It will be a few more days before I complete my analysis of the 2009-10 regular season. That analysis will include updating the Wins Produced model with the 2009-10 team data and evaluating each player’s productivity. So the numbers I have right now will change a bit (not very much, though). Given the numbers I have right now, though, I can say that Aaron Brooks posted a 0.037 WP48 [Wins Produced per 48 minutes] this past season. And in 2008-09, his WP48 was -0.004. So yes, Brooks improved. But not much. And he would not be the Most Improved Player for the 2009-10 season if Wins Produced was the metric of choice.
For the media, though, Wins Produced is not the tool used to evaluate the players. The story announcing that Brooks had won this award noted the primary metric the media uses to measure player performance: His (Brooks) scoring average went up 8.4 points from 2008-09, the highest increase of any qualifying player.
Yes – and this is not a surprise – scoring is the story. By the way, here is what I said last fall when discussing what we should see from Houston in 2009-10:
Aaron Brooks – who may lead this team in scoring – will be considered one of the best point guards in the game.
Let me close by noting that the article announcing this award argued that Houston failed to prove the doubters wrong this season. This argument rests on the observation that Houston failed to make the playoffs, as many expected. But this argument misrepresents what was said about Houston when the season started. Before the season started it was expected that Houston was going to be a below average team. I argued, though, that this team should win more than 40 games but struggle to make the playoffs (in other words, not be as bad as expected). This is indeed what happened. And as a consequence, Brooks got to collect some hardware.
Okay, I need to get back to those Freakonomics questions. This process is going slow since I am also watching the NFL draft (as I have noted in the past, this is the best day of the year for fans of the Lions).
– DJ
The WoW Journal Comments Policy
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
Someguy
April 22, 2010
Lions picked Ndamukong Suh, which should make you a lot happier than if they had picked Bradford.
Johnny Y.
April 22, 2010
Good day for lion fans i guess. Virtually impossible for them to blow that one.
dberri
April 22, 2010
Everyone tends to look great on draft day. Especially in the first round. That being said, as a Lions fan I am quite happy today (as I am on most draft days).
arturo
April 22, 2010
Willing to put a fiver on Durant as MIP.
And Suh is a beast you will be very happy.
Alvy
April 22, 2010
MIP is probably Durant, M. Gasol, or Josh Smith
arturo
April 22, 2010
I have (all estimates)
KD with .320 WP48 this year (.175 in 09,+ .145)
Gasol with .239 WP48 this year (.093 in 09,+ .146)
Smith with .232 WP48 this year (.116 in 09,+ .146)
So it’s either Gasol or KD (or just co-MIPs)
More interesting is that KD had a .240 WP48 in the first 41 games ( and I estimate .406 in the second 41 games).
That OKC team is going to be ridiculous next year.
todd2
April 23, 2010
Houston deserves some respect, what with injuries and the level of competition in the western conference.
Greg Dickenson
April 23, 2010
I’m making this post as a Rockets fan.
I think the key is what I take from Berri’s second to last paragraph: the Rockets were much better than expected. Brooks’s award was as much a recogonition of team effort as it is an individual performance.
Anon
April 23, 2010
except that the rockets were just about exactly as good as Berri expected….
Italian Stallion
April 24, 2010
The one minor issue I have with the Houston prediction is that part of it hinged on swapping Ron Artest (below average) for the highly productive Trevor Ariza.
The thing is, many of us predicted that Ariza did NOT have the skill set to expand his usage a lot and remain as efficient. It wound up he didn’t.
So I assume a couple of other players expanded their usage slightly (a more reasonable assumption) and a couple may have also improved overall (we’ll see when the final ratings are posted) and that’s how the team remained competitive.
Other than that, the prediction was pretty amazing in every other way.
(I ordered the new book. Looking forward to the read.)
ilikeflowers
April 24, 2010
IS,
This is one of those interesting questions about each player and how the role actually played differs from the typical role played for a given position. If a player scores efficiently on a limited number of shots do you want them to take more shots? The short answer is yes, but the more complete answer is that you want them to take more shots of the type that they are good a making. Of course Ariza may adjust to his new role in time and see his scoring efficiency return to its historic levels.
And then there’s the question of guys like Al Horford. He takes more shots now and scores slightly more efficiently now, but his rebounding has suffered. In his case the question is does he have the ability (or desire) to shoot more and still grab rebounds? Or does he just need more time to adjust to his new role?
It’d be nice to see players assigned positions (more than the standard box score 5) using a statistical method based upon shot-types, where they typically are on the floor, and how often they handle the ball. I’m sure somebody’s working on that or already has it done. I’m guessing that there’d probably be at least 10 positions using this method.
dberri
April 24, 2010
Brooks, Scola, Hayes, Landry, Lowry, and Ariza all saw leaps in their usage with the Rockets. If usage had such an impact on a player’s overall production, the Rockets should have become the team the media envisioned at the start. In other words, if you think your argument explains Ariza, then you have to explain why you are unable to explain what happened to everyone else on the roster.
Alvy
April 24, 2010
Italian S,
No, the media thought HOU would be worst off because they had no Yao Ming for the entire season, not because Artest left the team.
Anyway, d.berri, have you been watching the playoffs? The series between ORL and CHA is great, and I think Nelson is back to his good play. Consider his box-score for game 3. Madness.
ilikeflowers,
you said, “It’d be nice to see players assigned positions (more than the standard box score 5) using a statistical method based upon shot-types, where they typically are on the floor, and how often they handle the ball,” and to that I direct you to every team in the NBA. Just consider what Zaza Pachulia does, or Anderson Varejao, Jared Dudley or even Dwight Howard.
I hope that draws some light on your liking, mostly because it didn’t seem like an issue that is actually problematic in basketball—ever. Now, if I entire misunderstood, then I apologize :)
ilikeflowers
April 24, 2010
Alvy, I have no idea what you’re talking about.
ilikeflowers
April 24, 2010
Alvy, nothing that you said is relevant to what I said.
palamida
April 24, 2010
it’s true that at the time of the Artest\Ariza “Swap” some folks (including several avid commentators of this blog) Called Ariza a “role player” and thus have predicted that when forced to “create his own shot” and not enjoy the “spacing” as a result of having Kobe as his teammate – he will struggle.
It’s also true that Ariza has in fact struggled this season. but are the two really connected causally?
Let’s examine Ariza’s scoring a little closer:
Ariza was never an efficient scorer, not really.
The avg TS% in the league is about 55%. Ariza has only surpassed that mark once in 08/09 playing in 24 games for the Lakers.
In his first couple of seasons Ariza barely even attempted 3’s. It seems that after hitting a few big 3’s in that championship run Ariza has started to consider himself a valued shooter and that’s simply not the case nor has it ever was, but i’ll get to that soon enough.
the Avg 3p% in the league is around 35%. Ariza has never even made it that far. Mind you, if we only examined players who attempted as many 3’s as Ariza did this past season (5.7 per game, 8th in the league) the average would be higher as we would expect it to be from a group whose prime weapon is the long range gun. (Paging JR Smith- 33.8%? really?)
Where are coaches when u need them, right?
Even with the so called “free looks” he allegedly was a recipient of as a Laker (Kobe-spacing, Remember?) he wasn’t even shooting at an average clip. So i guess by the “role player” theorists theory he should have fared worse in Houston, right? well, he didn’t.
Despite almost tripling his Attempts per game (2.3 to 5.7) he shot the ball at a slightly better clip as a Rocket, in fact his 33.4% mark was a career high. Why is that?
Ariza’s “usage” rose from around 16.6 in his final Lakers season to 21 as a Rocket. In his Rookie season as an Orlando player he stood at 19.2, btw. That is not that a huge leap but what’s more interesting is how did that usage spike?
Let’s further examine where Ariza’s shots have come from:
Basically almost all of his shots were either close shots, 16-23 feet range shots and shots from beyond the arc.
He was much less accomplished “at the rim” converting only 56.3% from that range – a career low and under the league average from that range. The thing is, Ariza actually attempted the same amount of shots from that range per min, so no usage issues here, correct?
from the 16-23 range (a low value shot to begin with) Ariza was never accomplished. Per min again, he attempted almost exactly the same number of shots as a Laker and a Rocket.
As a Laker he was assisted on 59% of those shots (from this particular range) and shot 37%. Nice “assisting” Kobe and Co. Those are shots he never should have taken in first place considering how poorly he converts them (League avg. is around 40% from there). As a Rocket he was only Assisted on 27.7% from that range and his % dropped to a Paltry 30%. Seems significant at a first glance but actually, since we’re only talking about 2.2 Shots Per game from that range that’s not the real story here. Again though, roughly the same number of attempts per min – no usage issues here :)
From beyond the arc though we see a different picture: Ariza attempted per 40 mins 3.8 3’s in his last season as a laker (a Career high by a wide margin). A a Rocket Ariza put up 6.2 per 40 mins.
There’s your usage spike right there, 8th in the league in 3pa, remember?
Seeing as Ariza is a below average 3 pt shooter, that seems a bit of an odd shot selection, now doesn’t it? U wanna Blame it all on the Lack of Kobe? Make something up (that can’t be proved or disproved) about how failing to “create” a better % shot, Ariza was forced to put up 3’s? Go right ahead, but first see if u can explain why he shot all those 3’s at a better clip.
In addition Ariza was always somewhat poor and erratic from the line. in his last season as a laker he did however convert a semi-respectable 71% of his FTA. He dropped to 64.9% as a Rocket, which further hurt his production; Wanna blame that on Being “Kobeless” as well?
In summation let me just say that Ariza did in fact have a “disappointing” season. Even though he did improve his passing numbers and committed fewer PF his rebounds and steals dropped.
But mostly he failed to convert close shots (Let me restate he attempted a very similar number of these) and decided to showcase his non-existing 3 pt shooting prowess. He has done the opposite of what Josh Smith has accomplished this season:
Smith finally realized it’s not a very good idea for him to attempt 3 pt shots. If Ariza will learn that lesson next season and couple it with his career numbers converting shots at the rim – I suspect we will not be having this discussion again this time next season.
dberri
April 24, 2010
Palamida,
Excellent comment. We should make that a post of its own.
ilikeflowers
April 24, 2010
It sounds like Ariza was playing the ‘Mid Range’ SF in LA and in Houston he changed roles/positions to the ‘Full Range’ SF. Whereas Josh Smith when from the ‘Full Range’ PF to the typical PF role/position. If Ariza had continued to play the ‘Mid Range’ SF in Houston I imagine that his per minute production would be very similar to his LA per minute production. From this viewpoint he’s essentially playing out of position in Houston. Heck it might be possible to come up with more granular positional assignments just by looking at shot selection. Of course these more granular positional assignments would only be useful if they increased per minute production measurement stability year to year (assuming that a player maintains their position).
RAM
April 24, 2010
Regarding Ariza’s performance: “prediction is very difficult, especially about the future.” (Niels Bohr)
Alvy
April 24, 2010
NBA predictions are the epitome of ceteris paribus.
Tindall
April 24, 2010
palamida:
1. The Knicks drafted Ariza, not Orlando. His USG% in his rookie year was 17.8, not 19. FYI Hoopdata only goes back to 2007.
2. You are discounting Ariza’s 23 game playoff run with the Lakers, during which he took over 4 threes a game and shot over 40%.
2. Upon acquiring Kevin Martin, Ariza shouldered less of the load on offense (19.1 usg%) and played fewer minutes (33 mpg, down from 38mpg). Consequently, his scoring efficiency increased (44% FG, 41% 3pt%, up from 38% FG%, 30% 3pt%) , as did his steals and rebounding (offensive boards, in particular).
Ariza does not disprove WoW, but he is a unique case that embodies several of the arguments against it (increasing usage in certain players will decrease player efficiency, increasing minutes leads to decreasing production in certain players, shots do not materialize out of thin air, a player focusing his energy more on offense will perform more poorly defensively , etc.).
In most players, production remains consistent across teams and minutes played because coaches, aware of their players’ strengths and weaknesses, assign players suitable roles. Since coaching quality is relatively consistent, player’s do not assume drastically different roles from team to team. Ariza’s case is different. He joined a team without a defined ‘star’ and Morey paid him a fat contract (by NBA standards – WoW called for Ariza to get a pay increase). Rather than play his usual role, circumstance thrust Ariza into a role not suited to his skill set, and he consequently struggled.
To agree with me, you have to agree that shot creation requires skill. A shot does not appear out of nowhere; Andris Biedrins is incapable of taking 25 shots per game and shooting over 60% with his current offensive skill set. The same goes for Ariza. Other aspects of his season that appear to argue against WoW also do not conclusively break this model, but they point out the need to watch the game and take note of how coach’s utilize their players.
Why did Ariza have a large drop off this season? The answer is more complicated than “because he shot the ball poorly, didn’t rebound as well, grabbed fewer steals, etc.” And I think it’s more complicated than pure chance. But it could be just that. How about an analysis of Houston’s season, dberri?
marparker
April 25, 2010
Palamida just pwned alot of folks.
Thats not to say that WoW method is perfect. Its to say STFU to a bunch of folks who like to take anecdotal evidence as proof of a preconceived notion.
So those saying we told you so weren’t really saying much at all.
My personal hypothesis is that wp48s become less reliable in a couple of situations.
1. Top 3 players on teams where negative players get significant minutes.
2. Players not in top 3 on championship level teams.
I’m waiting for someone to do a study on year to year consistency and which players are the most inconsistent.
todd2
April 25, 2010
Something the prof covered in his book; maybe Trevor misses Phil Jackson.
Rockets fan
April 25, 2010
Ariza was also hurt for part of the year. I think next year, when he’s healthy and back into a more of a number 3 or 4 scorer role, he’ll put up good numbers.
Also, in regards to Brooks, he seemed to improve over the year. His shooting percentage improved over the course o the year and his assists seemed to as well. I think his numbers will continue to improve next year. He’s still young. Also, he played a fair amount of shooting guard this year. Probably a fifth of his minutes came with Lowry on the floor. When that happened, he was the shooting guard and Lowry ran the point.
I’m surprised Brook’s WS wasn’t better. He’s an efficient scorer. He takes a ton of threes and makes them at a good percentage. He also is a good free throw shooter. I suspect what hurt him is his bad turnover to assist ratio and his poor rebounding. The rebounding I don’t really care about (I understand there’s a debate about this . . .) and I believe his a-to-t ratio improved over the course of the year.
Italian Stallion
April 25, 2010
Dberri,
>In other words, if you think your argument explains Ariza, then you have to explain why you are unable to explain what happened to everyone else on the roster.<
I have tried to explain this, but it falls on deaf ears because it's common sense without the numbers.
I'll try to summarize it as simply as possible.
1. Each player has a specific individual offensive skill set. The more ways he can score efficiently, the higher his usage can be and retain a good efficiency.
2. Skill sets are not static. Every summer many players are trying to improve their skill set so they can score more, remain as efficient, and make more money (and win).
3. At any given time a player might be properly utilized, under utilized, or over utilized.
The reasons for improper utilization:
a. If a team is loaded with great scoring options there could easily be a player on the team capable of much higher usage if given the opportunity.
b. If a team lacks in efficient scoring, a individual player without a wide range of offensive skills might be asked to do a lot more than he can do efficiently. He would improve his efficiency if given a more limited role elsewhere.
c. Some players take a lot of poor shots. If they adjust they can improve their efficiency.
These are the exceptions, but there are multiple examples of them I could note on the Knicks alone. (like David Lee developing a mid range game last summer and successfully upping his usage this year. Wilson Chandler reducing his 3 point shooting mid season and dramatically increasing his efficiency etc…)
4. Most players have some ability to adjust their usage a little without much if any impact by running more plays for them etc….
5. If one specific high usage scorer is traded/injured etc… the upped usage elsewhere tends to be divided among 5 players. Since most have some ability to adjust and there may even be one with a lot of ability to adjust, the variance is often hardly noticeable statistically or can easily will attributed to other things.
Conclusion:
When you look at these things in aggregate you will see a mix of a lot of players with some ability to adjust, a few with a lot of ability to adjust, many relatively young players that are improving their skill set from one season to the next, and other small changes in usage among several players on the same team.
That why you conclude there is very little relationship between usage and efficiency.
The fact of the matter is that unless a player's skill set changes a lot, they are rarely asked to make huge changes in their usage. So you hardly ever see it.
However, within this giant mish mosh, an astute observer CAN identify individual players that are being under or over utilized, taking bad shots, about to be asked to do more than they are capable of etc… and predict a change in their efficiency beforehand if their usage is about to change.
That was the case with Ariza.
Notice, no one came here claiming that any of the other Rockets could not adjust. It was only Ariza.
That's because when he was on LA it was clear that Ariza had a limited offensive skill set and had the benefit of a tail wind of wide open shots created by Kobe and Gasol.
When he was on LA, Ariza appeared limited enough that defenses preferred to drop off him and double Kobe and Gasol. That gave him some extra wide open shots and the ability to get to the hoop a little better at times.
When he was sent to Houston, it was clear that he was likely to be pushed into a much bigger role on the offense and would also be covered better by defenses.
It was hard to imagine a way he could both increase his shots significantly, face better defense and also retain his efficiency UNLESS he improved as a player.
It was not impossible for him improve enough because his outside shooting had been improving for awhile. He actually made teams pay dearly for dropping off him in the playoffs last year (which is what made him appear attractive to begin with) But it wasn't likely he could sustain that.
Italian Stallion
April 25, 2010
I should have read Tindall’s response before I commented because he said as well as it can be said.
Italian Stallion
April 25, 2010
palamida,
You had a lot to say. I don’t have time to address it all. But let me address some of it quickly.
You seem to be arguing that Ariza wasn’t as efficient primarily because he took more 3 pointers and didn’t convert as often inside.
Well that’s exactly what some of us argued would happen!
Some of us said that to increase his usage he would have to increase the number of shots he takes that he’s not particularly good at because he doesn’t have a huge skill set. In this case it was 3 pointers, but it could have been others. That fact that he took more 3 pointers even though he’s not good at them helped prove the point.
It also doesn’t shock me that he had similar results from beyond the arc again this year (below average). He’s been improving his 3 point shot all along, but teams are going continue to allow him to have some of those until he starts knocking them down at higher rates (perhaps in the future if he keeps working on it). It’s a good thing for the defense if Ariza is taking a lot of 3s.
It was also argued he would face better overall defense (causing a further deterioration in his FG%).
That he missed more of the close in shots is probably a matter of him being a bigger focus of the defense in the spots where he is actually effective and not having the Kobe/Gasol benefit.
The only aspect of his decline in efficiency that was random was his FT%, but I doubt that had a huge impact.
Ariza is still young enough to improve his outside shot. With the addition of Kevin Martin to shoulder some of the offensive load, he should easily be able to return to his former efficiency next year by reducing his number of shots (bad ones).
I haven’t checked the stats, but he may have already improved as soon as Martin came to the team.
Italian Stallion
April 25, 2010
Sorry, but I feel compelled to say one more thing so I am clear.
Nothing I am saying (if true) says anything negative about WoW’s ability to estimate the productively and number of wins a player has contributed in the past or is likely to contribute going forward. Wow certainly captures the value of high usage efficient scoring relative to low usage efficient scoring.
I am talking about the rare occasion that a player moves to a brand new role that is dramatically different than his past role while retaining the same skill set. It rarely happens.
dberri
April 25, 2010
IS,
Just to summarize. You have a theory of basketball. But it only applies to one player in the universe.
ilikeflowers
April 25, 2010
IS’s point is not so easily dismissed. It applies to any player who changes their position/role (in Ariza’s case his shot selection) and who converts different types of shots at different rates (skill set). In Josh Smith’s case his productivity went up and in Ariza’s it went down (presumably). Likewise with Rashard Lewis playing out of position at PF in Orlando. Of course it would be nice to know how much shooting percentages from different ranges typically fluctuate from year to year, since it’s entirely possible that that could explain Ariza’s differences as well.
RAM
April 25, 2010
@ dberri
a) Given that IS has made arguments for how his ideas apply more generally your comment comes off as rather silly.
b) It wouldn’t really be appropriate for IS to present an entire “theory of basketball” in this forum. It would be appropriate, however, for you to respond in an intelligent manner to critiques of the theory that you present.
@ marparker
Thank you for telling us to STFU. Your discourse is appreciated.
dberri
April 25, 2010
RAM,
He has not shown that his arguments applied generally. When we look at players not named Ariza on the Rockets, they do not appear to apply at all. As I said before, it does not seem to make much sense to talk about usage and Ariza and then turn around an ignore the lack of a link between usage an overall performance for every other player on the Rockets.
marparker
April 25, 2010
RAM,
Come on dude, if I wrote that Palamida threw a knockout punch to a bunch of folks would you be claiming that I said I was going to knock you out. I only meant to point out that Palamida made several excellent points directly in contrast to points made on the other side of the arguement. I apologize that you took offense, my words weren’t meant to be confrontational.
Tim
April 25, 2010
It’s really great to see a guy like Daryl Morey, who believes so deeply in wp48, build a successful team around this model…
What?
He doesn’t??
Nevermind.
marparker
April 25, 2010
I gotta bite I can’t help myself
Successful teams that have won exactly 1 playoff series in 3 season?
Why are you bringing up Darryl Morey at all?
RAM
April 25, 2010
@ dberri
I think you’re misusing the word “apply”.
I also thing you are misstating IS’s arguments.
@ marparker
Fair enough. I’m just a bitter old man because my Lakers are getting spanked by babies.
ilikeflowers
April 25, 2010
There are two types of usage being discussed here. There is the usage of just shooting more and there is the usage of how you are being used strategically (shot types taken, position, etc…). On average it has been demonstrated that increasing a player’s minutes doesn’t change their wp48 much. But as IS notes it isn’t unreasonable to assume that most players fulfill the same position/role/shot selection from year to year which COULD account for much of the stability. I’m not aware of any analysis regarding how changing a player’s shot selection or position (both strategic issues really) does or does not change that players wp48. If Dwight Howard tries to play the point though, I imagine that his wp48 will go down and as we’ve seen with Josh Smith changing his shot selection/role (and I would argue position) has increased his wp48. Ultimately however this is an empirical issue. One can look at the shot selection of all the players in the league for a season and then compare that to their shot selection during the next season. For those whose shot selection has stayed similar one would expect their wp48 to be similar regardless of the number of shots (shot VOLUME usage) – subject of course to sample size. The interesting part would be to look at those players whose shot SELECTION usage has changed significantly and see how that affects their wp48.
As a specific test case one could look at the Rockets and note which players have maintained their historic shot selection, which one’s haven’t, and which one’s have seen a significant change in their wp48.
palamida
April 25, 2010
So much to say so little time :p
The truth is Mr Parker captured my motivation in writing that comment precisely: “That’s not to say that WoW method is perfect. Its to say STFU to a bunch of folks who like to take anecdotal evidence as proof of a preconceived notion”
The Stfu part… not so much, the anecdotal evidence as proof of a preconceived notion – that’s right on the money.
Moving on, I can’t possibly address all of the issues brought up in the comments but i’ll try to address a couple of them.
Tindall, your’e 100% right about that whole rookie season business that’s just a technical error on my end – I stand corrected.
IS, here’s my problem: If Ariza would take less 3’s next season and return to his Career Norms at converting shots “at the rim” – you would simply argue that he “improved” and Tindall would probably argue that this improvement is the result of playing a complete season alongside K. Martin.
How can anyone counter that or try to disprove it? that’s the embodiment of a preconceived notion.
More to the point I argue that the whole “role player” label is absurd to begin with:
The only “role” that exists is in every player’s “role” to outproduce his opponents.
Maybe we should call Aaron Brooks a “role player” since the only things he appears to be good at is shoot FT’s at a high % and Passing the ball, But no… When people say “Role Player” they usually mean a guy who can’t “create his own shot” but brings “energy” and does all the little things that fail to “show up on the boxscore”.
That’s ludicrous. Every player has a different skill set and it’s the Coach’s job to ensure that a player uses his skill set to his TEAM’s advantage.
Every min. Brooks spends on the court, like every other player – he’s expected to maintain a certain level of production, be it rebounds, Steals, Points or what have u. Brooks doesn’t do that. To the naked eye he’s doing nothing wrong, few people see in every possession the missed “rebound opportunity” or “steal opportunity” for example, but those are very real. Brooks just runs around, doing mostly nothing to help his team and every once in a while he dishes an assist or takes a shot (as said in the original post – he doesn’t appear to be all that gifted in the Latter Dpt. either).
Where am I going with this?
Usage, if it even holds any meaning has very little of it. Sure, biedrins can’t take 25 shots per game and maintain his FG% but that’s because he takes the shots he can make. The thing is, those shots apparently come often enough (among other things) making Biedrins a player who contributes considerably to winning.
Kobe’s usage this season for example was 32.2. His TS% was 54.5 – almost exactly the league average. Kobe’s scoring didn’t really change much. With all that “usage” Kobe is basically putting up average scoring in high volume.
What good does that do? What’s so great about it? What’s so great about the fact that Kobe can have one of the highest Usg rates in the league and can “maintain” it when he’s simply being average? Beats me. I guess I should ask Hollinger about that last one :)
Your’e arguing that that’s precisely your point:
If Ariza shouldered less of the offense he would undoubtably be able to make more of the shots he did take, not to mention having more energy to play at the other end of the court.
we’ll get to that in a sec.
That playoff run you accuse me of neglecting to mention (40%) was exactly the instigating cause I was referring to! Ariza hit an abnormally (for him) high amount of 3’s which in turn caused him to score more PTS, get more PT and then – Score some more.
In Houston with Mcgrady and Yao out (and His Newly signed hefty contract) he was expected by everyone (including himself) to live up to the hype and in today’s NBA that means…. Taking shots. Not converting them, just taking them. I argue that this has nothing to do with “Usage”. Ariza was and always been a poor jump shooter. Naturally if he’ll take more shots he’ll hurt his team! This is pure speculation but think about Ariza the “role player” off the bench in L.A wanting to take certain shots (off the dribble 3’s, anyone) because hey… everyone knows that’s how u get paid, but of fear of being benched and probably following the coaching staff’s instructions -*go in there, bring energy, Hustle Hustle Hustle that’s what we want from u*, elected not to take those shots. But he’s a starter now, right? a star. A contributing part to a championship team. He probably felt it’ll be “tolerated” or even encouraged now that he’ll do what he always wanted – and I suspect by now you can all guess what that is – Taking more shots.
Shouldering more offense has nothing to do with it. just poor shot selection on his part and poor decision making by the staff not realizing the harm in it and\or nipping it in the bud.
You can’t seriously tell me that devoid of a “star” player, that was the best strategy this Houston team, this unique “Role player” team had to offer. You can’t seriously argue that it had been any worse if Ariza simply didn’t take those shots- Ariza didn’t “suck” because his usage increased and he was merely a “role player” mistakenly thought to be something “else” and placed in a position to be and do “more”.
That’s absurd. All people ever mean by that is scoring. What is it with scoring? Basketball is a game of possessions: You get the ball You try to score until You either do, or you lose possession. Without the ball – no one can score, so why is scoring more significant than getting the ball? you clearly can’t have one without the other.
Let’s never use that silly term again: Ariza is limited offensively? fine, but Aaron Brooks is much more limited. Why is there a name only for being limited offensively? that’s just silly.
Ariza took shots he (if judging by past data) can’t make – how is that a usage issue? When he took those types of shots (assisted by the miraclous kobe\Gasol spacing) he didn’t make them either. This isn’t about more shots it’s about Ariza being the “victim” of his own “success”; not in the sense he now was asked against his will to “shoulder more of the offense” and he just didn’t have the skill set for it. but because he elected to take bad shots and no one instructed him otherwise.
I don’t think it’s a secret that Ariza isn’t the best offensive player in the history of the league. All this “Shouldering” business however can be accomplished in a variety of ways – Ariza and his coaches simply didn’t pick a good one.
Prof. Berry talked about the usage of other players in Houston. What about Artest?
He’s now getting all the Gasol and Kobe spacing.
In addition his usage dropped from 24.5 to 16.2.
That’s a considerably more significant change. and surprise surprise:
His Ts% in 09 was 51.3.
He increased it alright as a Laker… All the way up to 51.4%. Way to go Mr. Usage.
One can speculate all one wants but at the end of the day not one of the naysayers have provided a viable alternate explanation\model. That is one that can be assessed empirically. All we have is speculative talk of things that can never be proved or disproved, all the while using anecdotal evidence to take pot-shots at Berri’s work. At least have the decency to address my question: If Ariza’s move to the Rockets and his increased usage is indeed the cause of his reduced efficiency, why hasn’t Artest (who was basically swapped with Ariza to fulfill that role), benefited from his considerably decreased usage?
Anecdotal… I know but I guess we’ll have to fight fire with fire.
Good day Folks.
ilikeflowers
April 25, 2010
I just wanted to correct something that I posted:
“For those whose shot selection has stayed similar one would expect their wp48 to be similar regardless of the number of shots (shot VOLUME usage).”
It is of course more complicated than that since an efficient scorer who takes more shots will see their wp48 increase and an inefficient scorer will see their wp48 decrease. So you really need to look at PPS and wp48 to see what’s going on.
Tindall
April 25, 2010
palamida – I think we’re essentially in agreement over several things, mainly that circumstance thrust Ariza into a role for which he was not suited. As I demonstrated, once he began taking less shots and playing slightly fewer minutes, his production across the box score increased dramatically. In short, a bad jump shooter with limited ability to create off the dribble should not be a primary offensive option. Giving a guy like Ariza the green light to gun has consequences.
If I appeared to be arguing that Houston had no option but to give the ball to Ariza and get out of the way, I did not communicate myself very well. I tried to explain my view of the psychology involved, of which we’re both in agreement. However, I did expect Ariza’s production to drastically be diminished as a result of his role, and it did. You mention that taking more 3s is not an indicator of usage. I disagree. What about his increased assists per minute? He had the ball in his hands a lot this season. If you didn’t see that with your own eyes, you should be able to see it in the box score.
“If Ariza’s move to the Rockets and his increased usage is indeed the cause of his reduced efficiency, why hasn’t Artest (who was basically swapped with Ariza to fulfill that role), benefited from his considerably decreased usage?”
Because that’s not how usage works. You need to consider the context of the situation, what the player’s skill set is, how the player is being utilized, etc. If that’s not a tidy enough answer for you, then I’m sorry – go watch baseball.
For the record, Artest’s production does not surprise me. It probably surprises some Lakers fans who thought he would increase his efficiency in a lesser role, but I didn’t see that happening…
Since you want someone to offer up excuses for his play, I’ll give you a few typical Laker fans excuses: he isn’t a spot up jump shooter, he’s getting old, he doesn’t gel with the Lakers offense, etc. I think there’s definitely some truth to those excuses, but the simple fact is he isn’t very good, and he doesn’t play the type of game that benefits from diminished usage; a large percentage of his shots will be stupid, no matter how many he’s taking.
“Brooks spends on the court, like every other player – he’s expected to maintain a certain level of production, be it rebounds, Steals, Points or what have u. Brooks doesn’t do that. To the naked eye he’s doing nothing wrong, few people see in every possession the missed”
Not to my naked eye. Brooks is not very good, by my eye or by most advances stats. Lowery is better in most circumstances.
I’m not taking potshots at dberri’s work, but I do want him to address criticism of the model and admit its shortcomings in a more open way. Several years ago, the apbr community created several lengthy critiques of the model. dberri has yet to respond substantively, perhaps because the complaints are not worth addressing or perhaps because he has nothing to gain in admitting the model’s flaws. Both sides, those employed by NBA teams and statistics organizations and dberri have something to lose in the debate, so I guess it makes sense that both sides are eager to criticize the other… but I would really like to see the hostility, arrogance, and pettiness dropped in favour of a more open-minded approach.
As someone who does not come from a statistics background, all of this information is ridiculously confusing and difficult to digest, and the personalities, egos, and vested interests from all sides makes it nearly impossible to sort out who is right, who is wrong, and who is a little bit of both.
Thus, I would like dberri to definitively refute his detractors (that PER and efficiency are actually better predictive of team wins than WoW prior to the player adjustment kinda blew my mind when I read it years ago – that’s just one of the many criticisms of the model). I would like him to discuss cases like Ariza that inexplicably defy the model (it is quite unusual for a perennial .200+ player to drop below .100, correct?) rather than just the anecdotal cases that affirm it (Allen Iverson). And I would like him to approach analysis with a bit more humility. And so on.
So yeah, I think he’s doing good work, attempting to unravel the mysteries of basketball, but I don’t think he’s discovered a panacea just yet (or maybe he has – please destroy your critics and prove me wrong so I can use this wonderful little formula as a basis for nearly all of my basketball opinions), so I’d like to see a more even-handed approach. not that that will happen. There are books to be sold and a gospel that won’t preach itself. But maybe it will!
Tim
April 25, 2010
Re: “If Ariza’s move to the Rockets and his increased usage is indeed the cause of his reduced efficiency, why hasn’t Artest (who was basically swapped with Ariza to fulfill that role), benefited from his considerably decreased usage?”
Really? Oh boy.
Dberri, as RAM stated, it may be in your interest to draft cogent replies to IS rather than posting dismissive notes, while leaving the heavy lifting to palamida et al.
marparker
April 25, 2010
If I were the professor I wouldn’t waste my time responding either. His biggest detractors are those who won’t be convinced. Those who ask for Berri to be more open are essentially asking him to turn himself in to the proper authorities to be judged by a jury consisting of none of his peers.
dberri
April 25, 2010
Tindall,
The critiques you noted were supposed to be published in an academic article. That never happened. If you read — and understood– the article where the critiques were presented, you would see why these were not published.
Although the criticisms were never published, I did respond. You can see the response in an article I published with JC Bradbury. And also in Stumbling on Wins. So I suggest you go buy a copy of the book and look at a few of the end notes. You might discover that much of what the on-line stats community was telling you over the years didn’t stand up very well to scrutiny.
Alvin
April 25, 2010
In response to why Artest has not seen an increase in efficiency with his decreased usage, I ask you to actually read what IS and others wrote. No where in what they wrote did they claim there was a relationship that can be applied to ALL players. While there may be a relationship (as I believe Dberri has claimed there was, stating that increased usage leads to increased efficiency, correct me if I am wrong), they say that does not mean that relationship applies to every player. In the case of Trevor Ariza, they believed that given his skill set, with a dramatic increase in usage (and a different role), his efficiency would not hold up (despite what the relationship says). In the case of Ron Artest, they never made a remark, so I don’t know what they would think about what would happen to him with decreased usage (although if I were to guess, I would imagine IS and others would have predicted that his efficiency would not have increased much based on his skill set)
I know this is not going to satisfy Dberri and others because it is not based around a model. But does the model (or relationship that Dberri notes to), really fit everybody? I highly doubt it. Until more work is done regarding usage and efficiency, the best way to judge may be to individually look at skill sets, keeping in mind the relationship that Dberri notes, and look at context and take it from there.
John Giagnorio
April 25, 2010
I believe Prof. Berri has stated that Wins Produced explains about 93% of the variation in wins, so it clearly does not “fit everybody.” With that said, I’m not sure why the Ariza usage arguments are so fascinating; all of the proponents seem to admit it applies to a very limited pool of players and haven’t provided any useful tool for identifying players who fall into this category. For now I’m most inclined to believe Ariza’s poor performance is a combination of variance and Houston punting on the season (which would somewhat explain the bad shots) but who really cares?
Alvin
April 25, 2010
Ariza is a fascinating case because he is one of the rare players who has jumped from being a fourth scoring option to a first or second, seen in his dramatic increase in shots. Most players (at least according to my memory, not according to statistics), when they increase their usage increase it fairly gradually, especially if they have already been playing in the league for 5 years.
Also, this is a hypothesis (and I don’t know how to test it, maybe somebody else does), but my guess would be that players with low free-throw rates would be the type of player that would see a decrease in efficiency with an increase in usage. These types of players, I believe, usually are the ones who wold have trouble creating more shots that are capable of making.
Another note, I think this only would for a substantial increase in usage. I do not know what a substantial increase in usage would be, but again, somebody who is more statistically inclined may be able to look into this. This won’t have much of an effect if the player really doesn’t change what he does much (as a slight increase in usage would imply) Now if somebody wants to go test this (haha), maybe this debate can go farther
Tindall
April 26, 2010
dberri – I will take a look at your new book in the coming weeks.
I don’t remember your response to the criticisms being wholly convincing, though. You chose to respond to only their weak criticisms or mis-represented their criticisms at times. In fairness, I think they were doing the same thing.
I was not introduced to the online stats community until I discovered your blog in 2006, so my natural stance was to view their critiques with skepticism. Over the years, though, I’ve become much more amenable to their way of thinking and believe the story is much more complex than you sometimes suggest.
I’ve come to accept that it is weak in at least two areas, both of which you readily admit:
“1. Wins Produced does not give credit to players for creating shots. This is generally perceived as the biggest weakness. This perception is based on two beliefs. a) it is difficult to get a shot off in the NBA and b) the more you shoot the lower your efficiency. I am not sure there is much empirical evidence behind either proposition. Plus, if you give credit for taking shots then inefficient shooters will look better. Nevertheless if you believe shooting is difficult, Wins Produced will disappoint. And connected to this point…if you believe scoring is the most difficult thing a player does in the NBA, and therefore scorers are the most valuable, Wins Produced will disappoint also.
2. How vs. Why: Wins Produced tells you how productive a player is, but it does not tell you why. Performance is impacted by age, injury, roster turnover, coaching (in a few cases), and the productivity of teammates. The latter issue is the idea of diminishing returns. The more productive your teammates, the less productive you will be. This is a real effect in the NBA, although in the aggregate it is rather small. Still, these effects are not part of Wins Produced and these issues can impact what we see in the future from a particular player.”
Once you begin incorporating more statistics into Wins Produced (some of which do not exist yet!), I think your model will be more “why” than “how” and then we will have a more complete picture. Until then, I am hesitant to draw any major conclusions from it without considering the context of each situation. To establish the context – to establish the how and why – other statistics are necessary.
ilikeflowers
April 26, 2010
I went ahead and looked at some former and current Houston Rockets over the 2008-2010 seasons. Specifically I’m looking at the per 48 minute scoring changes from last season to this season. All stats are approximate. I assumed that FTA were from fouls on 2PA.
Artest – 7 fewer shots. Shot selection and PPS unchanged.
Landry – 4 more shots. Shot selection unchanged. PPS went down some (-0.08). 2nd year player. Stats are for the Rockets and Kings.
Scola – 4 more shots. Shot selection unchanged. PPS went down some (-0.09).
Ariza – 4 more shots. Shot selection changed. Took more threes (+8%). PPS went down a lot (-0.15).
Brooks – 3 more shots. Shot selection unchanged. PPS went up some (+0.07). 2nd year player.
Hayes – 3 more shots. Shot selection unchanged. PPS went up a lot (0.28) (!).
Battier – 1 more shot. Shot selection and PPS unchanged.
Lowry – 1 more shot. Shot selection changed. Took a lot more threes (+12%) and was fouled much more often (+16% per 2PA). PPS unchanged (!).
It looks like two players changed their shot selection: Lowry and Ariza. Ariza scored less efficiently and Lowry stayed the same which is interesting because he arguably changed his role the most.
Hayes improved his scoring efficiency tremendously and he also had the largest positive percentage change in minutes played. Which led to the other is a chicken and egg kind of thing. Any Rockets fans have any idea what happened?
Brooks improved his scoring efficiency some which is probably not unusual with a second year player.
Landry’s efficiency went down some owing to his time with the Kings after the trade. His efficiency was unchanged with the Rockets though and he was taking 5 more shots with them.
This is too small a sample to base any conclusions on and there are better ways to determine shot selection, so make of it what you will. Also this doesn’t examine wp48 changes at all. Something like this with better shot selection stats done over many seasons with all players and including wp48 might be helpful The critical issue with determining causality would still remain though. Presumably, a good coach is going to give more minutes to a player who is improving their PPS. I guess you could run the sim, assume causality one way or the other and then test that assumption against cases in which the causality is known.
palamida
April 26, 2010
Alvin, you claim Ariza is a rare case seeing as he jumped from being a 4th scoring option to a 1st or 2nd one. If usage rate is any indicator of the offensive “pecking order” please be kind enough to attest to the fact that Scola, Landry and Brooks all had higher rates than Ariza. Seems to me he was still the 4th option.
Tindall and co., lemme get this straight: Artest wasn’t affected positively (meaning no increase in efficiency followed the large decrease in his Usage rate) because… what? “he’s going to take stupid shots anyway” hmm. fascinating.
or is it….
“In response to why Artest has not seen an increase in efficiency with his decreased usage, I ask you to actually read what IS and others wrote. No where in what they wrote did they claim there was a relationship that can be applied to ALL players”
Seems to me, it only applies when it fits your preconceived notions. Can you supply any evidence?
Of player that saw a significant change in their Usage rate, how many fit your bill?
Do you use anything other than subjective judgment to decide that Ariza’s case can and should be explained by him being a “role player” thrusted into a situation not befitting his skill set?
Is there any evidence at all to support that view? If we’re just guessing here, let’s ask ourselves: Maybe Ariza plays worse on Tuesdays and the Rockets had more Tuesday games this past season? who knows…
You have not demonstrated cause and effect anymore than I just did, or rather didn’t.
“No where did anyone claim there’s a relationship that can be applied to all players”. What then? to what % of the players can it be applied? seems to me it only applies to “role players” whom Usage’s rates go slightly up, or maybe we should just call it The “Trevor Ariza Law of basketball”.
Alvin
April 26, 2010
Again, you just don’t read the whole post. I stated that I don’t have statistical evidence. I even said at this point that all it is is a judgement call. Yes, it is subjective. I stated a hypothesis and said I dont have the capabilities of proving it.
” What then? to what % of the players can it be applied?”
Once again, it is a certain skill set, that only some players have. I hypothesized players with low free throw rates, but I was basically asking for more research to be done (as I myself am not able to do it). I am not claiming to be right. But until further research is done, we can’t really say much, can we?
So what exactly was your point again?
Alvin
April 26, 2010
The fact is, before the season many people (such as IS and others) predicted that Ariza’s efficiency would decrease, knowing that his usage would increase. That is in fact what happened. Whether it was for the reasons stated (his being a role player and not capable of taking more shots and maintaining the same efficiency), we can’t say.
But the fact that we predicted this, despite the fact that according to Dberri that an increase in usage will lead to an increase in efficiency, makes IS and others believe they are onto something. Ariza was picked as somebody whose efficiency would decrease not arbitrarily, but with certain reasons. Whether or not you accept those reasons doesn’t change that. You are right in that until it is proven statistically, nothing can be claimed. But a pretty good hunch can be seen from this
palamida
April 26, 2010
Alvin, First of all, I addressed my reply to you mostly to challenge your claim that Ariza went from a 4th option to a 1st or a second. As prev said, three other players (and that’s disregarding Kevin Martin) had higher usage rates then him:
Scola’s spiked from 19.3 to 22.7 ; Brooks’s spiked from 22.9 to 25.7 ; Landry’s spiked from 18.1 to 23.8 which is actually a larger Increase in usage rate than Ariza’s.
None were effected the way Ariza Was.
That’s what I wanted you to address.
You’re correct in stating that you did not argue you had any evidence, good for you. that’s a wonderfully effective defense: “hey I never said I can back anything I say up with any form of evidence”.
that one is hard to counter.
One Final note to Tim:
Essentially you’re attacking me Ad Hominem when you say: “it may be in your interest to draft cogent replies, rather than posting dismissive notes, while leaving the heavy lifting to palamida et al”. Be advised, that quoting a comment, followed by a “Really?, Oh boy” has very little to offer on the subject at hand.
If you’re not interested in actually expressing your opposing view, why are you even here?
marparker
April 26, 2010
Is Ariza the new Kobe? Bring him up and watch the comment thread go on for days.
Tindall
April 26, 2010
“Scola’s spiked from 19.3 to 22.7 ; Brooks’s spiked from 22.9 to 25.7 ; Landry’s spiked from 18.1 to 23.8 which is actually a larger Increase in usage rate than Ariza’s.
None were effected the way Ariza Was.”
Because none of those players have Ariza’s skill set.
Prior to this season, I was a huge fan of Landry’s, even though many people considered him a garbage man at best — according to them, his efficiency would drop if he became a bigger part of a team’s offense. I disagreed because he has a skill set that allows him to score efficiently in several ways: he’s a great mid-range jump shooter, he is an amazing finisher around the basket, he can catch difficult passes in traffic, and he can take his man in isolations. I knew that if you gave him the ball more often, he would continue to be efficient. David Lee is a similar case. He’s always been an efficient offensive player, and it’s not because he’s hand-fed high percentage shots. That is why, in contrast to Ariza, I did not predict his efficiency to drop off. Earlier in this discussion you agreed with me, saying, “Sure, biedrins can’t take 25 shots per game and maintain his FG% but that’s because he takes the shots he can make” so you clearly understand that increasing usg% affects certain player’s – those with limited offensive game’s – efficiency, so where is the source of our disagreement?
Also, calculate his usg% pre- and post-all star break. You’ll note it’s much higher prior to Kevin Martin’s arrival (it’s particularly high in the first month’s of the season). Watch a couple Rockets games from the start of the season and you’ll see his role has drastically changed.
btw, I ctrl + f’d this page, and you’ve said role player more than anyone else – what exactly is a “role player theorist” that you’re talking about? How do you know I consider Ariza a role player? You don’t.
Alvin
April 26, 2010
I still don’t understand your point. It is not about comparing him to other players, it is about comparing him to his prior self. Maybe he didn’t lead the team in usage, but his usage increased a significant amount and he became a much more significant part of the offense in Houston than he had been anywhere else. I was wrong in using the words “fourth option”, but that doesn’t change the fact that his role changed.
And it is not my role to bring evidence. The fact is Berri stated that with an increase in usage comes in increase in efficiency. In the case of Ariza this did not occur. People on this forum have offered an explanation; one that is not made up after the fact, but one that was stated before the year as a prediction and is being repeated now.
It is a possible explanation for what went wrong with Ariza. How is it any different than your first post when you said “let’s examine Ariza’s scoring a little closer”, than offered your opinion on what the deal with Ariza’s scoring was this year?
dberri
April 26, 2010
I don’t recall saying that an increase in usage comes an increase in efficiency. After looking at thirty years of data, I have found that players who see an increase in shot attempts can see a very small decline in shooting efficiency. This is reported in Stumbling on Wins (which I sense many people haven’t read yet). The decline in shooting efficiency, though, is very small. A two standard deviation increase in shot attempts only causes shooting efficiency to decline about 1%.
In sum, the idea that players should be rewarded for taking more shots is not supported by the data. Shot attempts are the easiest thing on a team to replace (NBA players really love to shoot and will do this if you let them).
marparker
April 26, 2010
Dberri,
I’m no statistician so I’m not sure how to word this question. Do the distributions for your results have fat tails?
Alvin
April 26, 2010
Ok, sorry for the that, I clearly just recalled incorrectly
palamida
April 26, 2010
Tindall and other folks, When I comment I’m usually addressing several arguments from several different commentators. This nitpicking has gone too far – I never said that you or anyone else specific for that matter, sees Ariza as a “role player”. I did suggest however that the term basically means one thing but in fact when Folks use (or rather misuse) it usually has certain connotations and is just a fancy word for a non-scorer type player.
You want me to calculate his usg% pre and post the AS break. Let’s examine that particular split:
It has been suggested that the drop in Ariza’s Reb, Stl, etc. is a result of his increased offensive Usg and also that his Increased Ast number is a clear indicator and a direct result of that Usg spike. If so, Why is it , that his Reb and steals remained virtually unchanged Post the break while his assist went up from 4.5 (p48) pre berak to 6.4 post break?
Again, anecdotal but it’s purpose is to demonstrate that the model\picture you are offering simply does not coincide with the facts.
I could look at some other splits as well:
It has been said that Ariza played a portion of the season while not being 100% healthy. I have no idea when that was – Perhaps he was overly reliant on his poor Jump shooting because of that? Perhaps that was the partial cause for his decline in converting close shots?
I can support those claims as much as you can support yours. (not to be misunderstood i’m not actually claiming them). Let’s try another take, just for kicks:
Around 10-12% of his PT (depending on where the data comes from) Ariza was deployed at SG.
Those mins mostly came in 5 man units that featured Chase Budginger as a SF. Ariza clearly does not possess the mentality and skill set of an NBA Guard. Prime example of the result of such poor deployment we observed just recently with one Durant, Kevin. Remember him?
One could argue that the negative effects on his productivity from such a deployment were equally detrimental to his production this past season.
Personally, I don’t know if that’s the case, I do know that I can conjure up alternate (unsupported) explanations from here till Xmas.
If your model and views only allow you to say something meaningful about Ariza, Trevor and furthermore – we can debate the validity of your conclusions\assessment from here to next week, I don’t really see the point in having it in the first place.
As for our disagreement?
You (and others) seem to have this whole detailed view of different offensive schemes and statistics that when coupled with subjective observations lead you to predict and analyze with great confidence (where’s Khandor when you really need him, right?).
You KNEW that if Ariza would increase his Usg his efficiency would drop. you also knew that as for Landry such would not have been the case. All that preoccupation with Usage and Shot taking leads me to conclude that much like the “national media” aka the “casual NBA fan” aka the much maligned “conventional wisdom” – you simply overvalue scoring. That’s my real “beef” with you. Yes I agree that Biedrins couldn’t take 25 shots per game and retain his TS%, but the players who usually sport high usg% (incidentally those would be players that PER and WoW are usually in disagreement on) don’t usually offer much above average.
Biedrins could hoist shots all day long and most probably he’ll end up somewhere near Kobe’s avg. mark. What you argue then? that he’s a better player then he was before because he scores 20+ ppg and has a Usg rate of idk… 27%?
When Biedrins gets the ball in the post he mostly takes shots he perceives as being a high % shots and the same goes for putbacks etc. when he doesn’t see a good opening\look he looks for his teammates. For the likes of Kobe every look is a good look. Since he’s a talented player all that mix of good shots and ill-advised shots end up being converted at an average rate. Let us indulge in speculation for a minute and play a little what if game:
Throughout his career Kobe puts up a shot about every 1.8-2 mins. Biedrins (i’m disregarding this last season where he basically played injured every time he was on the court) puts up a shot every 3-3.5 mins or so. One caveat to our little imaginary experiment – The same way Kobe doesn’t just pull up from halfcourt, let’s assume Biedrins would not pull up from outside his “reasonable” range let’s call that range… 15 feet.
on every “good” shot he takes biedrins would have to pull up from somewhere and put up an additional shot,more or less. Do you honestly believe he’ll start hitting those extra shots at idk.. 30%? Maybe…
I see him hitting somewhere from 45-50% putting him exactly where Kobe has been (in terms of Efg%) throughout his career.
As long as a player can score efficiently, he (and his team\coaches) should attempt to get him the ball in those spots where is he EFFECTIVE from.
That’s common sense. Kobe isn’t really helping his team with his Average scoring he’s mostly helping himself.
One last thing – Where are the commentators supporting self-concocted models of “rebounding schemes” or “PF Schemes” – they’re none to be had.
Only the Scoring skill gets the royal treatment and for a perfectly no good reason – that’s where I think our true disagreement lies.
palamida
April 26, 2010
Alvin, just saw your last comment after I posted my latest comment and I want to address it briefly:
What I “did” is nothing like what you and\or others did. It was stated as FACT that Ariza’s decreased efficiency was a result of his increased Usage.
I toyed (in my initial comment, which you referred to) with an alternate explanation. I never suggested this actually took place, or is.
I was merely pointing out that the “Usage” explanation, is far from undeniable fact.
Not once did I claim that my “explanation” was even remotely factual, I simply offered a different outlook for those who would be interested in one.
Alvin
April 26, 2010
Nobody has stated anything as fact. Everybody, myself included, is offering their opinions. I have even stated this is just what I think. Just as you offer no facts in your long winded Kobe-Biedrins rant, just a bunch of what-ifs and hypotheticals.
While I was wrong about Dberri’s take on usage and efficiency, the original point still holds true. Given Ariza’s increase in usage, he suffered an abnormally large decrease in efficiency. This has to be explained somehow. It could just be random.
But to me, and a few other people on this board, such as IS, the most logical explanation is the fact that Ariza simply could not handle the increase in usage. He was asked to do more than he was capable, and suffered. Nobody has stated anything as fact or offered evidence to show this. The fact is, the evidence this time (Dberri noting that a 2 SD increase in usage leads to 1% decrease in efficiency) failed to account for what happened to Ariza. So an alternate explanation is needed.
While I can’t factually prove what I believe is right, all I can do is make the most logical argument I believe I can make. If you believe it, you do. If you don’t, so be it. But it must be admitted that something is going on that is causing Ariza an unusually high drop in efficiency
John Giagnorio
April 26, 2010
I’m having a hard time buying that so many people who read this blog are so accurate in predicting who will and will not be able to handle an increase in usage (or won’t benefit from a decrease). Maybe what we need before the 2010-2011 season is a big predictions thread where people can share their “expertise.” It’s annoying that so many people are so quick to jump into this thread and announce “look, I was right about Ariza, Artest, …”. Make the prediction publicly before the season, and make a number of similar predictions so it’s less likely that you got lucky with that one call.
Alvin
April 26, 2010
Ariza this year is just one of the few cases where multiple people have stated their opinion. Rarely do players that play the same position, but are used in completely different manners on offense, just switch teams as Ariza and Artest did. This offered a very unique and interesting case study allowing people to have more clear opinions.
As Dberri noted, a 2 SD increase in shot attempts only leads to a 1% decrease in efficiency. So most of the time, a player’s usage won’t have much of an effect on efficiency. Ariza was a rare case that people felt strongly enough to voice their opinions on it. I know that personally if you were to ask me about the efficiency of most players next year, I would say it would stay roughly the same just because Berri notes that player performance is relatively stable from year-to-year. It would take a unique case such as Ariza for me to attempt to confidently predict something that the numbers don’t show
Gabe
April 26, 2010
Two links:
and,
http://www.basketball-reference.com/players/a/arizatr01.html
First is Ty of courtsideanalyst doing an analysis of Ariza and Artest’s past seasons. Second is Ariza’s career numbers.
Looking at Ariza’s seasons with Orlando. In 05-06 he played 21 games for Orlando and scored even less efficiently than this past season (eFG% of .400%), and had a usage % of 19.3%. The following season he played 57 games with Orlando, scored super efficiently (.539 eFG%), and his usage % was 19.2%… Identical usages, huge difference in efficiency, so it seems unlikely that a “role” change could account for the huge difference.
It’s pretty clear from Ariza’s career history that the only statistically confirmable assertion about him is that he is an unusually volatile player.
Gabe
April 26, 2010
my comment with links is awaiting moderation, but I’ll go ahead and add a “small sample size” rider to it.
Michael
April 26, 2010
Good call about Brooks Professor.
Can’t wait to see the end of season numbers.
ilikeflowers
April 26, 2010
Gabe,
21 games is pretty small sample. No surprise at the volatility there. Also before going to LA he almost never shot threes so he was likely playing a different offensive role.
Gabe
April 26, 2010
His 3 point % this past season was the highest of his career.
Ty looked at Ariza’s shot distribution and assisted stats, and his assisted % on close jump shots fell quite a bit, which suggests that he may have taken too many 1v1 pull-up jump shots, some of the lowest % shots in basketball.
I don’t think that makes the “role-dependent” argument, as those are shots that you don’t want anyone to take, regardless of what their “role” is.
simon
April 26, 2010
Wow, a big thread! I was just enjoying reading the whole discussion, but Tindall brought a blast from the past that should be commented on.
Tindall seems to think, like Prof. Rosenbaum has claimed, that WP has its predictive power only because of team adjustment and other measures (PER, eff,etc) do just as well or better when you add adjustment to the model. But that seems to be wrong. Once you add another (unexplained) term to a model, that’s no longer the “model” and it’s just a giant hodge-podge of blackbox that’s there just to prop up the explanatory power. You can read more about it here in the epic thread https://dberri.wordpress.com/2007/11/01/what-the-box-score-data-says-about-shane-battier/ An anonymous scholar makes an appearance!
ilikeflowers
April 26, 2010
Gabe,
I don’t know if you were responding to me or not, but just to put some context on his career high 3p%. His 3p% with the Rockets is 33.4 while his percentage with LA was 32. He increased his 3p% by 4% moving from one team to the next. I would call this essentially unchanged. I don’t know what a typical player’s year to year 3p% variation is but it’s probably close to or greater than 4%.
And if a player is taking different kinds of shots than they have in the past it’s not unreasonable to assume that they are fulfilling a different role, whether or not it is a role that you want them to play or not.
The professor has empirically addressed how shot volume impacts PPS and wp48. Likewise how shot selection impacts PPS and wp48 is an empirical issue, just one which hasn’t been addressed yet to my knowledge. When/if it is don’t it may or may not explain Ariza’s (and others’) scoring efficiency changes.
ilikeflowers
April 26, 2010
C’mon let’s keep this thread going! I need something to do while I program…
Wade and Ginobili are WAY better shooting guards than Kobe!
Kobe is better than Jordan.
There are lies, damn lies, and statistics!
These numbers are #@$#*% STUPID. How dare you disrespect Mr. Joe Dumars!!!
Gabe
April 26, 2010
ilikeflowers,
I was more just putting it out into the universe, but you’re right that it’s not too huge difference in 3 pt %, though it came from over 200 more 3 pt attempts.
I guess my issue with the “different roles” argument with Ariza is that the main difference in his scoring efficiency comes from having shot a worse % on close shots, even though they made up a smaller % of his overall fga’s (and he shot jump shots at a similar efficiency to his previous season despite taking more). So, a huge increase in long range jump shots, with an improved %, and a lower proportion of close shots, with a worse %, IMO, seems contradictory to the “different roles” argument.
random Pistons fan note- we’ve been calling Joe Dumars, “JOD” at detroitbadboys for a while, which still amuses me for some reason.
dberri
April 26, 2010
Random question for fans of the Pistons… when did the average fan of the Pistons surrender this season. I gave up last July.
Gabe
April 26, 2010
Probably right around Nov 4th, 2008
http://sports.espn.go.com/nba/news/story?id=3679931
or, that was me. The average fan held out for quite a while.
dberri
April 26, 2010
Okay, you win. I had hope that Dumars would do better in free agency. So I held out until he made his choices in July.
ilikeflowers
April 26, 2010
How DARE you disrespect MR. Joe Dumars!
John Giagnorio
April 26, 2010
As long as we’re going off on tangents (and since Simon reminded me :), I can’t stand Dan T. Rosenbaum. I do hope that he returns to the comment section with another clever fake name though.
Italian Stallion
April 26, 2010
Dberri,
>IS, just to summarize. You have a theory of basketball. But it only applies to one player in the universe.<
That unfair and not true.
I have a theory of basketball that is not far from yours but that also accounts for many of the exceptions that fall through the cracks.
I am a fan.
No one pays me to watch every single NBA game and chart the shot selection and defensive adjustments of every single player so I can better evaluate how variances in usage might impact the efficiency of each player.
However, it is clear to me that someone with the resources to do so could make better predictions on these things because on the occasions that my causal observation reveals such insights, I am invariably right.
John Giagnorio
April 26, 2010
“I am invariably right.”
What else there to argue, really?
Italian Stallion
April 26, 2010
Palamida
>>S, here’s my problem: If Ariza would take less 3′s next season and return to his Career Norms at converting shots “at the rim” – you would simply argue that he “improved” and Tindall would probably argue that this improvement is the result of playing a complete season alongside K. Martin.<<
I think you are trying to make the game simpler than it actually is.
The box score tells you what happened, but it doesn't always tell you why it happened. That's why some of us advocate using the box score as a one tool and observation and perhaps more detailed stats as another.
If Ariza stopped shooting as many 3s, I would argue that either he or the team realized he was inefficient at them and they tried to substitute more efficient alternatives. That's a type of usage change.
If he started converting more inside shots, I would have to evaluate how defenses were playing him.
If they were playing him the same way I would argue he probably got better as a player +/- some randomness that could be part of it.
If he was getting open a lot more because defenses were more preoccupied by new players on the team (like Martin or Yao etc..) , I would argue the conditions for his success improved.
Italian Stallion
April 26, 2010
>>“I am invariably right.”
What else there to argue, really?<<
LMAO.
I didn't mean for that statement to come off as arrogant as it sounds.
What I was trying to say is that I have very few strong opinions on these matters, but sometimes the evidence is so strong even a casual observer like me can notice.
When I do see something like that it seems I am invariably right.
Obviously, I think someone with greater dedication, resources and skills could do better than me.
Italian Stallion
April 26, 2010
This thread is so long and thoughtful, instead of responding to everyone that may have addressed something I said, I want to clarify one key point.
IMO there are at least two types of increased usage.
1. You can increase the number of shots you take that you are already efficient at.
2. You can expand your shot selection to include shots that you normally don’t take because you aren’t any good at them.
In the first case, I think there is some ability to increase usage with little or no impact by running more plays for a player.
An example of that might be Phil Jackson encouraging Kobe Bryant to get the ball to Bynum and Gasol more often because they are more efficient inside than he is at taking tough shots from the outside.
However, eventually IMO they will bump up against the limits of that kind of thing because defenses will adjust and try to take the most efficient shots away.
To make my point using an extreme example, even Wilt Chamberlain couldn’t score 100 points every night no matter how hard he and the team tried, but when he averaged 50, 55 at similar efficiency was probably not out of the question . :)
It’s the second case where we get into the actual skill sets of the players.
I hope few would debate the following.
Some players are good around the basket, but not so good from mid range and outside.
Some players are good from the outside, but have a tough time finishing around the basket.
Some players can create off the dribble, but some cannot.
Some can get a shot off quickly, but others need time to set.
Some have the leaping ability or wingspan to get over other players and others do not.
I could go on endlessly, but the bottom line is that the more offensive skills and attributes a player possesses, the higher his usage can go while retaining satisfactory efficiency.
As usual I will use a NY Knick as an example because I know the players so well.
David Lee has always been a very good player, but his mid range and outside shot was previously erratic and limited enough that he didn’t take them very often (he’s very smart). When he did, the defense gave him space.
He worked hard on that last off season and finally got really good at it. That “allowed” him to increase his usage and retain high efficiency (his stats are available)
However, if Lee decided he wanted to up his usage a lot further he would eventually have to add some shots he’s currently not very good at because there are limitations on the shots he’s taking now (otherwise he’d score 100 a night – lol).
He might have to stretch the range out an extra few feet, shoot more 3 pointers etc…
The problem is he’s still not any good at those yet. So if he tried his efficiency on those shots would not be satisfactory and his overall efficiency would suffer a bit (the more of those he took the more it would suffer).
The proof that this is a reality comes from David Lee himself, as he has already said that this off season he plans on working on a 3 pointer so he can add it to his arsenal and expand his usage further.
The thing is, occasionally players get throw into situations where their usage changes sharply without a corresponding change of skills.
If the change is drastic enough to involve either an increase or decrease in the number of shots a player takes that he is not very good at, then his efficiency will change.
The reason this does not happen very often is because coaches are usually smart enough to use players within their existing skill set and not outside it.
However, on the occasions that injuries, trades, unusual circumstances, foolishness on the part of the player/coach etc… causes a player to go outside that range, an astute observer of that player can make predictions about efficiency that a more general model cannot.
conanthelibrarian
April 26, 2010
IS,
Beautiful analysis, my friend. For the record, I used to see you post over at Knickerblogger, and have seen your recent posts on APBR forums, as well as here, obviously. I’d like you to know that I’ve always thoroughly enjoyed your commentary. Your efforts to continually better understand the game are commendable, and your opinions are usually well though out and grounded. I have some questions for you that are more suitable to e-mail than dberri’s blog. If you’re interested/curious, please contact me at michaelmcnicoll @ hotmail.com.
ilikeflowers
April 26, 2010
I think that one of the main things to take away from this epic (but not Kobeesque) thread is the difference between what is happening and why it is happening. This is the classic correlation does not necessarily imply causation. The What that is happening is that on average increasing a player’s shot volume doesn’t significantly affect their scoring efficiency. The Why this is happening boils down to three possibilities that I can see:
[1] There is an intrinsic scoring aptitude that is being captured by the box score data. Having a player shoot more frequently therefore doesn’t cause a significant decrease in their scoring efficiency whether or not they change their shot selection.
[2] There is self-selection in the data in the form of coaches and/or players generally limiting a player’s increased shots to those types of shot attempts that that player is perceived to convert well.
[3] There is self-selection in the data in that players who fail to shoot as well when their shot attempts are increased will see their shot attempts reduced (by themselves or by their coaches or teammates) and thus they will not be recognized in the data set as players whose shot attempts were increased and whose scoring efficiency suffered as a result.
Each of these situations alone will result in the empirical result seen. The actual why/cause is likely made up of some proportion of each of these factors. The relative proportion will determine the strategic value of this knowledge for decision making.
If case 1 is dominant then one can just say, hey player x scores efficiently so let’s have them take more shots of any type and it’s likely that they will maintain their scoring efficiency.
If case 2 is dominant then one can and should choose to increase the shots of efficient scorers within their typical shot selection type and you might let them experiment with additional shot types to see what happens.
If case 3 is dominant (or even just significant) then we really can’t make a recommendation for an individual player based upon the general trend because the players who are declining in scoring efficiency as a result of increased shot frequency can’t be reliably identified in the data.
Hopefully, since scoring efficiency often isn’t properly valued in the NBA when it comes to who gets to take the most shots, the reality is that case 3 is not very significant.
As long as 3 isn’t significant then empirically looking at shot selection’s effect upon scoring efficiency coupled with the demonstrated general stability of scoring efficiency as shot attempts are increased will lead to a sound strategy of when, how, and if a player’s scoring load should be increased.
palamida
April 26, 2010
IS, You tell a lovely story but i’m finding it extremely hard to believe that of all the active players in the NBA, somehow poor ole’ Ariza got himself in this “mess”.
For every singular case there seems to be an explanation, a way to adjust these extremely particular models: “Player x is still young and improving”, “Player Z was underutilized to begin with”, and so on and so on.
Here’s an arbitrary (and incomplete) list of players whose Usg rates increased drastically in the past season with little or none negative effects. Some saw slighter increases compared to Ariza and some saw more significant ones – here they are in no particular order: (09 usg, 09 ts%, 10 usg, 10 ts%)
Derrick Rose – 22.6 ; 51.6 ; 27.2 ; 53.6
Ryan Anderson – 18.5 ; 53.3 ; 24.3 ; 57.4
Chris Kaman – only played 31 games last season I used his average for the past two season, both with similar Usg (higher Ts% last season)
– 21.4 ; 54.5 ; 27 ; 52.7
S jax – 25.2 ; 53 ; 27.8 ; 52
Corey Brewer – played only 15 games last season I used an avg of his past two seasons – 16; 45 ; 21.4 ; 50.3
Luke Ridnour – 18 ; 50 ; 22 ; 57
Dahntay Jones – 14.7 ; 53.3 ; 20.14 ; 52.7
Rodney Stuckey – 23 ; 50.8; 26.4 ; 48
Tyrus Thomas – Small sample in Charlote (25 games) but the larger increases in Usg are hard to come by :) used his numbers from the last season and a half – 19.5 ; 52.4 ; 25.1; 49.9
Andray Blatch – 21.9 ; 50.8 ;25.8 ; 51.9
I think that’s enough for now. I tried to include players as diversified as I could find: Young player, Veterans, Players who changed teams, those that didn’t, Different positions, starters, reserves etc.
Some of these players saw an increase in efficiency, some saw saw a slight decrease, some saw a larger decrease. Not one,even those who saw their Usg increase more drastically than Ariza showed the complete free fall he has demonstrated in efficiency – from 54.4% to 48.8%. Not even close. not one has moved from an average TS% (or higher) to below average mark, as Ariza has.
I’m oversimplifying, you say?
Maybe I am, but this seems to be quite simple:
We can speculate, conjecture and invent “stories” about “training in the summer” and “extending ranges” and lest not forget my personal favorite – “changing roles”, but the fact remains this simply does not happen very often, to say the least. Not one “side” can prove or disprove the other side’s position, IMO it’s just not plausible that a dangerous beast is roaming free in NBA courts across America. It’s name – Increased Usage. It lied quietly, biding it’s time, for years seeking the perfect prey. Finally it found poor old Trevor Ariza, ravished him and crawled back to it’s hole, never to be seen or heard from again.
Naturally, I’m exagarating, but it’s all in good fun (I hope) and I respect every single commentator who put thought and time aiming to contribute (I guess that excludes you, Tim) to this growing thread.
It’s simply not a very plausible theory.
P.S I’ve just redIlikeflowers latest comment and I have to say that the self-selection argument (which has been brought up in the past) could theoretically explain the lack of precedent. The only problem with it is that the same staff (I.E Adelman) has been around the block once or twice and I for one, Can’t say I recall any, one particular player to slip within the cracks of this “self selection process” so heinously as Ariza did.
Not under Adelman, or anyone else for that matter. It still seems to me like Occam’s razor is the cure to all our troubles: Ariza simply had an off year: Got his contract and was less motivated, injured, didn’t care for Houston’s couisine or found out his wife was cheating on him – who cares? – an aberration is what I mean.
Bottom line is, it’s very likely that a simple decision by him (or the staff) to limit his ill-advised 3’s coupled with a return to his career norms from up close would render this whole discussion moot.
P.P.S Who knew Aaron Brooks was such a popular player?
brgulker
April 27, 2010
Dr. Berri,
Gabe and I comment frequently at the same Pistons blog/fan site.
I’d say most people held on until around the All Star break. But I’m with Gabe. I had little hope after the Billups trade.
And now, the latest news from the Detroit Press is that Detroit is looking to upgrade at the PG. The irony … it’s so thick.
Italian Stallion
April 27, 2010
palamida,
Clearly you want to believe what you want to believe and I lack the time, energy, resources and mathematical skills to prove I am right.
I’ve tried to persuade you by using common sense, examples of specific players I am very familiar with, and general basketball sense but that did not work.
I totally understand your perspective.
I would guess the only way someone like me can more or less prove their case is by making predictions that go against other theories and then seeing how they turn out.
I love basketball and want to understand it better. But this is not my career. I don’t watch every game, looks at stats all day etc…
I watch every Knicks game and most games on TNT, ESPN and NBA TV. So the types of exceptions I can talk about are few and far between. They have to be extreme enough for me to notice with such casual effort. I suspect there are many others, but they are outside my range.
In the end you can read my thoughts and see how my occasional predictions turn out. Best of luck.
l
todd2
April 28, 2010
Trevor Ariza has been given (and compensated for) a role that he’s not living up to. It looks as if the Rocket’s management has blown it. Objective and subjective evaluations have been an ongoing debate on this forum. Box score data is one way to evaluate a player but there are others that are equally important, albeit, less tangible/measurable. IS and ILF have both commented on being able to identify the causes of a player’s performance. With all due respect to the prof, his work is like a snapshot of player/team performance but really isn’t diagnostic or go to great lengths to qualify. Rodman and Jordan were both .400 players but night and day. Successful teams are built by a combination of size, speed, depth and the right mix of complimentary skills. Individuals can be evaluated based on size, speed, strength, stamina, footwork, shooting mechanics, shot selection, coordination, agility and decision-making, to name a few. Many of these variables are subjective, can be determined by an experienced observer and don’t turn up in box scores.
Phil
April 28, 2010
palamida,
Excellent stuff.
A caveat behind the “increased usage can lead to diminished efficiency” concept is that it is not simply a matter of shooting more which lowers efficiency, but of shooting shots to which players are not suited. It’s not just his 3pt% that was poor in a high usage role; his FG% was also way down.
I’ll admit I’m not wholly familiar with all the players you listed. But for most of them, their role hasn’t really changed – they’re shooting the same shots as before, but just more of them.
Rose still gets most of his stuff in pick-and-roll situations and in isolation. Ditto Brooks. Ditto Kaman.
Ryan Anderson shoots almost exclusively open jumpers. I’d wager good money that the Magic are more adept at creating open jumpers for him than the woeful Nets. The same is true for Jones, going from the Pacers to the Nuggets.
Stuckey increasing his usage by 10% isn’t much of a change at all. Ditto SJax – both of whom are, again, fine creating their own shot. Especially when you’re comparing it with Ariza who had, what, about a 40% increase in usage?
I could go into more detail, but there’s times where players can or should shoot more shots, and there’s times where there can’t/shouldn’t. Ariza was a case of case of the latter, and that was pretty obvious even before he started playing for Houston.
todd2
April 28, 2010
“increased usage can lead to diminished efficiency”—great stuff! Sounds eerily similar to the law of diminishing returns.
ilikeflowers
April 28, 2010
Epic thread is still alive!
palamida,
“Bottom line is, it’s very likely that a simple decision by him (or the staff) to limit his ill-advised 3’s coupled with a return to his career norms from up close would render this whole discussion moot.”
This is exactly what I’m talking about though. One of my points is that when you significantly change your shot selection, you change your role/position. There are different types of small forwards. A SF who doesn’t take 3’s is not playing the same role/position as a SF who does take 3’s. If Ariza stops taking (and positioning himself for) so many threes what’s he doing in the meantime? He’s cutting to the basket more, posting up more, attacking the rim more, taking more midrange jumpers, passing more, in position for more rebounds, etc… I would note that Josh Smith is the poster child for this kind of role/position change.
Also we don’t really have the shot selection data for Ariza when it comes to what sort of 2pt shots he was taking this year as opposed to last. Courtside Analyst has speculated that he’s shooting more short-range jumpers as opposed to layups and dunks this year. But we just don’t know.
To expand upon positions/roles. I don’t think that anyone here would argue that for most players, a change of position (the standard 5) will cause a change in wp48 or that a change in offensive efficiency is likely. I’m of the opinion that the five positions used in the box score don’t really provide enough granularity in assigning a player’s position and that wp48 stability MIGHT be enhanced by systematically separating players into additional positions based upon shot selection.
ilikeflowers
April 28, 2010
I said,
“I don’t think that anyone here would argue that for most players, a change of position (the standard 5) will cause a change in wp48.”
Obviously since wp48 is adjusted by position changing position is likely to change wp48. I should have said change in WinScore or significant change beyond what would be expected from just a different positional adjustment.
Gabe
April 30, 2010
ilikeflowers,
I wasted about 10 minutes on http://www.nba.com/hotspots/ to try and figure out Ariza’s shot selection. Quick summary:
In 08/09, 47% of Ariza’s total FGA’s were at the basket, and he made them at .617%. Also, 21% of Ariza’s total FGA’s (123 attempts) were 2 point jump shots, and he made them at 32%.
In 09/10, 37% of Ariza’s total FGA’s were at the basket, and he made them at .531%. And, 22% of Ariza’s total FGA’s were 2 point jump shots (216 attempts), and he made them at 27%.
If Ariza had kept his 08/09 ‘at the basket’ FG%, it would have resulted in about 31 additional made baskets over the year, and would have raised his eFG% from .462% to .493%.
Stu
May 1, 2010
Here’s an interesting critique at APBRmetrics from David Lewin:
“The best way to test whether a statistical model is accurate capturing player ability is to test whether the model can predict out of sample how a lineup will perform based on the ratings of the players in that lineup and the ratings of their opponents. When tested, Wins Produced fared significantly worse than many other metrics (PER, Win Shares, etc) at out of sample predicting the results of each shift [a shift is a period of a game between substitutions, so just two five man lineups playing against each other].
This is a damning result. Basically, it seems, that Wins Produced’s high season level fit is due to overfitting and mis-attribution of credit and that when each shift of the season is considered, in order to disentangle player collinearity, Wins Produced does a poor job of predicting how teams will perform.”
I was wondering what Prof. Berri’s thoughts on this were? It does seem like wins produced, if the best metric, should be able to outperform PER and winshares in predicting the on-court performance of a 5-man unit. What is the explanation for why it didn’t do that? And doesn’t that totally invalidate wins produced as a metric? What good is it if it can’t predict how a given 5-man unit will play?
dberri
May 1, 2010
This is another example of what we saw back in 2006. Lewin likes to see how other models explain his approach. Such an approach, though, leaves one wondering what is being tested (your approach or Lewin’s approach).
The problem with plus minus data — and adjusted plus minus data — is that these measures are very volatile over time. So we are not sure what is being captured. I do not see how changing this to looking at five man units changes the story. At least, I want to see how consistent five man units are over time.
Let me add that Lewin is taking a very odd approach to evaluating a model. In evaluating a model we should at least look at the following issues
a. the theory behind the model
b. does the model explain what it purports to explain.
c. is the model consistent over time.
PERs — and models like this — are not theoretically strong. They also do not explain current wins (which seems to be what you shoudl try to explain). This is true whether you add a team adjustment or not (unless you define the team adjustment as the residual of your model, an approach that is very incorrect).
By the way, this point has been made in refereed journals.
I would also add… I do not know serious academics who argue that their critique of another person’s work is “damning” or “devastating”. Serious academics (again, at least the people I know) also do not argue their work is on the “cutting edge”. People who take this approach appear more interested in selling their work to non-academics. They do not seem interested in serious academic discussions.
stu
May 1, 2010
Okay, but it seems to me like Lewin’s approach was this – take the play by play of every game of a season, or whatever. Identify ‘shifts’, which would be stretches between subs, so everyone on the court stays the same. Look at how much one unit outscored the other unit in that shift. Do this for every shift of every game of the season. Then take metrics like PER & wins produced, and say ‘we know which ten players were on the floor at the same time…. Using this metric, what would we predict the outcome of this shift to be?’ And then see which metric most closely matched the actual outcome, thereby proving that metric to have the best predictive power.
What am I missing about the flaws in this approach? It seems really strong – you’re basically saying, if I know the 10 players on the court and their wins produced etc., can I accurately predict the outcome of that matchup? Isn’t that what basketball analysis is really all about, predicting how a lineup will do against another one? In Wow, you make the case for point differential as the building block of wins, and in that study he tried to predict point differential of every lineup matchup in every game. If Wow doesn’t do well at predicting that, I think that’s a really serious problem. Or am I still missing something?
palamida
May 1, 2010
Dberri, I have to admit I don’t truly understand your reply. Stu’s comment, btw isn’t “another example of what we saw in 2006” it’s basically quoting an old post from the APBRmetrics forum – It’s literally what “we” saw in the past.
I completely understand and accept your approach to the whole “damning” business – that’s not a legitimate approach, however I fail to see how +/- got thrown in the mix here. Your critique of the the whole PM branch is well known. While it’s true that “5 man units” usually are used in the PM context, Lewin’s “study” (has he ever actually released the data and specified his methodology,btw?) has nothing to do with PM.
Sample size can surely be a factor here, but regardless of “how consistent 5 man units are over time” and regardless of PER and winshares’s weaker explanatory power (“current”) – isn’t it reasonable to expect superior prediction “power” of the actual performance of different “shifts” using WoW over say, PER and\or Winshares?
If I recall correctly one of the main issues some of the folks over at the aforementioned forum had with WoW was the value of the Defensive Reb.
the argument (I won’t go into it at length, we all know what I’m talking about) was that while the value derived from the regression was correct when applied on the team level, the fallacy lied in attributing the “entire” credit to the “fetcher” I.E the player who controlled the DReb. The “Shifts” study is an excellent way (at least theoretically) to expose that “fallacy”, is it not?
It can (if done correctly) demonstrate that the “Current” explanatory power derives from the “team level” but that in the WoW model the credit is in fact assigned incorrectly, individually.
The whole D. Rosenbaum “Harvard paper” fiasco (residual, etc) also has nothing to do with this and yet you chose to bring that up as well.
What am I missing here?
Personally, I haven’t read an actual study examining performance across “shifts”.
But can you explain to a layman, (without bringing up suspect usefulness of PM, or other metrics for that matter) why is it wrong to expect the WoW model to “outpredict” other “Linear weights” metrics across say, a season’s worth of “shifts”?
dberri
May 1, 2010
Let me clarify my thinking…
Lewin is choosing as his dependent variable the performance of 5-man units. This appears related to the approach taken in 2006. At that time, adjusted plus minus measures were regressed on PERs, Wins Produced, etc… My problem with that is approach is that adjusted plus-minus measures are very inconsistent over time. So it is not clear that these measures are actually capturing the value of an individual player.
Now we have moved to plus-minus for five man units. Again I ask, what is being captured with this number? Are you capturing the value of the 5 man unit? If that is the case, these numbers should be consistent over time. In other words, if you are really capturing something important to decision-makers, you should get roughly the same value over time. So I spent some time downloading data on 5 man units from 82games.com. Again, just like we saw with adjusted plus-minus, these numbers are very inconsistent over time. In other words, past values for five man units don’t seem to explain future values. If that is the case, what is the value of a model that uses these as a dependent variable?
This returns me to a point I was making earlier. Why do we need to use such an odd approach to test a model? Why can’t we simply ask…
1. Is your model theoretically sound?
2. How well does your model explain wins (what a model should be trying to explain)?
3. How consistent is your model over time?
When we look at these three questions (which is not an exhaustive list of issues we consider in evaluating a model, but it is a good start), we see that PERs and models of this sort do not explain current wins. And there are good theoretical reasons for this lack of explanatory power. When we turn to adjusted plus-minus, we see very little consistency over time. So what value are these models to decision-makers?
palamida
May 1, 2010
That does clarify some issues (for me at least), but i’m still not getting the relevance of PM to this whole discussion. Let me present a simple example:
Let’s take an imaginary 5 man unit consisting of 5 career 0.100 Wp48 – that’s 0.500. Let’s suppose that unit played an X amount of minutes vs. a unit whose wp48 summation was 0.250, and Y Mins. vs. a unit whose sum was 0.750.
I would expect (assuming the sample wasn’t particularly small) that the the 0.500 unit would fare better in terms of Points scored and allowed per possession (Efficiency Differential) against the 0.250 unit, than it would when matched up with the 0.750 unit.
It’s my understanding that Lewin’s numbers suggested than when you use PER or Winshare values rather than WoW’s, you get “closer” to the actual outcome, meaning they both “predicted” the actual scoring deficit – better.
I don’t see how PM and it’s shortcomings are pertinent to this specific issue I.E the results of this “shifts” study, in any shape or form.
To clarify once more, what am I missing?
Yes, 5 man units are usually used in the context of PM, but in this particular instance the data isn’t evaluated in terms of on\off court issues.
And while i agree with you that as you argue – without consistency there’s little (if any) value to these metric as far as decision making is concerned. I don’t see how this is relevant to Lewin’s claims.
A game of basketball is merely being dissected here into smaller portions – Shifts. and the outcome of each mini-game (shift) is then recorded. Then our “expectations” of each mini-game (shift) are devised, based on the values derived from three different metrics. Then the expected value each metric yields is compared with the actual outcome of each shift. have I completely misunderstood Lewin’s methodology?
As I said, I never saw it’s details (just the conclusions) but if i’m reading that post correctly, those are in essence the steps that were taken, were they not?
stu
May 1, 2010
Thanks for clarifying. I guess I’m still confused about why the consistency of the 5-man groups matters? Wins for teams aren’t consistent from year to year either, but it seems like you’d be more comfortable predicting out-of-sample wins from wins produced than out-of sample point differential for shifts. But like palamida said, aren’t shifts just ‘mini-games’? I keep coming back to the emphasis you placed on point differential in WoW, and how it related to wins. If point differential for shifts is important to predict, then why is it more important to predict wins for seasons, when season wins are simply made up of the point differentials of a lot of shifts?
dberri
May 1, 2010
I think the analogy that these are mini-games is fine. The problem I have is twofold…
1. We already have data on entire games. Why can’t we just use that data (which frankly seems like the more obvious approach)?
2. The problem with inconsistency is also important. The dependent variable in this case does not appear to be a very reliable measure. If you are truly capturing the value of the five man unit, then it should be the same value over time. But it doesn’t appear that is the case. So then, what are you regressing on all your different measures?
Let me point out again… the box score measures are consistent over time. So they seem to be capturing the player’s value. The inconsistency of these so-called “advanced” metrics suggests that these measures are not much of an advancement.
palamida
May 1, 2010
“We already have data on entire games. Why can’t we just use that data (which frankly seems like the more obvious approach)?”
First of all we can and we do. However if we limit ourselves to “that” data alone we’re left with two issues. 1.) outside of players changing teams we have very little in the way of “out of sample” data.
If for instance the Dreb is indeed over valued (On the individual level, that is) in the WoW metric as many critics have argued, we have very little data to use when countering WoW’s premises, reasoning and\or conclusions.
The only form such criticism could be demonstrated would be through anecdotal evidence which is by definition – anecdotal.
When an “extreme” player (on either side of the Dreb scale) switches teams we can examine how the teams in question have performed before and after. Even then it’s impossible to isolate all the constants and variables making this type of anecdotal evidence prone to subjectiveness and lacking any objectiveness in the true sense of the word. That only leads us to the “either side can’t prove or disprove” their position kind of debates of which this thread is a glaring example of.
That’s one very good reason to invite other sources of out of sample data which are less anecdotal or even not all all.
2.) As for the validity of this particular dataset, you argue that it’s inconsistency renders it all but useless (can’t even explain itself, etc).
I don’t have the knowledge to address that head on, but let me ask you this – is that really the case? are 5 man units inconsistent?
The innate problem (one of them) of PM is the whole starters vs. subs issues it has.
Let’s ignore that for a second.
If we take say two specific 5 man units:
The first, the Celtics (most common) starting five: Rondo\Allen\Pierce\Garnett\Perkins. this unit played 1154 mins together this last season.
The second is Toronto’s most commonly used 5 man unit (and starting 5) :Jack\Derozan\Hedo T.\Bargnani\Bosh. Both are starting units which by in large (if not 100% of the time) were matched up against other starting lineups across the league.
I would expect The Celtics Lineup outproduced it’s opponents more than Toronto’s outproduced it’s respective opponents ,in a very consistent manner. Haven’t looked into it, but i’d be very surprised if that’s hadn’t been the case.
If THAT isn’t consistent then sure, you can flush all that data down the toilet – it’s useless.
But when you say : “If you are truly capturing the value of the five man unit, then it should be the same value over time. But it doesn’t appear that is the case”, What exactly are you referring to?
It is my strong belief that a reasonable person would have to agree that we have a conundrum of sorts:
In this thread (among others) certain commentators (IS, for example) have expressed their views of the game of basketball. Their views are hinged on a premise that states that the game of basketball is a complicated game to evaluate.
They claim that Coaches and players use a somewhat intuitive approach in assessing player’s abilities and skill sets. In their “world” NBA players are like snowflakes – no two are quite alike. In their “world” that is precisely why NBA teams employ highly payed staffs who possess an expertise (some naturally, more than others) in this field. Those “experts” make constant adjustments in play-calling, matchups or what have you. Where do I see the conundrum?
Well, if that were to be true it’s well expected that 5 man units would not be consistent in such a world and it’s ever evolving dynamics.
In their eyes the very thing that excludes that data in your eyes, is the very proof of their view. Naturally they haven’t (or even attempted to) extablished cause and effect here, not in the least bit.
They continue to argue that there is no “one true metric”. The experts should (and usually do) use a wide variety of metrics and other non-stats related methods to aid them in their thought process and decision-making.
You’re correct when you say that Box score Stats are consistent over time, but that in itself does not validate the WoW metric. “They” will argue that on the Team level everything works fine, a rebound is a rebound ,a To is a To and a point is a point – All equal to the value of a possession; But when applied indvidually, WoW miscredits the individual. In the past You argued that if the consistency was only at the team level, how can we explain the year to year in consistency even in players who switch teams? “they” have an answer for this as well: “roles”. E.G: Camby was deployed in a manner that allowed him to record a high level of Dreb%. When he switches teams (but not “roles”) which is often the case since teams that desire his services are usually interested in what he already brings to the table, he’s teammates still “defer” to him, and his numbers remain all in all – consistent. That’s a conundrum that in my eyes, can only be truly resolved by “out of sample” data.
Perhaps rejecting this particular dataset is a wise move (if it is indeed inconsistent), but arguing that “Why can’t we just use that data (which frankly seems like the more obvious approach)?”
seems inadequate to me. We should be pursuing other datasets that could demonstrate and illustrate in new ways the true nature of the game of basketball. If that “new light” reveals that the WoW metric was “right” all along, so be it, wonderful. But further examination and doubt have long been the staple of rationality. Let’s not sight of those, along the way.
dberri
May 1, 2010
When I say it is inconsistent, I am saying that it is inconsistent when I looked at the same five man units in consecutive years. When I say box score data is consistent, I am saying that at the player level (not the team level) the data is consistent.
I would add, box score data is also consistent from team-to-team. Plus-minus and adjusted plus minus data is completely inconsistent. If you look at a sample of players who switch teams, a player’s APM is not explained at all by his lagged APM. So when player’s switch teams, APM doesn’t appear to tell you much at all.
palamida
May 2, 2010
Just to be clear, I agree with you on probably 80-90% of the Major issues.
Since this old comment was brought back from the dead, I tried to represent the “naysayers’s” arguments, for the sake of this discussion and also intended at readers who perhaps aren’t all that familiar with your past responses to some of the common criticisms.
Personally I just find it a bit frustrating that when discussing these matters with rational, intelligent people (IS, Tindall, etc. as opposed to say… Khandor) all we can agree on is that we disagree.
I’d be great with convincing them that my views are just and i’d be great if I could be persuaded by them – neither seems to be happening. I don’t care much for this “Mexican stand-off” these discussions tend to end up at as of late.
Quick question and a follow up side note:
When you say inconsistent do you mean the 5 man units (year by year) were inconsistent in terms of Points scored and allowed?
On that side note and to follow up on that, to what do you attribute this inconsistency? I mean, since Box score stats are consistent (on an individual level as well) and they are highly correlated with wins, and efficiency differentials are obviously correlated with wins, what then, causes a certain 5 man unit, which remained under contract and on an active roster in the same franchise to produce different results year by year?
that seems rather odd to me…
stu
May 2, 2010
You implied that the year-to-year inconsistency of 5-man units means that the point differential of a 5-man unit is not reflecting its true ability. If that’s true, how better could you measure the ability of a 5-man unit? And aren’t games just a series of 5-man unit matchups? Teams aren’t that much more consistent game-to-game, are they? Then that must indicate game data is flawed too, that game point differential isn’t indicative of true ability. But aren’t seasons made up of games? And teams aren’t that consistent from year to year, either. So that must mean season point differential is not indicative of true ability. But isn’t point differential the foundation WoW is based on?
Do you see why this means your complaint about 5-man units being inconsistent makes no sense? You can’t pick and choose which dataset to test on. If a metric is valid, it will work on all samples, not just the ones you hand-pick. 5-man unit point differential is the basis of games, which are the basis of seasons, which are the theoretical basis of wins produced. Saying that 5-man unit performance isn’t relevant means that game performance isn’t relevant, which means that seasonal performance isn’t relevant, which means that your books aren’t relevant. That’s an awful lot of cognitive dissonance, if you don’t mind me saying.
simon
May 2, 2010
stu//
Wow, that’s one of more ludicrous usage of slippery slope logic I’ve ever seen, even on the internet.
stu
May 2, 2010
And Simon, that’s one of more ludicrous usages of hyperbole I’ve ever seen, even on the internet.
simon
May 3, 2010
I know you’re being clever stu but that really does sound ludicrious. Trying to assess the usefulness of the WP model based on its ability to predict line up just seems awkward. You make it sound like 5 man units play like hockey lines, but most of the units do not play that many minutes and their year-to-year performance consistency is rather iffy. If the WP model has failed to predict yearly player performance since it came out you’d have a much more solid ground for complaining.
But more so, based on that test, you somehow link the logic all the way to the usefulness of his book and claim it’s a “cognitive dissonance.” Sorry, but that’s really ludicrous.
stu
May 3, 2010
I don’t see why. The game is made up of matchups of 5-man units. That’s basketball. If it can’t predict how any given stretch of actual basketball will play out better than something like PER, what’s the point? I thought we were trying to model real basketball here, and 5-man unit matchups are the core of basketball. It doesn’t get any more elemental than that while still maintaining the team aspect of the game.
simon
May 3, 2010
Because the “elemental” part isn’t as clear cut as you imply. If we get set lineups where 5 players consistently play with the same players game after game, then it’ll work better but the NBA basketball isn’t played that way. It should be noted that a similar claim has been made before . The advocates of +/- insisted all you need in basketball game boils down to point scored and allowed, just like in team point differential. And thus if the WP can’t predict each player individual adjusted +/-, the WP model is flawed.
At the end, the model was designed to predict seasonal individual performance with some consistency that can be linked back to team wins in that season, and you can easily check the model’s performance by looking at how well it has performed since its inception. Shouldn’t that be the test of its strength?