects you to fall, assuming no injuries. So that is the continuum, and given the current state of Usage, I would rather be making decisions using the model that offers more certain outcomes, Wins Produced, at least until Usage is systematized and incorporated into a model like Wins produced.
So in conclusion, I do not agree with your assertion that it is a “monumentally foolish idea to suspend common sense in some of those situations just because we don’t have the data or ability to prove something.” At least in as far as it relates to decision making in the NBA. Those “common sense” observations can be incorporated into decision making, but if you do so, especially if you are not careful and systematic in doing so, you run the chance of hurting yourself more than you would have had you just let the model pick for you.
P.S. my interest in this blog is that it is a combination of two of my interests, economics and basketball. Economics is concerned with decision making, so that’s why I keep harping on that aspect of using the model, and less on the “OMG Lebron has a better WP48 than Kobe!!!!” aspects of it)
Shawn Ryan
June 9, 2010
Sheesh! 868 words. Sorry, I just kind of got on a roll…
Tom Mandel
June 9, 2010
This is an interesting thread!
WP48 is a great tool for a GM to use in assessing players for the draft, trade acquisition, FA signing, etc.
Of course, a through-the-roof WP48 (well… WP40) in college doesn’t guarantee equivalent or even good results in the NBA. I give you Michael Beasley. And average or below average results in college don’t guarantee the same in the league. I give you Tyreke Evans.
At the same time, these 2 cases and other similar cases don’t *invalidate* WP48 either. They just tell you that a statistical analysis, however helpful, is not itself sufficient to understand an individual case.
This point should be obvious, but it is often forgotten.
In a similar sense, the fact that *statistically* usage variations don’t impact performance level does not tell you anything specific about Trevor Ariza’s play. It may lead you to look at other issues, but it does not rule out this exact effect.
Thus, when Dave writes “I have yet to see systematic analysis that supports the usage story” — does he mean *in this case?* In what might such “systematic analysis” consist?
And when he writes “A number of players on the Rockets increased their usage. But only Ariza saw a significant decline in productivity. If Ariza supports the usage argument, don’t all the other player lead us to reject this argument?” how is this different from saying “statistically it’s not raining, so it’s not raining today.”
So, use the WP48 tool to do what it is *very* good at. E.g. if as a GM you consistently draft, trade for or other sign guys with high WP48 results you will do *very well.* But that’s not to say that you’ll never miss out on someone terrific.
Ovid
June 9, 2010
You mention in the beginning that the commentators are so terrible that they need to be muted. Mark Jackson called Rasheed a “knock down shooter” and talked about how great it was that Kobe will keep shooting no matter how many times he misses. Rasheed’s the worst player on the court most times and Kobe could just pass to Gasol and get easy baskets on a regular basis. I know this is a complete tangent, but I’m just sayin’.
Shawn Ryan
June 9, 2010
Ovid:
Yeah, Marc Jackson is one of the very worst. I am also particularly unfond of Kenny Smith.
Baby Changing Table
June 11, 2010
I did like the article really much, was really informative and the best part was that only the required part was elaborated, to the point concise information always helps and keeps readers running around digging for the information’s will never require a reread. I really wish spammers read these articles and check how easy it is to be human and respect knowledge.
16:42
Shawn Ryan
June 11, 2010
BCT,
Thanks, glad you liked it.