*New Adj. +/- Rankings. Best Off/def and overall players

[mysticbb]

I watch Kobe tape all the time with different Coaches in Florida and these rankings are BS and stats don't mean nothing in basketball and things like an extra pass or a tipped steal or huslte, defense ect. don't show up at all on the stat sheet.

Well, we are talking about +/- numbers. If something a player does on the court has any effect on the result of the game, it will eventually show up in those numbers. So, your argument that stats mean nothing in basketball is pretty stupid, because that would mean the result of the game (in that case either a win or loss) would mean nothing at all. ;)

Anyway, jinxed, is there a description of the method somewhere? Just giving out some "adj +/- numbers" aren't that special at all.

[Tim Lehrbach]

And I guarantee you no teams are using this stat right now.

http://blogmaverick.com/2009/02/08/nba- ... e-numbers/

NBA fans are not quite ready for adjusted plus/minus, I think. And honestly, I don't know that the OP is helping the cause by misrepresenting its value. As Cuban notes, and as Dan Rosenbaum and others working on it have noted, adjusted plus/minus is NOT an answer to the question of how to rank every player in the NBA from best to worst. Just as Hollinger says of PER. It is, however, probably to most exciting trend in basketball statistics. It's a noisy measure of a player's contributions to winning, but it offers a certain elegance in that the raw numbers are simple fact (what happened to the scoreboard when the player was on the court) and the adjustments are all easily explainable moves towards placing a player's measurement perfectly in context (i.e. the effect of the other nine guys on the court). Where it loses me is in its relationship to the counting stats (the "statistical adjustment"). As a layman, I really don't get how that's done, and I've read Rosenbaum's old article a couple of times and follow the conversations at APBR.

I think what adjusted plus/minus does better than any other metric is not settle arguments but force people to come up with better ones. Somebody who wishes to convince you that Kevin Martin is a passable defender will have to reckon with this measurement. There are circumstances where a player's unexpected rating can be explained, but on the whole it's a phenomenal tool for keeping us grounded in what actually happened on the court.

[jinxed]

Here are some articles on adj plus/minus

http://espn.go.com/blog/truehoop/post/_ ... plus-minus

http://www.82games.com/ilardi1.htm

A more down and dirty article on how it's done

http://www.82games.com/comm30.htm

[jinxed]

I watch Kobe tape all the time with different Coaches in Florida and these rankings are BS and stats don't mean nothing in basketball and things like an extra pass or a tipped steal or huslte, defense ect. don't show up at all on the stat sheet.

AND WATCH KOBE DOIN WORK SO YOU SEE WHAT I AM TALKING ABOUT.

This just shows you have no understanding what plus/minus is. That tipped steal, hustle, defense...all of it shows up in adjusted plus minus. All of it. That backdoor screen, that three point shooter who opens up the floor, that boxing out so your teammate can get the rebound ..basically ANYTHING that is done on a basketball court that either helps your team score points our contributes your team stopping or giving up points is measured. How is this done?

Because +/- stats COMPLETELY IGNORE box score stats..All it does is measure how many points your team scores when you are on the floor, and how many points your opponents team scores when you are on the floor. Then using advanced statistical techniques such as REGRESSION ANALYSIS, they take into who the opponent is, is it home or away, who your teammates are when you are on the floor, are you playing against the other teams good lineup or bad lineup, in order to isolate a particular players contribution to the teams bottom line...scoring points and defending the other team from scoring..Let's take Kobe Bryant..

Your claim is that Kobe is a great defender, but if we look at his on/off stats from 2008-2009 it shows that the Lakers were a better defensive team by 2.5 points per 100 possessions when Kobe was OFF THE FLOOR. Now of course, you just can't look at this stat and say Kobe is bad at defense, a lot of factors could come into play such as who was one the floor when he came off, who his opponents were when he was on and off the court etc..what ADJ +- does is account for all of those things. That's why it is called ADJUSTED and what we find is that after doing those adjustments Kobe is still hurting the Lakers on defense every time he is on the court by and average of 1.7 points per 100 possessions (as opposed to the NBA's average defender being in that position)

You're right that Kobe is a great offensive player. He is without a doubt one of the best offensive players in the game. In fact he is the 7th best offensive player according to these numbers. But all of the guys in front of him are also perimeter players, so he isn't the best. In terms of pure offense Nash is #1, then Wade,LBJ,CP3... his mid range game is great, his three point shooting not so much. Kobe is 5/26 for the year on 3's, a dismal 19%. For his career he's 34%, which is below the league avg of 37%.

[_-IvIadNice-_]

Hmm I see what you are talking about. Makes sense. My bad if I went off but I just couldn't believe my eyes. I have another question why would Nash be a much more better offensive player than Wade?

Also I agree with Wade being number one in your offense/defence ranking due to his all around game.

And YES I know what +/- stats are but I just don't pay any attention to them.

[jinxed]

Classy response. Shows you're a guy who actually cares about basketball and knows the game well.

Nash routinely finds himself at the top of these offensive rankings year after year. What can I say that you don't already know, he's a maestro with the ball in his hands. He makes the offense run, even if he's not getting the assist, he's putting the ball in his teammates hands where he can make a good play..Steve Nash's teams (the Suns and before that the Mavs) have been some of the greatest offensive teams in the history of the game. It's no coincidence that it's because Nash is at the helm.

He's not much better than Wade..Nash makes his team 11.2 points per 100 posessions better on offense and Wade 10.6..so it's really close.

[_-IvIadNice-_]

Thanks by the way but his team's defense hasn't so shabby though for the past few years. Why is that they are exploding this season? Maybe defense? To me it looks like Frye is making an impact for them in rebounding and shooting (which he brings other centers to the three point line which opens up the lane for Stoudemire). Also Stouedmire is getting it right this year and Richardson is putting his dunks and three pointers back to the way it was.

[mysticbb]

Here are some articles on adj plus/minus

http://espn.go.com/blog/truehoop/post/_ ... plus-minus

http://www.82games.com/ilardi1.htm

A more down and dirty article on how it's done

http://www.82games.com/comm30.htm

Well, neither of those links gives me a description of this particular method. I find it interesting that he has such a low error. Usually the error is twice as much, even with some sort of stabilization.

[azuresou1]

Ballboy topic.

Who needs Dwight Howard when you can have Amir ****ing Johnson?

[wreck]

jinxed,

The statistics look about right, in my opinion. You cannot expect the general public to agree with the data shown because 1) Their minds can't comprehend what they are seeing 2) If that statistics don't show what they believe to be true, they will discount the statistics much quicker than their own lack of knowledge and understanding of the game.

Great thread. Keep up the good work.

[azuresou1]

jinxed,

The statistics look about right, in my opinion. You cannot expect the general public to agree with the data shown because 1) Their minds can't comprehend what they are seeing 2) If that statistics don't show what they believe to be true, they will discount the statistics much quicker than their own lack of knowledge and understanding of the game.

Great thread. Keep up the good work.

See, this is what major statheads always claim: that rather than that their stats or methods are flawed, that everyone else isn't 'smart' enough to see things the same way as them.

You're telling me you'd rather have Amir Johnson than Kobe Bryant and Dwight Howard? Of course you would, your stats tell you so.

+/- is at best a judgment of the gap between a player and his backup.

[bert stein]

any interpretation of plus/minus numbers, adjusted or not, should take into account that such measures do not take complementarities between players into account. this may not be such a big issue when comparing role players, but adjusted +/- will overestimate the value of a superstar who has the rest of his team constructed around his skillset, compared to one where the rest of the team is not so dependent on the superstar's skillset.

for example, I would guess that adjusted +/- overestimates lebrons / cp3's value and underestimates kobe's value (as opposed to impact). this is consistent with the observation that kobe has adjusted +/- number's that aren't commensurate with subjective estimates.

this is not meant as a criticism of the way adjusted +/- stats are constructed, since the attempt to measure a player's contribution without taking into account the nature of the rest of his teammates implicitly assumes that there are no complementarities between players, and thus all stats that attempt to estimate player value will have this flaw to some extent.

[mysticbb]

See, this is what major statheads always claim: that rather than that their stats or methods are flawed, that everyone else isn't 'smart' enough to see things the same way as them.

Well, you even give the "statheads" the proof with arguments like this:

You're telling me you'd rather have Amir Johnson than Kobe Bryant and Dwight Howard? Of course you would, your stats tell you so.

The stats doesn't tell me something like this, it just says that in his limited minutes against his direct opponents Amir Johnson helped his team winning by a certain margin. If I have to choose, I take the production of Howard or Bryant over Johnson's just because I know they can give me a better chance to win games due to their ability to be more constant in more minutes. That is what the best players in the game seperates from the other, consistency.

There is no particular stat which would determine the "best player" or "better player" in terms of the overall game.

+/- is at best a judgment of the gap between a player and his backup.

Uh, no. Anyway, I doubt that you even want to understand, therefore any further explanations might be casting pearls before swine.

this is not meant as a criticism of the way adjusted +/- stats are constructed, since the attempt to measure a player's contribution without taking into account the nature of the rest of his teammates implicitly assumes that there are no complementarities between players, and thus all stats that attempt to estimate player value will have this flaw to some extent.

Well, the adjustments takes exactly the rest of his teammates into account. Your basic idea about the adjusted +/- numbers is simply wrong. Read those articles about that and you will see that those adjustments have no bias towards Bryant (or any other player). In fact, Bryants adj. +/- for the last decade was awesome, only because some players are better in those advanced stats doesn't make it an insult. ;)

[bert stein]

this is not meant as a criticism of the way adjusted +/- stats are constructed, since the attempt to measure a player's contribution without taking into account the nature of the rest of his teammates implicitly assumes that there are no complementarities between players, and thus all stats that attempt to estimate player value will have this flaw to some extent.

Well, the adjustments takes exactly the rest of his teammates into account. Your basic idea about the adjusted +/- numbers is simply wrong. Read those articles about that and you will see that those adjustments have no bias towards Bryant (or any other player). In fact, Bryants adj. +/- for the last decade was awesome, only because some players are better in those advanced stats doesn't make it an insult.

no - you just didn't understand my point (but it's ok - it's a rather subtle point ). the adjustments take his teammates into account by assuming a _linear_ contribution of each teammate/opponent. It fails to adjust for complementarities between players' skillsets. Put it another way, the adjusted +/- procedure takes into account how good teammates are, but not how well they complement each other. In doing so, it overvalues players whose teammates complement their skillsets particularly well.

[mysticbb]

no - you just didn't understand my point (but it's ok - it's a rather subtle point :)). the adjustments take his teammates into account by assuming a _linear_ contribution of each teammate/opponent. It fails to adjust for complementarities between players' skillsets. Put it another way, the adjusted +/- procedure takes into account how good teammates are, but not how well they complement each other. In doing so, it overvalues players whose teammates complement their skillsets particularly well.

Uh, it wasn't that subtle, don't worry about that. But you might misunderstood my post. You basically wanted to imply that Bryant's adj +/- isn't that good, because his teammates didn't fit that well with him together. Instead of saying that this works both ways, you said it would have a specific "bias" towards Bryant. It is probably an ability of the players to make things work between each other. ;)

[bert stein]

Uh, it wasn't that subtle, don't worry about that.

I agree it wasn't that subtle - but the alternative was to imply that you were stupid.

But you might misunderstood my post.

I highly doubt that. Your original reply read:

Well, the adjustments takes exactly the rest of his teammates into account. Your basic idea about the adjusted +/- numbers is simply wrong.

so, without editing that post, there is no getting around the fact that you were flat out wrong and didn't get my original post.

You basically wanted to imply that Bryant's adj +/- isn't that good, because his teammates didn't fit that well with him together. Instead of saying that this works both ways, you said it would have a specific "bias" towards Bryant. It is probably an ability of the players to make things work between each other.

actually, if you reread my original post, I pointed out that (i) it was an example, and (ii) I was speculating.

The reason I chose LBJ, cp3 and kobe was that all 3 were superstars, but that the first two play in an offense that is utterly reliant on and built around them while kobe plays in a structured offense, with players that fit that offense well, that is able to make plays even when he is out. thus we might expect, based on my argument, that adjusted +/- would overweight the value of the former and underweight the value of the latter.

[jinxed]

he reason I chose LBJ, cp3 and kobe was that all 3 were superstars, but that the first two play in an offense that is utterly reliant on and built around them while kobe plays in a structured offense, with players that fit that offense well, that is able to make plays even when he is out. thus we might expect, based on my argument, that adjusted +/- would overweight the value of the former and underweight the value of the latter.

I understand your point and it is definitely an interesting one. However in respect to Kobe, ultimately it fails for three main reasons.

1) Kobe Bryant's usage % was 2nd in the league last year to Melo, higher than both LBJ and CP3

2)Before Pau got there Kobe was even more the focal point of the offense (he took an estimated 9 million shots a game) and his ADJ +/- is similar to what it is now.

3)We can look statistically at how well their respective offenses performed when LBJ or Kobe was out of the game. The Lakers without Kobe had an Off Rating (points per 100 poss) of 102.The Cavs w/o LBJ had an off rating of 101. So those numbers are VERY similar. It's not as if the Cavs were inept w/o LBJ in the game.

[bert stein]

3)We can look statistically at how well their respective offenses performed when LBJ or Kobe was out of the game. The Lakers without Kobe had an Off Rating (points per 100 poss) of 102.The Cavs w/o LBJ had an off rating of 101. So those numbers are VERY similar. It's not as if the Cavs were inept w/o LBJ in the game.

good point. I'm surprised cavs do so well with lebron off the floor, and that does suggest that the effect I hypothesized isn't significant, at least in comparing lebron and kobe.

[bert stein]

Where it loses me is in its relationship to the counting stats (the "statistical adjustment"). As a layman, I really don't get how that's done, and I've read Rosenbaum's old article a couple of times and follow the conversations at APBR..

I'm basing my explanation here on Rosenbaum's article on 82games. The basic idea is as follows: pure adjusted +/- is a noisy but relatively unbiased measure of performance. box-score based measures are less noisy, but biased (in the sense that they miss a great deal of the player's contribution). to combine the benefits of both, Rosenbaum does the following.
1) he runs a regression of adjusted +/- on box score variables. because he runs this regression over a large sample of players, the noisiness of adjusted +/- is not a significant issue. this gives him a "model" of the form: plus-minus = a*points + b*FGA + ... , where a,b,c etc are constants which capture the weight each stat is given.
2) he then estimates each player's statistical +/- by plugging the player's statistics into the "a*points + b*FGA + ..." formula. This gives us statistical +/-.
3) He then combines statistical +/- with adjusted +/-, weighing linearly, to obtain overall +/-.

In other words, statistical +/- is at heart a box-score-based performance measure, and thus is susceptible to the same criticisms as PER, win shares, etc. It does have an advantage over other box score stats because it is estimated from adjusted +/- rather than some team measure of success, but it doesn't manage to get around the problem that box stats miss many valuable contributions that may vary across players.

[jinxed]

I'm not a big fan of Rosenbaum's method of using Stat +/-. I don't think Ilardi does that, nor does Wayne Winston.

Also, I think it's important to keep in mind that APM is a measure of PERFORMANCE. And when you try and extrapolate it, by saying things like who would you rather have on your team, player X or Y- you are taking the player out of the system in which he is performing.