Okay, let’s continue with the “history” theme. Previously I posted rankings of every player in the history (since 1977) of the Utah Jazz and the Boston Celtics (see below):
Ranking Every Player in the History of the Utah Jazz
Ranking Every Player for the Boston Celtics since 1977
Given the response to these “rankings”, I thought I would analyze another team. Across the past three decade no other team has gone to the finals as often as the LA Lakers. So in honor of their latest trip, here is what every player on the Lakers has done with respect to Wins Produced since 1977.
Table One: Ranking the LA Lakers (1977-78 to 2008-09)
Recently much has been said about the relative value of Kobe and LeBron. And some have even compared these two players to Michael Jordan. However, maybe people should think about Kobe and Magic. If Kobe and the Lakers prevail – over the Magic – Kobe will have won four titles. This is only one less than Magic (the player, it would be four more than the Magic franchise).
So is Kobe that close to Magic? Table One suggests otherwise. In 13 seasons, Magic produced nearly 300 wins and posted a 0.429 WP48 [Wins Produced per 48 minutes]. This mark leads the franchise across the past thirty years. Second on the list is Kobe. But as one can see, it’s a very distant second. In 13 seasons (and more than 1,000 additional minutes), Kobe has produced 149 wins with a 0.207 WP48. It’s important to emphasize, Kobe is twice as productive as an average NBA player (average WP48 is 0.100). But he has never offered anything close to what we saw from Magic.
Kareem vs. Shaq
Just like we see with the first two names on the list, there may also be a debate about the next two players listed, Kareem and Shaq. The ranking is based on Wins Produced since 1977, and as one can see, Kareem did produce slightly more for the Lakers across the years examined. But if we look at WP48, Shaq is clearly the more productive player. Before fans of Kareem get too upset, it’s important to note that Kareem was 30 years old in 1977-78 and his numbers reflect what he did until the age of 41. Kareem did manage to produce 12.3 wins [with a 0.225 WP48] at age 37 (which is great for an old guy). But in his last four seasons Kareem only produced 14.0 wins (which is still pretty good for a really old guy).
In contrast to Kareem, Shaq’s numbers begin at age 24 and end when he is 31. So we are looking at Shaq in his prime. If we compare Shaq and Kareem at age 30 and 31 we see the following:
Kareem at 30: 18.6 Wins Produced, 0.394 WP48
Kareem at 31: 22.9 Wins Produced, 0.348 WP48
Shaq at 30: 17.0 Wins Produced, 0.321 WP48
Shaq at 31: 14.8 Wins Produced, 0.287 WP48
In sum, I think Kareem in his prime did more than Shaq. Of course, if Shaq could have hit his free throws more consistently it might have be a different story.
It’s also important to remember when looking at this list that it is simply a ranking of what each player did – with respect to Wins Produced – since 1977 (in other words, it is what it is). The Shaq and Kareem comparison tells us that you have to think about why each player posted the numbers we observe (in the case of Kareem, age is clearly an issue) before leaping to any conclusions.
The Top Ten
For those who don’t wish to click on the above table, here is the list of top 10 players (again, with respect to Wins Produced for the Lakers since 1977).
1. Magic Johnson
2. Kobe Bryant
3. Kareem Abdul-Jabbar
4. Shaquille O’Neal
5. James Worthy
6. A.C. Green
7. Vlade Divac
8. Michael Cooper
9. Lamar Odom
10. Byron Scott
There are 193 players ranked, so you will have to look at the table to see more.
Let me close by essentially repeating what I said when I posted the ranking for the Celtics.
In general, reactions to such analysis follow two paths. If the person reacting likes the analysis (i.e. I always knew Magic was the best), then the reaction will look like this… “Professor Berri – with some of the best analysis I have ever seen – has confirmed that Magic is the greatest player for the Lakers since 1977.”
If the person, though, does not like the analysis (i.e. Kurt Rambis is 11th?), then you see… “Berri – who no one thinks knows what he is talking about – actually thinks Rambis is a great player. That is all you need to know to see how stupid all the advanced stats are. Why can’t these geeks put the computer down and watch a freakin’ game.” Or something like that (usually the language is more colorful).
Although I like the first approach (and I am not too keen on the second), both reactions have the same problem. In both instances the person reacting is arguing from conclusion back to evidence. In other words, their reaction to Wins Produced is entirely dictated by whether or not what it says confirms what the person already believed. If it does, then Wins Produced is great. If not, then it’s stupid.
Unfortunately, this is not how one should do analysis. When we do research we start with the evidence and work to the conclusion. And if we think a conclusion is incorrect, we have to actually go out and find sufficient evidence that allows us to reach a different conclusion. Oh, and by sufficient, I mean the new evidence shouldn’t be accurately described as “horseshit”.
Okay, after I said that on the post for the Celtics there were still a few comments (just a few, I don’t want to paint everyone with the same brush) that suggested reading comprehension is not important to every fan of the Celtics (or to put it another way… my impression of the unhappy fans was pretty good). Now we get to see if fans of the Lakers react any differently.
- 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.
Now this is the Best.Post.Ever!
Thanks Teacher!
I am actually surprised to see Michael Cooper so high up there. He was mostly a bench player, but he was a defensive force to be reckoned with, so it’s not shocking to see his WP up there.
My biggest shock is seeing Derek Fisher Wins Produced so low. Seems that every Laker team he’s played with has gone to championship. Like a friend used to tell me, film don’t lie. Numbers don’t either, and while Derek Fisher provides other intangibles, his production has always been lackluster. He’s held in such high regard here in LA that people will forgive his low productivity.
It’s funny seeing a bench player like Coop beat out a starter like Fisher. Then again, those 80′s Lakers were the best team ever!
Mike – now this is an interesting observation regarding Cooper. If the WOW matrix is, as many commenters claim, insufficient in accounting for any defensive “effort”, what do defensively-renowned players like Cooper do (that show up on the stat sheet) to make them productive players?
On the other hand, what if a “defensive guru” does not have a good WP48 value – e.g. Bruce “Lee” Bowen – does it mean that their label of a defense force is only a mirage, or something the media / fans given to them based on impression, aka “watching the game”?
Since I’m a regular player who put a lot of energy and effort in defense, I do think the value of defense is somehow, some way reflected on the stat sheet. I suppose the truth is that there’s no defense in the world (basketball world) to completely shut down opponents from scoring any point, and the game of basketball really is about putting the ball in the opposite basket and scoring more points than your opponent does, the end result (win or lose) does come down to the most obvious offensive measures.
Thanks Dave.
Obviously, the star players on Showtime and Shaq and Kobe lead the list.
Just want to say that this blog has become significantly more awesome now that people consistently make fun of the repetitive complainers. Many props to jbrett for the idea.
i woulda thought Bob McAdoo would’ve been higher on this list than Sun Yue.
Something seems off there
Berri,
Is there anyway you can rehash the lack of success of teams who lacked a .3wp48 player. Thanks in advance
@Brizzle
The fact that Sun Yue is higher than McAdoo is a reflection of the fact that he barely played. Because both players had wp48 of less than .000, the one who played less is likely to have harmed the team less. In fact, if you look at wp48, you’ll see that McAdoo is rated much higher. Also worth noting that all of McAdoo’s best years come before 1980. He arrived in LA in 1981.
Professor I asked a question in the last thread, about the relative strength of eras, I’m just wondering is the league stronger now than in the 80s/90s, or was it stronger then?
Cool analysis btw, will you be doing one for the Bulls?
Michael,
I expect players to get better over time, but WP48 does not answer that question.
I can do the Bulls soon.
Please do one for other NBA teams… maybe this offseason?
Thanks!
Hi Prof. Berri – I know the team histories are popular right now, but I was wondering if you had any thoughts on the Dwight Howard touches issue? The story is a bit past its prime, but it’s very weird – Dwight is one of the most efficient scorers in the league, but Orlando doesn’t do better when he takes more shots. Any idea on what they do worse when he shoots more? Thanks!
I’m surprised no one has commented on Lamar Odom’s WP48 (higher than Kobe’s, but i’m guessing not significantly so).
On a different note, I can’t wait to see the wins produced (and other statistics) for lebron in 2010 when he’s running and gunning under the d’antoni system.
Got it. thanks for the explanation Zach
Alex: I suspect that when he takes more shots FG% decreases. That is, they are better when he only takes good shots (which for him are basically only dunks) and is eschewing forcing the issue by taking more, worse shots.
But I’m probably wrong.
The basic ranking is probably consistent with the views of most observers. I suspect that most would rate Kobe a little closer to Magic, but still far off.
I believe the basic difference is that regardless of whether the author agrees or not, any 15 year old basketball player can tell you that offensive output is a function of personal skill set and aggressiveness.
A player can up his scoring significantly, but not without taking extra lower percentage shots.
A player can up his FG% significantly, but not without lowering his scoring.
The only way a player can really improve is by upping his skill set and making better decisions.
One will improve usage and efficiency and one improve efficiency.
The problem that most teams face is that they need enough scoring to win, but there aren’t enough dunks, layups, shots inside the paint with a foul, and wide open 3 pointers by a gun slinger to go around etc… Those are the kinds of shots that lead to high levels of efficiency.
As a result, on any team, there are some skilled players that wind up taking some of the more difficult shots that space the floor, that draw double teams, that are forced because the clock is winding down, or because they are a good enough option in that sequence etc…
Those guys are contributing in ways that allow other players to attain adequate or even superior efficiency despite inferior skills.
Kobe is extremely high usage, extremely skilled, constantly drawing double teams, and has defenses totally geared towards stopping him etc… . He is in a completely different role than Magic was in. Magic was not the “primary scorer” on his team, let alone did he have to carry the team at times.
IMO, this model does not reward these highly skilled super high usage guys properly.
It’s puts too much weight on scoring efficiency and not enough weight on usage and the benefits that accrue to teammates and the team as long as that player’s efficiency doesn’t drop sharply and they aren’t taking stupid shots when better alternatives exist.
IMO, this is the opposite problem of PER where very low efficiency players and players that make poor shooting decisions are undeservedly rewarded for their high scoring.
Ganesh,
IMO, Lebron is so obviously the best player in basketball any debate is just plain silly.
He’s extremely skilled, team defenses are totally geared towards stopping him, he draws double teams constantly, is super duper extremely high usage, and despite all that remains extremely efficient.
And if that wasn’t enough, he rebounds well, plays good defense, blocks shots, and has assist totals that are better than many fine PGs.
And if all that wasn’t enough, he accumulates those numbers on a slow paced team.
This is the one and only player that can be compared to MJ and actually be part of a serious conversation.
IS,
I feel some obligation to read comments left in this forum. And I have noticed that you are leaving essentially the same comment over and over again. This is what a 15 year old might do. So you can stop doing this now.
Wow, IS, that’s…so much of what has been said ad-nauseum before I don’t know where to even begin. What are the numerical codes again?
It’s all good and dandy that you bring up these points, but.. do you have any empirical evidence that high usage rates impact things in the way you describe?
Also wasn’t Magic actually more skilled than Kobe in numerous ways, just not in scoring?
Darn it, Prof. Berri, you beat me by mere seconds!
Simon,
I should just leave policing this forum to the readers. Generally you do a better job than I do.
prof. berri//
Careful, now some people will liken you to Hitler and call some of us Nazis who “police” these forums against those who will not follow your dogmas blindly. :O
But honestly I do think most of what could’ve been discussed have already been mentioned somewhere between 2006~2008.
I forgot to say this before, but I really liked the economics-related posts, and would love to read some more econometrics pointers in the future when you have more time.
Oh and notice that I just fulfilled Godwin’s Law by bringing up the ex Fuehrer’s name. ;)
dberri,
I apologize for the repetitiveness.
I would say I am reptitive because the premise of just about every one of your articles is that some high usage big scorer is overrated or some efficient low usage player is underrated.
I happen to think you have the best model I have seen to date, but IMHO, you have not addressed the usage/efficiency relationship and other aspects of the game I am being repetitive about to my satisfaction (nor has anyone else).
IMO, they are basic to an understanding of the game and a players ability/value and account for many of the extreme differences of opinion between various stats oriented models, NBA players, NBA executives, fans etc…
Again, I apologize, but it’s equally frustrating for fans to read the same criticisms of certain players over and over again when they miss the mark in some obvious and basic ways.
IS,
We addressed the link between shot attempts and shooting efficiency in 2006. And we will do it again in this next book.
Quick review…changing the number of shot attempts does not dramatically impact shooting efficiency. We looked at this across 30 years of data and that is what we see. So you can insist that usage and efficeincy are related. But this does not show up in the data.
Simon,
I believe that Magic was a much more skilled player than Kobe in several ways. He was also a better decision maker. On a net basis, I agree that he was a much better player than Kobe. It is the degree I am disputing.
You are asking for evidence of the relationship between usage/efficiency etc…
Well, IMO that’s the problem.
It’s not very easy to provide the data required to analyze this properly because coaches usually don’t decide to have their least skilled players take an extra 10 shots per game and their most skilled players take 10 fewer shots so they can see how well the team did and how their respective TS%, PPG etc.. changed.
Managements are also not likely to change their compensation incentives towards rewarding efficiency over what the coach perceives to be the best possible use of the team’s scoring resources (all of which are different).
Usuage tends to be fairly constant. It changes as a player’s skill level changes, the makeup of the players on the team changes, the defenses change etc… IMO it would be difficult to isolate what I am talking about in circumstances like these, but I am not the stats guru.
The way you know it exists is by playing and observing the game. That’s why I said I knew it at 15 years old. It was to stress the point that everyone that has ever played the game understands the concept and knows the impact is not insignificant if you change usage significantly.
Other than that, I am sure others, including those with a much better grasp of stats, have brought up these points before. But you have to remember, I have not read every article and discussion on this and every other basketball forum. There are always new readers with the same questions and observations.
If people would prefer not to cover the same ground, that’s understandable. But if the questions haven’t been answered satisfactorily, they remain questions.
D. Berri,
I look forward to your next book. I may be pain in the ass, but it is not by design. It is because I am genuinely interested, like to express opinions, and tend to be contrarian by nature. I make my living as a value investor and gambler – both of which require going against the prevailing consensus . I was unreasonably barred at Knickerblogger for this kind of thing.
I would have to see your data before I could comment.
However, I suspect I will not be satisfied. It’s not the shot count that matters, it’s the types of shots taken and avoided that count. Sometimes the player’s postion and role on the team dictate the types of shots he takes. Then, reducing the count will not matter. It’s the shot distribution that has to change.
As I said in the prior post, IMO there are also going to be complications with looking at the historical record in this regard because sometimes usage changes for sensible reasons that lead to a steady or even improved efficiency.
It is best explained by example:
IMO, if Kobe were to simply reduce his shot count it would not translate into higher efficiency.
He would have to reduce the very specific lower probabilty ouside shots he takes now because so much of the scoring burden is on him. He would have to allow lower skilled players to take them instead.
That would never happen because the coach would go crazy!
In fact, I think it did happen one time in a single playoff game a couple of years ago and everyone did go crazy.
I believe Kobe was taking some heat for not getting his teammates more involved (which is actually a fair criticism of his game). So he decided not to shoot at all in the second half. The team couldn’t get anything going at the offensive end. Then he took all sorts of heat for not being the superstar leader and taking control of the game and providing the scoring down the stretch etc… LOL
I may have the details slightly confused, but that was the gist of it.
IS,
Aren’t you essentially saying that there should be some correlation between # of shot attempts a player takes and the % of that player’s baskets that are unassisted? The reasoning here is that the hard, “end of shot clock”/”manufacture your own offense” types of shots that the Kobes and Carmelos of the world take might tend to be unassisted. Let me know if this is the kind of thing you’re thinking of.
ZK
You always want efficient scorers, but you also want guys who can create shots, make plays for teammates.
A lot of efficient fg % players have very specific offensive roles. They usually feed off of players who create space or draws double teams.
E.g., if you have a center who feeds off double teams or pick and rolls and scores 5 ppg at 55% clip, you are not going to ever see him averaging 20 ppg. If he ever tried to score that many points, you’d probably see his efficiency drop dramatically because he’ll be taking a lot of bad shots. But in a professional league like NBA, guys with limited skill sets would never try something crazy like that – otherwise they’d be out of jobs.
Therefore, it’d be extremely hard to find evidence for what IS is talking about. However, I think that if you put together a bunch of efficient role players with limited offensive skillsets, their WP48s will probably all go down quite dramatically, because guys will be forced into roles they’re not familiar with. Fortunately, most GMs are not dumb enough to put together a team of non-scorers.
We’ve already seen this in Cle-Orlando series, where Cle’s offense basically became a one-man show, because besides Lebron, no one else could create anything. It was one of the ugliest offenses I’ve ever seen from a 66 win team.
simon,
A basketball prospectus article a couple of years back found an inverse correlation between usage and efficiency. There is a lot of variation between players and teams, however, but it seems that most players have a “wall” that they hit, and from there their efficiency steadily drops as their shots increase. But for most players, that tends to be around the 25-30 possessions/game range.
There is a reason that Joel Pryzbilla only shoots the ball 5 times a game, and there is a reason that LaMarcus Aldridge gets around 18 looks, despite Joel being a more efficient scorer. There’s a reason Anthony Morrow gets fewer looks than Stephen Jackson. Etc. It’s difficult to tell this from the box score, though it is usually fair to infer that players that score little, but get a lot of offensive rebounds and/or 3-pointers are not able to increase their usage and maintain efficiency without support from their team.
Kobe actually bucks this trend – his TS% has remained stable the past few years despite an increase in usage. But the “taking tough shots because there’s no time on the shot-clock” situation has only a marginal effect upon his efficiency, and he is actually pretty middle-of-the-road when it comes to high-usage players. He just takes a lot more shots without losing efficiency, which few players are doing.
Whether they can’t do it is another question entirely.
Because the Wins Produced is a counting stat (not a rate stat) the compensation for being high-usage is already built in. The high usage player will score more points. Provided the player is shooting over about 33% from 3 and 50% from 2, the increase in shots will increase the total wins produced. Below 1 point per possession on shooting possessions, added possessions decrease wins produced–as they should; they are now below the approximate efficiency of the NBA (~1PPP).
dsm,
Its about 1.08 for the NBA this year. I think
Alex, I never bothered reading all these posts so I may be repeating something someone else already said, but I imagine the Magic suffer when Dwight shoots a lot because they are a perimeter oriented team. They take a lot of 3s which is one of the reasons they are successful. I remember Hollinger did a piece on it a couple of weeks ago.
What I really want to hear is dberri’s impressions of Orlando manhandling Cleveland.
I was as surprised as anyone… but I he picked the Cavs in 4 or 5.
My initial reaction is that this would be a case in point example of how match-ups trump stats on paper… but I lack the statistical knowledge or experience to even know how to go about investigating that initial impression.
brgulker, you don’t need statistical knowledge to make that argument! You just need to say three little words ‘small sample size’! I think the prof’s stated before that basically anything can happen in a playoff series because it’s such a small sample. What about the Magic Lakers, who do you see winning in that series? The Magic owned that match up in the reg season as well.
But, what if the Magic are repeatedly better against teams who are “better/more productive” than they are?
The Magic owned the Cavs in the regular season, and that continued into the Playoffs.
If the trend continues in the Finals, is the sample size big enough to make any conclusions?
If prof. Berri is using a 30-year dataset, then injuries of high-use players and “scaling up” of high-efficiency, low-use teammates should provide an interesting natural experiment! They don’t happen every day, but there’s always a couple of major incidents every season, and together they should add up to a decent sample. Gilbert Arenas – the king of making ill-considered shots more often than he reasonably should? – alone may suffice. :)
montecristo,
Arenas just about the exact same career TS% and eFG% as Kobe.
Not saying I really have much of a point, just putting it out there that maybe, just maybe, Arenas isn’t a rarity.
As IS and simulator have pointed out, the data’s confounded. Wins Produced is better than whatever Hollinger uses to create his Holy Grail Stat™ (Eye of Newt mixed with ppg, I assume), but it is still fundamentally flawed and should not be taken on its own.
Also, we need a summary of the Finals and an explanation for why the data utterly failed to predict Cleveland’s implosion. We’re paying for this service, after all, so make it snappy…
KBRC- While WP may or may not have flaws (and Berri has never claimed that it has 100% explanatory power), the Cleveland – Magic series is not an example of a possible WP flaw. Do you see why?
Also, my impression — and I haven’t gone back to look at the data — is that the Magic team that played against Cleveland was different in makeup from the reg season and even the rest of the playoffs. That might be a further nail in the coffin of an already dead argument.
Prof Berri –
I’ve suggested several times that you add a few commenters as siteadmins to moderate comments. I’m sure there’s a few regular commenters here who you can trust to police the site on a reasonable basis.
KBRC,
1. Six games is a small sample size. And until the final game, every single one was very close. The Cavs could have won 4-1 or been swept if a few possessions had gone differently.
2. Though the Cavs were the best team in the league during the regular season, the Magic were also very good. There was not a huge discrepency between them.
3. If you want to look at individual players, Delonte West and Mo Williams had terrible shooting outings, a product of both the Magic’s defense (best in the league) and missing tons of open shots. Z gave them nothing. Ben Wallace was injured. Besides Lebron, no one by Varajao had a respectable outing.
Orlando shot crazily well from downtown as a team. Dwight Howard was ridiculous. Lebron’s performance over these playoffs has been unreal, even though he was a bit turnover prone against the Magic. But his teammates regressed to average, while Dwight’s did not.
“Wins Produced is better than whatever Hollinger uses to create his Holy Grail Stat™ ”
Actually for the most part wins produced and per seem to agree with which players are good and which aren’t so good. Its just per tends to overvalue a particular type of scorer and wp can seemingly overvalue a particular type of role player. I think Hollinger does good work, his articles are usually quite astute and he uses a wide range of data. Obviously professor Berri is much more well versed in the use of statistics etc, but Hollinger’s work has value too.
Brgulker, I was actually asking your opinion, who do you think will win in the finals? :-)
I second the recommendation for comment moderators on the conditions that repetitive comments be deleted and that all comments disparaging Glen Davis in previous posts are never deleted.
Re: “What I really want to hear is dberri’s impressions of Orlando manhandling Cleveland.”
I’m not dberri but I’ll offer my 2cents anyway.
The point differential between ORL and CLE was a single basket (Hollinger has CLE being up 2 on ORL, BRef gives CLE 2.7). That’s it. The series went 6 games, each team had home court thrice, let’s call that a wash. So you have a series of 6 games that will come down to a basket. That means some drama. Take away Lebron’s game winner or any of Lewis’ two game winners and the series outcome is likely another. This is just what you’d expect for two teams that are a basket apart, yes? It’s like flipping a coin that is not quite fair, instead of 50-50 maybe it’s 53-47. Still plenty close to not be surprised when the “weaker” side comes up. Not really much of a story here I think. I have a hard time calling this an upset. Now, LAL losing to HOU would have been quite a meltdown/upset but ORL beating the CAVS, no…
A brief follow-up:
Playoff rounds with point differential (series winner on left)
1st Round
CLE (10) vs DET (-0.6)
ORL (7.3) vs PHI (0.1)
BOS (8.2) vs CHI (-0.3)
ATL (1.7) vs MIA (0.2)
LAL (8.1) vs UTH (2.9)
DEN (3.6) vs NO (1.7)
DAL (2.1) vs SAS (4.2)
HOU (4.4) vs POR (6.1)
2nd Round
CLE (10) vs ATL (1.7)
ORL (7.3) vs BOS (8.2)
LAL (8.1) vs HOU (4.4)
DEN (3.6) vs DAL (2.1)
Semifinals
LAL (8.1) vs DEN (3.6)
ORL (7.3) vs CLE (10)
Finals
LAL (8.1) vs ORL (7.3) –my pick is LAL 6
I just don’t see an upset. The “statistical upsets” were all a push in my eyes (teams separated by a single basket). This holds irrespective of discounting Garnett and Ginobili absences on their teams’ differential, altho this skewed the numbers against their respective teams. The statistical oddity I found intriguing is DEN. Things came down to a basket, yes, but it is odd when that one basket always goes your way. That’s how DEN got thru Rd 1 and 2 and pushed LAL. So DEN gets the overachiever award for the playoffs…
Zach,
It’s hard for me to explain exactly what I am talking about.
D Berri has been very patient with me so far and I really don’t want to get barred. Here’s my last attempt at this.
I think the goal of every player should be to maximize the relationship between the team’s need to score and the team’s scoring efficiency (not his own scoring and efficiency).
I think the balance that maximizes a player’s own stats and the balance that is best for the team is not always the same. A team needs to score “X points” to win, but not all the shots needed to get there will be easy or even desireable. Some players will have a bigger burden in taking those tougher shots.
I believe it is foolish to argue that there isn’t a relationship between usage and efficiency. Here, I am not talking about shot count alone. I am talking about the types of shots taken.
For example:
If a player insists on taking only dunks and layups, he can surely become an extremely efficient scorer, but he’s not going to score as much as he would with a more balanced approach and will often hurt the team.
If a player gets hyper aggressive and is looking to score every time he touches the ball, he can surely increase his PPG, but he’s probably going to reduce his efficiency and hurt his team.
Depending on the makup of the team…..
Some guys are going to take more tough low probability outside shots than others because they are relatively skilled at it and it is “requirement” to successful basketball.
Some guys are going to take more shots around the basket because they are big, skilled at getting to the basket and it’s a requirement of successful basketball etc…
Some guys are going carry a very large offensive burden that forces them to take some shots they would otherwise pass on if they were trying to maximize their own stats.
Some guys are so skilled, they are going to draw double teams, have the defense geared specifically designed to stop them, get much better looks for their teamates etc…
When you add it that up (an other factors), most models reward higher scoring and better efficiency etc.. That’s the way it should be.
What I am saying is that IMO no one has it “just right”.
When we see huge discrepancies between the perceptions of players/coaches/media and specific models or descrepancies between individual models about a specific player, IMHO it is often a function of this balance between scoring and efficiency, the role of the specific player on the team, etc…
In aggregate, the models may be fairly close or even correct, but in specific circumstances I think they miss the mark.
Semifinals
LAL (8.1) vs DEN (3.6)
ORL (7.3) vs CLE (10)
Factor in losing Jameer Nelson (as Berri did), and the disparity between Cleveland and Orlando is significant… yet Orlando pulled it out. You can say it simply comes down to statistics or you can say there are variables beyond your comprehension.
Michael – I agree that Hollinger has some good stuff, but all of it is unrelated to PER.
You haven’t taken too many stats courses, have you, KRBC?
Btw, I have greatly enjoyed saying your name in a robot voice over and over. Don’t know why.
In Italion Stallion’s defense, he’s making intelligent points, so I’m hoping he doesn’t get disowned on this board. He’s not posting “Allen Iverson is the greatest guard of all time, screw this site.” IS is presenting articulate, thoughtful points, but is getting insufficient responses. When he posts something, people are just like, “Don’t you mean A and B?” And I’m betting IS doesn’t have the luxery of reading every post and comment in the history of this site. So maybe he is rehashing old arguments, but can someone give a thoughtful answer? Give up the letter code joke for a minute, and maybe IS will understand what you guys want him to understand.
Italian Stallion,
The reason that it is so difficult for you to explain yourself is that you are relying only on vague anecdotal evidence. I have seen very persuasive critiques of economic/econometric models before, and while most of them use some form of anecdote they all provide data that can at the very least have one interpretation that supports the critique. You *must* do the same to be taken seriously here. There are plenty of critics at apbrmetrics, so being unable to do the analysis yourself isn’t really an excuse.
Ray I think people just get tired going around in circles with this stuff. It leaves a bitter taste in the mouth. Much better to just appreciate the work done here. Took me a little while to learn that lesson but I realised that this is a cool site, its better to just enjoy what’s said here with the
knowledge that you don’t have to agree with all of it all the time.
What Michael said. I doubt that anyone agrees with everything Professor Berri writes, and most, Berri included, are very open to discussing critiques with even a little bit of thoughtful (read: more than just your subjective anecdotal opinion) analysis behind them.
Italian Stallion,
I think the argument of the model is that scoring is linear. Basically, regardless of whether I take one shot or fifteen shots, I have the same chance of hitting any given shot. As such, it doesn’t matter how many shots I take.
I understand that you disagree with this and believe that players that shoot more shots are more valuable then those that don’t. Perhaps the way you should use this model is by separating high volume shooters from low volume shooters.
Instead of comparing Kobe to say Trevor Ariza, you only compare Kobe to Lebron or Arenas. By doing this, you can tell which high volume shooters are productive and which are not.
To elaborate: the Italian Stallion style posts detract from the comment section because they lead to diversions like this and crowd out thoughtful discussion. IS and all the rest may have any opinion they choose, but please stop spamming this forum.
I don’t think that the WP model inherently has any kind of player or scoring model built into it. I think that the only thing it does is computing weights for different statistics (points, rebounds, fg%, etc. in some non-overlapping fashion) wrt differentials (and there have already been a lot of knowledgeable critics for the equation, so I won’t go into that).
There may have been some analysis of a correlation between # of shot attempts and fg% for and finding that there’s no significant correlation, but that is not equal to saying that efficient scorers will always be efficient scorers no matter how many shots they take.
To show that, we’ll need a large database of many players in the league who have alternated between taking a small number of shots to taking a large number of shots. Unfortunately, you are not going to find enough data for that type of analysis because players are pretty much stuck in their roles in the league.
“You haven’t taken too many stats courses, have you, KRBC?”
Enough to know that correlation doesn’t equal causation :).
My point was that Orlando’s efficiency differential without Jameer Nelson was closer to 5 rather than 7.3 in the regular season, yet they still beat the Cavs soundly. People are revising their predictions now, saying that Orlando going into this series was the same team as it was in the regular season with an efficiency differential of 7.3, as if that explains why they prevailed. This isn’t the same team, though, and pretending that it is is disingenuous.
Sports are more complicated than the flip of a coin (although I’m sure some statisticians and economists will disagree); the model is flawed.
Honestly, I’m just a bit bitter because I decided to quit making subjective picks in the NBA PicknRoll contest (you pick who will win each round in how many games prior to the playoffs starting) and went purely by efficiency differential, home court, and efficiency differential between teams during the season…
As a result, I went against my better judgment and picked the Blazers to defeat the Rockets in 6 (I thought the Rockets matched up well against the Blazers and should pull it out). In the conference finals, I picked Cleveland over Orlando in 6 and the Lakers over the Nuggets in 5. Harrumph.
John G.
What makes it difficut to explain is that at the micro level (as opposed to modeling things at some aggregate NBA player level) there are sometimes many detailed issues that can change perceptions.
I suspect it’s extremely difficult to isolate all these issues statistically and that accounts for why we have discrepancies from model to model. I know I don’t have the skill to do it.
I have not criticized this model exclusively. In fact, I have been saying all along that I like this one the best. I don’t think anyone has a better one. So everyone, please don’t take my commentary personally (I have already had enough experiences with hyper sensitive stats guys when I discuss their work and beliefs).
What I am suggesting is that everyone suspend their love affair with their current favorite numbers and models and just use some common sense. If you were previously a basketball player, some of this should be rather obvious.
Does anyone on earth really believe that David Lee could score close to 30 a night without resorting to a lot more low probability jumpshots etc… that would result in a lower efficiency?
Does anyone on earth really believe that Kobe couldn’t score 10-15 points a night at a much higher efficiency if you offered him an extra few million to do that instead of score more?
Most models try to account for this by looking at total shots, misses, points etc… The resultant ratings reflect this information rather well in the vast majority of cases and are even quite consistent from model to model.
However, there are occasional exceptions.
D. Berri has pointed out how and why PER tends to over rate some high scorers to my satisfaction in several articles.
Now I am simply suggesting that perhaps this model has the opposite problem.
It tends to rate players like Kobe, Maravich, AI, etc… quite a bit below the perceptions of NBA players, coaches, and other astute observers (and my own observations of the stats and games).
It also tends to rate some of the lower usage efficient players much higher.
I actually have no problem with saying that Kobe, Maravich, AI etc…. are generally overrated. I think the stats clearly reflect that.
What I have a problem with is the extent of the difference between the model and the more bullish general perceptions because when I look at the details, I believe some of the value of these players is not being reflected properly in the model.
I don’t know why people take my comments personally, consider them off topic, or get annoyed when the article in question is almost always making commentary about the relative merits of the player I am discussing.
No one has to agree, but if you disagree and want to convince me, you have to make a case similar to one that D Berri made about PER.
Until, then I and many others will believe some of the value of these players is not being reflected and continue listing some of the specific reasons why we feel that way.
If you simply don’t want to hear criticism, don’t want to hear tough questions, can’t answer those questions to my satisfication yet, etc… then I understand why I am being considered a pain the ass. But really, I am not the problem.
I’m just a guy that wants to rate players better than I can now and hasn’t drank the kool-aid on any of the models that currently exist.
(sorry for any inconvenience)
Berri,
Please…please…please rehash your article on the superstar theory. It is the only reason I picked Boston last year and I have a feeling it will reel its head again this year.
rear its head
I vehemently vote AGAINST commentors acting as admins. All that will do is set up a war between the antagonistic cliques of pro- and anti-PER camps, one of which will have the predominance of commentors who can censor the other. The site will rapidly degenerate when one side is driven away and criticism is stifled. The last thing we need is a team of censors, each of whom has the power to further degrade debate based on their own peculiar diea of what “acceptable” means.
I think I see your point IS.
One side effect of what you’re describing (I think) is that all other four players may run the offense horribly, give it to Iverson who misses a last second off balance shot, and the system will put the brunt of the blame on AI.
Or Jordan can draw a triple team, force up a shot that misses, Longley grabs the rebound and putback and the system will reward Luke, above all.
More generally, (I hope i’m not completely off), the system rewards or punishes individuals for actions that are a great deal team effort or team screwup.
I think…
1. Kobe Bryant
2. Magic Johnson
3. Shaquille O’Neal
Pingback: Disagreeing with Doug Moe « The Wages of Wins Journal