About a month ago an Ian Thomsen story appeared in Sports Illustrated detailing the resurgent New Orleans Hornets. At the Huffington Post – again, about a month ago — I took issue with Thomsen’s story. For Thomsen, the story was all about changes to both the team’s roster and the team’s front office. This focus, though, struck me as largely misplaced. It seemed to me that the numbers suggested the big story was Chris Paul getting healthy.
Chris Paul and Not Much Else
Okay, another month has passed (as I mentioned). And now we can see that all the changes the Hornets have made – to both the roster and front office – haven’t really made a difference. After 32 games, the Hornets have scored 101.8 points per 100 possessions while allowing 100.2. The team’s efficiency differential – offensive efficiency minus defensive efficiency — of 1.62 is consistent with a team that will win about 45 games across a full season. Yes, this is better than what we saw last year. But this is not much of an improvement. In sum, it doesn’t look like the Hornets – despite all the changes to the roster and the front office – are ready to contend for an NBA title.
When we move from efficiency differential to Wins Produced, we can see why this team is not quite a contender. As the following table notes, Chris Paul is really very, very, good. In fact, Paul is once again among the leaders in the NBA in Wins Produced. And as noted a few weeks ago, Paul is also the primary source of this team’s improvement. Unfortunately – despite all the changes to this team (did I mention, changes to both the roster and front office?) – Paul is still not getting much help.
David West and Emeka Okafor – two of the only four teammates Paul has left from last year – are above average this season (average WP48 – or Wins Produced per 48 minutes — is 0.100). And that result is not much different from last year.
Unfortunately, the rest of the roster is below average. And this result is really not that surprising. The Hornets added eleven new faces. So far, all eleven are below average players. Ten of these players played in the NBA last year, and when we look at what each player did in 2009-10 we see that only Jarrett Jack was above average. In other words, it’s not a surprise that so far the Hornets have struck-out with these additions.
The Hornets experience appears to put an interesting twist on what was observed about the Detroit Pistons about two weeks ago. The Pistons – as has been noted before – brought back essentially the same players this year and are getting essentially the same result. Each time this is noted (and it has been noted more than once) it is argued that players are quite consistent in the NBA. So if you want your results to change, you need to employ different players.
The Hornets experience this year suggests an addition to this rule. Specifically, if you want your results to change, you need to employ “better” players.
And until the Hornets find “better” players, the tragedy of Chris Paul (like the tragedy of Kevin Garnett observed once upon a time in Minnesota) will continue.
Trevor Ariza and the Usage Story
One player the Hornets added was Trevor Ariza. In the past Ariza has been above average. Last year, though, Ariza went to the Houston Rockets and saw his production drop. Fans of the “usage” argument immediately cited Ariza as evidence that the “usage” story is correct. Just to review… the usage story is that players who see their shot attempts rise will see their shooting efficiency decline. Last season, Ariza saw his shot attempts rise with Houston — and just as the usage folks expected — Ariza’s shooting efficiency dropped. So naturally, the usage story was clearly confirmed.
Last year I noted, though, that a number of other players saw their shot attempts rise with the Rockets and their shooting efficiency didn’t really decline. In response to this observation, though, people argued that all that matters is what happened to Ariza (yes, people really made such an argument). Apparently, you only need one piece of evidence to believe a story.
Obviously I found this story unconvincing. And this is because I have actually looked at the link between shot attempts and shooting efficiency.
Here is what we say in Stumbling on Wins (p. 206):
The key issue in looking at scoring is not shots taken but shooting efficiency. It’s thought that a player who takes more shots will see his shooting efficiency decline. If you regress shooting efficiency—measured with points per shot or adjusted field goal percentage—on shot attempts you do not see this relationship. However, if you look at the link between the change in a player’s shooting efficiency from season to season and the change in his per-minute shot attempts, the expected link is seen. The effect, though, is small. To see how small, imagine a player who takes 16.3 shots per 48 minutes and has an adjusted field goal percentage of 48.4% (these are the league average marks). If that player increased his shots per 48 minutes to 25.3 (a two standard deviation increase), his adjusted field goal percentage would be expected to decline to 47.1%. This is a very large increase in shot attempts, but it only appears to reduce shooting efficiency by about 1%. Consequently, players have a clear incentive to shoot as much as they can. Even large increases in shot attempts don’t diminish efficiency very much. But such increases do add to scoring totals, and more scoring will lead to more minutes, money, and fame. One should note that there is no statistical link between a player’s shooting efficiency and the shooting efficiency of his teammates. In other words, playing with teammates who tend to hit their shots will not make a player a more efficient shooter.
The above paragraph reviews a study that looked at thirty years of player data (and thousands of observations). Although such a study might be convincing, we all know that all that matters is Trevor Ariza. And this year Ariza — relative to what we saw in 2009-10 – is taking fewer shots. Again, the usage story is that fewer shots will lead to increases in shooting efficiency. So we should not be surprised that Ariza’s shooting efficiency has suddenly… okay, it has actually gone down a bit more.
Wait, that can’t be right. The usage story is that fewer shot attempts leads to much better shooting efficiency. Again, we see there is a link between shot attempts and efficiency, but it is quite small. Ariza’s performance is not even consistent with even the small effect we uncover (and that is because the changes in the number of shots a player takes doesn’t explain much of the change in a player’s shooting efficiency) . And Ariza’s performance in 2010-11 is certainly inconsistent with the usage story people trumpeted last year.
Hopefully, though, another player will start taking more shots in the NBA and see his shooting efficiency decline (or take fewer shots and see his efficiency rise). Again, fans of the usage argument only need one anecdote to run with. So let the search begin for a new poster child of the usage argument!! Surely we can find someone to take the place Trevor Ariza has unfortunately abandoned.
– DJ
Wages of Wins Journal 
Evan
December 30, 2010
This should be a great thread! Where’s khandor?
Italian Stallion
December 30, 2010
With all due respect, I still think you are wrong on this issue and it hurts the credibility of the model when it shouldn’t because usage usually changes with changes in skill. That shows up in the model as more wins produced. The rest are lower level changes within the player’s existing range.
To begin with, it obvious that just increasing shot attempts per 48 minutes does not mean efficiency will change.
1. It depends on whether the player was fully utilized before the increase and how much more you are asking of him.
For example, we all know that Lebron James can up his usage from where it is in Miami and remain highly efficient because he’s already done so in his career. But could Landry Fields up his usage to Lebron’s level and remain as efficient? I’d bet my life against a dollar he couldn’t. His efficiency would fall off a cliff. He’s a massively less skilled offensive player at this stage of his career (and probably will be forever) . If he could, he’d do it tomorrow and become a max player.
2. Players are constantly working on their game specifically trying to expand their offensive skills set, shooting range etc… So when we do see increases in usage, it’s often because the player has improved and the team can go to him more often and expect favorable results. That’s why there’s little or no change in efficiency.
3. While an increase in usage sometimes doesn’t have an appreciable impact on a player’s overall adjusted fg%, if you compared the adjusted fg% of “just the incremental shots” to the previous level, the impact would sometimes be much larger and more noticeable.
4. Sometimes a system or offense will get more of the same types of shots for a specific player. In that case, he should have no problem staying as efficient despite an increase in usage.
5. Most players have at least a little flexibility to get a few more of the same types of shots they usually get, but great players seem to have a lot more flexibility than lower level players.
The bottom line is that every player has a different skill set, different shooting range, different strengths and weaknesses, is often playing in a different system or with different players etc… and there is no magic formula that will apply to all players in all situations and tell you how a change in usage will impact their efficiency.
That why people observe games, takes notes on shot locations, evaluate athleticism, take an inventory of skills etc… and try to assess these things. It isn’t a neat mathematical equation (at least this stage).
To deny that some players can’t increase their usage much without negative impact is to deny something that every young kid understands within a few days of first playing the game. During the first week they understand that layups are easier than jump shots, they can’t all get layups all the time, and some kids are better at jumps shots than others etc…. So they start distributing things in a way that makes sense. The kid that doesn’t is called a chucker and they all get pissed off at him.
statsguy
December 31, 2010
It simply means that # of shot attempts is an inadequate variable to capture ‘usage’. Defining ‘usage’ == # of shot attempts may seem like a nice simplification, but obviously it’s woefully inadequate to capture all the nuances that are correlated with player’s role changes. Additionally, have you explored conditional expected values? Average behavior may contain too much noise to tell you any correlation. You may have to control some variables (disadvantage would be that you’ll have fewer samples).
From what I’ve seen, Ariza has declined due to the fact that he’s been asked to create and score whereas in LA, he was primarily a catch-shoot and an opportunistic slasher. Of course things like this are not captured by official stats but you can get a sense if you actually watch the game.
There are so many things that go on the court that are not captured by official stats. Relying on the official stats (and these models based on official stats) to explain everything is obviously a doomed effort. You gotta see what’s going on the court together with numbers to make sound judgments.
nerdnumbers
December 31, 2010
Statsguy,
Love the name. I really need a sound effect to play whenever someone uses a word or phrase like “obviously” or “of course”. The goal of a model is to simplify a problem. When you come back with “the model doesn’t have all of the data” or “the model doesn’t explain everything” the answer (for a good model) should be “True, but it has the important data and it explains what I need it to explain.”
Using classic box score stats and the Wins Produced formula does a fairly good job of modeling wins and losses. Saying it does not capture everything is not actually a valid reason to dismiss it. To dismiss it you need to show why it does not work at modeling what it attempts to model.
statsguy
December 31, 2010
Well, I work extensively with machine learning algorithms at my work and published at respected machine learning conferences so I have some knowledge. And the goal of a stat model to be as accurate as possible about real world phenomena it’s trying to model. It’s NOT to simplify a problem for the sake of simplicity (though simplicity is often preferred for various reasons).
And if you read my post, it wasn’t about wins produced at all. The post was about feature inadequacy (in this case ‘usage’). DBerri is basically modeling all of his beliefs on official boxscore stats (only because a lot of feature variables can’t be computed yet), and then claims that a particular phenomenon doesn’t exist because boxscore stats and his models can’t explain them.
I’m 100% sure that better/more-accurate models exist if anyone would bother doing further research on these. In other popular fields (computer science, statistics, etc.), we’d probably already had 100s of related papers. If I can find some time, I’d spend a couple of weeks on this to try new models and new features.
mr obvious
December 31, 2010
Am I crazy or was ariza’s production his last year in LA an outlier? His first year in houston was actually regression to his historical level of efficiency. Same with this year.
Adam C. Morrisson
December 31, 2010
It was an outlier imo. Ariza was a great outside shooter for 2-3 months.
I remember him being decent in Orlando but nothing starter-worthy.
arturogalletti
December 31, 2010
I find the usage argument interesting. I think I actually disagree with everybody. Wrote a piece on it (http://arturogalletti.wordpress.com/2010/10/31/gimme-the-rock-part-3-the-shooters-dilemma/).
The gist is, players are who they are. Each player has a usage which will provide optimum effienciency and/or win value. Deviating significantly from this point in either direction leads to diminishing returns. By the time players get to the NBA thousands of basketball hours have gone to determine that point. Smart coaches and players stick in the optimum range.
The big problem come when you put incompatible pieces together (like say Iverson and Melo or Melo and Amare, you listening New York??).
curtains
December 31, 2010
Hmm, Dwight Howard has increased his usage. His efficiency has gone down. Well, the added shots are metastasized in midrange shots and paint post ups, not dunks and layups. This isn’t rocket science; if the shots you add are of inferior quality, your overall efficiency will decline. If you continue to add quality shots, you efficiency will not decline – it may even improve.
Ariza’s 3pt shooting has fallen off a cliff. Well, not really; he’s never actually been a particularly competent shooter, outside of one very fluky/lucky finals run. The major difference is he’s not getting shots at the rim this year. It’s not clear why, perhaps he’s not suited as a 3pt specialist. Much better role for Marco Bellinelli.
Michael
December 31, 2010
role ≠ usage.
Michael
December 31, 2010
Also, a long time ago now Kevin Pelton wrote an article about the significant positive impact Kobe Bryant has on the offensive efficiency of his team mates. This Ariza piece reminded me of that.
Patrick
January 1, 2011
Most of these blog posts are formed in a similar fashion. There is the first part:
“Hey, there’s this interesting topic or trend in the NBA. Let’s use WP to look at this and draw some conclusions.”
I find this part to generally be very enjoyable. Then there’s the second part.
“Some people disagree with me and it obviously annoys me. Let me make some snide comments about it.”
I generally find this part to be very unenjoyable.
I think there is a ton of valuable insight on this blog, but I fear the message can get lost because of the presentation.
Happy New Year!
-Patrick
todd2
January 1, 2011
There could be two approaches to improving a team’s success regarding efficiency differential: improving a team’s offensive or defensive efficiency. Have any studies been done on the latter? In the Hornets case, it’s striking that their opponents are averaging more assists than they are when the NOH have one of the best point guards in the league. The assist disadvantage could be the result of a number of things: poor transition, post or help defense, or pressure on the ball at the point of attack (which is unlikely with CP3). David West is a statue and Okafor is an undersized pivot and that may be part of the problem. It could be a cheaper way to go for a team to improve by shoring up defensively if there was a reliable way to evaluate defensive talent. Speaking of defensive shortcomings, Andrew Bogut is leading the league in blocks. Their back court must be a sieve.
Adam C. Morrison
January 1, 2011
After reading the FAQ, all I can say is it’s quite difficult to refute.
BadgerBucco
January 7, 2011
I think Professor Berri mischaracterizes the usage story, at least as I think about it. Usage v. efficiency isn’t really just about shooting and shooting efficiency. In my mind, it’s about used possessions (whether shooting or not) and overall scoring efficiency. Therefore, you should take things like turnovers and ability to get to the line into account. Has anyone done a study that includes those factors?