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

[mysticbb]

but the alternative was to imply that you were stupid.

That's what you wanted to imply in the first place and the reason why I started with "uh". Well, maybe you aren't as smart as you think ...

I highly doubt that.

And that is the reason why you failed to understand it again.

As I said it before it works both ways. Even if you assume there are non-linear contributions, it doesn't mean Bryant would be the only one effected by this. If Bryant has a lesser "value" due to the "structered offense" (I would call the Princeton Offense also "structured" and the Lakers running the TPO maybe only in 30% of the plays anyway, but who cares), it should also be true for his teammates. Which makes your point redundant, because Bryant just isn't as necessary for his team as James and Paul (well, you basically said that in the first place), which is exactly what you can get out of those adj. +/- numbers. Arguments like "under-" or "overvalue" are only make sense, if you think this adj. +/- determines the "best player" in the league. But that is not the way how that works. It would be stupid to assume that Odom or Gasol are the better players in comparison to Bryant. Both just had an bigger impact on the Lakers last season due to some specific circumstances.

Well, now what you are saying is, if a team is build around a certain player, it might be rather obvious that his adj. +/- is higher. The point of those +/- numbers is to determine how much impact a player has. And that takes both ends of the court into account. Looking at the numbers, Bryant had a significant lower value on defense than both James and Paul. It is rather unlikely that this has something to do with their "structured or non-structered offense". Now applying your idea to the past would also mean, that Bryant's +/- should be higher between 2004/05 and 2006/07, well in fact it is even slightly lower than his +/- between 2007/08 and today. While the Lakers offense was built more around Bryant than in recent years the numbers are implying the complete opposite of what your idea would propose (well, that is jinxed's 2nd point).

And as I said in my first answer to you, using this method of regression to determine adj. +/- should exactly take those complementarities out of the equation. It should be rather obvious that some players +/- numbers are benifit from playing together with better players, while it lowers the unadjusted +/- numbers of the better players. So, that is why I wrote your idea how that works is wrong. Adj. +/- want to take those things out, which you wanted to take in. If you want to see complimentarities between players, you can use those 5-man unit stats, which can help to determine which players are playing well together on a team.

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.

This is a very important point!

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

You are right, neither Winston nor Ilardi/Barzilai are using boxscore stats to determine the adj. +/-. That is their whole point "thinking outside the boxscore".

[bert stein]

but the alternative was to imply that you were stupid.

That's what you wanted to imply in the first place and the reason why I started with "uh". Well, maybe you aren't as smart as you think ...

probably not. but I do think I accurately gauge how smart you are.

I highly doubt that.

And that is the reason why you failed to understand it again.

I'll reiterate. You claimed that "the adjustments takes exactly the rest of his teammates into account". as I've already pointed out (although I don't think you understand this yet), this is incorrect. so there was no misunderstanding

As I said it before it works both ways. Even if you assume there are non-linear contributions, it doesn't mean Bryant would be the only one effected by this. If Bryant has a lesser "value" due to the "structered offense" (I would call the Princeton Offense also "structured" and the Lakers running the TPO maybe only in 30% of the plays anyway, but who cares), it should also be true for his teammates.

this last point is incorrect, but I won't quibble, since it is irrelevant. my point, as was clear in the original post, was that if we try to compare player _value_ across teams (as opposed to within teams), adjusted +/- will miss some part of that.

Which makes your point redundant, because Bryant just isn't as necessary for his team as James and Paul (well, you basically said that in the first place), which is exactly what you can get out of those adj. +/- numbers. Arguments like "under-" or "overvalue" are only make sense, if you think this adj. +/- determines the "best player" in the league. But that is not the way how that works. It would be stupid to assume that Odom or Gasol are the better players in comparison to Bryant. Both just had an bigger impact on the Lakers last season due to some specific circumstances.

yes. and, as was clear in my original post, I was pointing out that any attempt to measure player value (ie, who is the "best player" in the league) as opposed to player impact was flawed. you explicitly disagreed with my post in your reply. so either you are confused or you are backtracking.

Well, now what you are saying is, if a team is build around a certain player, it might be rather obvious that his adj. +/- is higher. The point of those +/- numbers is to determine how much impact a player has. And that takes both ends of the court into account. Looking at the numbers, Bryant had a significant lower value on defense than both James and Paul. It is rather unlikely that this has something to do with their "structured or non-structered offense". Now applying your idea to the past would also mean, that Bryant's +/- should be higher between 2004/05 and 2006/07, well in fact it is even slightly lower than his +/- between 2007/08 and today. While the Lakers offense was built more around Bryant than in recent years the numbers are implying the complete opposite of what your idea would propose (well, that is jinxed's 2nd point).

correct. which is why I conceded that my hypothesis was incorrect after jinxed's post. this lack of empirical support does not detract from it's consistency as a theoretical hypothesis - which was what you objected to in your original reply.

And as I said in my first answer to you, using this method of regression to determine adj. +/- should exactly take those complementarities out of the equation. It should be rather obvious that some players +/- numbers are benifit from playing together with better players, while it lowers the unadjusted +/- numbers of the better players. So, that is why I wrote your idea how that works is wrong. Adj. +/- want to take those things out, which you wanted to take in. If you want to see complimentarities between players, you can use those 5-man unit stats, which can help to determine which players are playing well together on a team.

ugh. no. you continue not to get the point. the notion of complementarities is orthogonal to the concept of player quality (which you implicitly apply when you allude to "better players"). I'll be explicit: it is incorrect to claim that "using this method of regression to determine adj. +/- should exactly take those complementarities out of the equation". I'll be charitable here and assume that you don't understand the concept of complementarities.

[mysticbb]

I'll reiterate. You claimed that "the adjustments takes exactly the rest of his teammates into account". as you admit below, this is incorrect. so there was no misunderstanding

Well, maybe the fact that I'm not a native speaker (in terms of english) makes it harder for me to explain what I mean, but you obviously (at least for me) don't understand the point.

this last point is incorrect, but I won't quibble, since it is irrelevant. my point, as was clear in the original post, was that if we try to compare player _value_ across teams, adjusted +/- will miss some part of that.

Well, and if you would have read the whole post in which I said the adj. +/- doesn't tell me that Amir Johnson is the better player in comparison to Howard and Bryant, you should have been able to understand that I didn't reply to this specific part of your post. Well, in fact I agree with you on this (obviously).

so either you are confused or you are backtracking.

Well, you obviously didn't follow the whole thread otherwise you wouldn't write something like this ...

ugh. no. you continue not to get the point. the notion of complementarities is orthogonal to the concept of player quality. I'll be explicit: it is incorrect to claim that "using this method of regression to determine adj. +/- should exactly take those complementarities out of the equation".

Adj. +/- wants to determine the individual impact of a player which is (and that is the reason for doing it) likely hidden by complimentarities between players. To get rid of those effects to determine the individual impact that the reason to make those adjustments. Otherwise it would be rather pointless, because you could just use raw +/- numbers.

[bert stein]

so either you are confused or you are backtracking.

Well, you obviously didn't follow the whole thread otherwise you wouldn't write something like this ...

ugh. no. you continue not to get the point. the notion of complementarities is orthogonal to the concept of player quality. I'll be explicit: it is incorrect to claim that "using this method of regression to determine adj. +/- should exactly take those complementarities out of the equation".

Adj. +/- wants to determine the individual impact of a player which is (and that is the reason for doing it) likely hidden by complimentarities between players. To get rid of those effects to determine the individual impact that the reason to make those adjustments. Otherwise it would be rather pointless, because you could just use raw +/- numbers.

sigh. no. the adjustment of adj. +/- corrects for the effect of the quality of teammates/opponents on raw +/-. it does not correct for complementarities between teammates.

here is an extreme example to illustrate the point clearly. assume player X is an average point guard on team A. no other player on team A is a point guard, although all the other players are quite average. In this case, we can expect player X to have an incredibly high adjusted +/-, simply because the team will be abysmal when he is off the floor. The reason is that other players function well as substitutes for each other, but no other player is a good substitute for X. That is, in the example I constructed, there are nontrivial complementarities. And as a result, although X is merely average in quality, he has extremely high adjusted +/- (which captures his impact on the team, but not the effect of complementarities).

[mysticbb]

(which you implicitly apply when you allude to "better players").

Uh, you added this one later, which makes your argument true. As I said I agree with you that you can't use those things to determine the "better player". If you wanted to say that in the first place, I'm sorry, because I couldn't get that out of your post.

And regarding the adj. +/- we maybe argue about two different things. If I see it right, you are talking specifically about Dan Rosenbaum's approach, while I talking about Winston/Sagarin or Ilardi/Barzilai. The latter don't use boxscore stats to determine the quality of the teammates. The idea behind this is really to take those effects like player A is playing well together with player B out of those numbers to get an impression about the individual impact on his team.

With your example you just make again an argument that you can't use +/- stats to determine how good a player is. Player Y on another team (also a PG) with a worse adj. +/- can be still better than your player X. Well, as I said it before in this post, if that's what you wanted to say in the first place, I can only agree with you. But I obviously had an different impression after reading your first post.

[bert stein]

(which you implicitly apply when you allude to "better players").

Uh, you added this one later, which makes your argument true. As I said I agree with you that you can't use those things to determine the "better player". If you wanted to say that in the first place, I'm sorry, because I couldn't get that out of your post.

And regarding the adj. +/- we maybe argue about two different things. If I see it right, you are talking specifically about Dan Rosenbaum's approach, while I talking about Winston/Sagarin or Ilardi/Barzilai. The latter don't use boxscore stats to determine the quality of the teammates. The idea behind this is really to take those effects like player A is playing well together with player B out of those numbers to get an impression about the individual impact on his team.

With your example you just make again an argument that you can't use +/- stats to determine how good a player is. Player Y on another team (also a PG) with a worse adj. +/- can be still better than your player X. Well, as I said it before in this post, if that's what you wanted to say in the first place, I can only agree with you. But I obviously had an different impression after reading your first post.

i'm referring to the ilardi/sagarin approach.
I'll accept that we mostly agree at this point.

[TrueLAfan]

As bert stein notes, basketball is a team game in a way that is going to make +/- a little iffy, no matter what. Teams often have (very) different defensive schemes and styles depending on who comes off the bench and how they manage rotations. Do you pull your elite player with one or two other good players at the same time, and try to rest them all at once? Does your primary complementary defender to your elite player stay on the court? If so, does he have a different scheme when the elite player isn't on the court? Does the team play differently--more traps, slowing it down more? These aren't the kind of distinctions that will make small differences; they'll be pretty large. And there's simply no way to determine values of individual roles and outcomes of that in a statistical analysis. That's the "noise" that Rosenbaum refers to. It does not even, really, correct for the qualities of other players. It corrects for what the same analysis determines that quality of the other players are...which means if the tool either has a design flaw (possible) or has made a misjudgment (given the qualities/styles/number of complementary players--very likely to happen somewhat often), it will compound it with lengthier and deeper analysis.

This is, obviously, a problem. I think it effectively refutes the original idea that this particualr stat/analysis (or any) can provide "quite simply, THE BEST STATISTIC WE HAVE FOR MEASURING HOW GOOD A PLAYER IS." Which doesn't mean that this sort of regression analysis isn't valuable. But, like most statistical tools, it is really a secondary analytic tool--meant to be used with (lots) of other tools and firsthand observations. It shouldn't be used to as a primary way of judging player value.

[DrunkOnMystery]

As bert stein notes, basketball is a team game in a way that is going to make +/- a little iffy, no matter what. Teams often have (very) different defensive schemes and styles depending on who comes off the bench and how they manage rotations. Do you pull your elite player with one or two other good players at the same time, and try to rest them all at once? Does your primary complementary defender to your elite player stay on the court? If so, does he have a different scheme when the elite player isn't on the court? Does the team play differently--more traps, slowing it down more? These aren't the kind of distinctions that will make small differences; they'll be pretty large. And there's simply no way to determine values of individual roles and outcomes of that in a statistical analysis. That's the "noise" that Rosenbaum refers to. It does not even, really, correct for the qualities of other players. It corrects for what the same analysis determines that quality of the other players are...which means if the tool either has a design flaw (possible) or has made a misjudgment (given the qualities/styles/number of complementary players--very likely to happen somewhat often), it will compound it with lengthier and deeper analysis.

This is, obviously, a problem. I think it effectively refutes the original idea that this particualr stat/analysis (or any) can provide "quite simply, THE BEST STATISTIC WE HAVE FOR MEASURING HOW GOOD A PLAYER IS." Which doesn't mean that this sort of regression analysis isn't valuable. But, like most statistical tools, it is really a secondary analytic tool--meant to be used with (lots) of other tools and firsthand observations. It shouldn't be used to as a primary way of judging player value.

Whatever, TrueLAFan. Clearly you just aren't willing to accept that Dwight Howard is clearly one of the worst defensive players in the league, and was just the 4th best player on the Magic last season(behind Lewis, Turkoglu, and Courtney Lee!) like the other sheeple.

[Jimmy76]

As bert stein notes, basketball is a team game in a way that is going to make +/- a little iffy, no matter what. Teams often have (very) different defensive schemes and styles depending on who comes off the bench and how they manage rotations. Do you pull your elite player with one or two other good players at the same time, and try to rest them all at once? Does your primary complementary defender to your elite player stay on the court? If so, does he have a different scheme when the elite player isn't on the court? Does the team play differently--more traps, slowing it down more? These aren't the kind of distinctions that will make small differences; they'll be pretty large. And there's simply no way to determine values of individual roles and outcomes of that in a statistical analysis. That's the "noise" that Rosenbaum refers to. It does not even, really, correct for the qualities of other players. It corrects for what the same analysis determines that quality of the other players are...which means if the tool either has a design flaw (possible) or has made a misjudgment (given the qualities/styles/number of complementary players--very likely to happen somewhat often), it will compound it with lengthier and deeper analysis.

This is, obviously, a problem. I think it effectively refutes the original idea that this particualr stat/analysis (or any) can provide "quite simply, THE BEST STATISTIC WE HAVE FOR MEASURING HOW GOOD A PLAYER IS." Which doesn't mean that this sort of regression analysis isn't valuable. But, like most statistical tools, it is really a secondary analytic tool--meant to be used with (lots) of other tools and firsthand observations. It shouldn't be used to as a primary way of judging player value.

Whatever, TrueLAFan. Clearly you just aren't willing to accept that Dwight Howard is clearly one of the worst defensive players in the league, and was just the 4th best player on the Magic last season(behind Lewis, Turkoglu, and Courtney Lee!) like the other sheeple.

The data suggests that Howard is actually a good defender but he hurts his team on offense is why his ranking is low.

[azuresou1]

As bert stein notes, basketball is a team game in a way that is going to make +/- a little iffy, no matter what. Teams often have (very) different defensive schemes and styles depending on who comes off the bench and how they manage rotations. Do you pull your elite player with one or two other good players at the same time, and try to rest them all at once? Does your primary complementary defender to your elite player stay on the court? If so, does he have a different scheme when the elite player isn't on the court? Does the team play differently--more traps, slowing it down more? These aren't the kind of distinctions that will make small differences; they'll be pretty large. And there's simply no way to determine values of individual roles and outcomes of that in a statistical analysis. That's the "noise" that Rosenbaum refers to. It does not even, really, correct for the qualities of other players. It corrects for what the same analysis determines that quality of the other players are...which means if the tool either has a design flaw (possible) or has made a misjudgment (given the qualities/styles/number of complementary players--very likely to happen somewhat often), it will compound it with lengthier and deeper analysis.

This is, obviously, a problem. I think it effectively refutes the original idea that this particualr stat/analysis (or any) can provide "quite simply, THE BEST STATISTIC WE HAVE FOR MEASURING HOW GOOD A PLAYER IS." Which doesn't mean that this sort of regression analysis isn't valuable. But, like most statistical tools, it is really a secondary analytic tool--meant to be used with (lots) of other tools and firsthand observations. It shouldn't be used to as a primary way of judging player value.

Whatever, TrueLAFan. Clearly you just aren't willing to accept that Dwight Howard is clearly one of the worst defensive players in the league, and was just the 4th best player on the Magic last season(behind Lewis, Turkoglu, and Courtney Lee!) like the other sheeple.

The data suggests that Howard is actually a good defender but he hurts his team on offense is why his ranking is low.

First off, 35th is a pretty mediocre rating.
Second off, if his defensive rank is 35, and his overall rank is 96, his offensive rank must be damn near Jason Collins status.
Third, Shard is 8th in adjusted +/-, while Dwight is 96th. Do you POSSIBLY think that this is because when Shard sits, they're playing Tony Battie, while when Dwight sits, they're playing Marcin Gortat, and the talent dropoff isn't anywhere near comparable?

IMO adjusted +/-, or at least the current formula being used, is one of the worst basketball stats created, and is certainly not "THE BEST STATISTIC WE HAVE FOR MEASURING HOW GOOD A PLAYER IS." The only other stats that I can think of that are worse are 'clutch stats' and the jumpshot attempt % at 82games, which lists Shaq at 21%. Maybe WS.

[azuresou1]

Another thing to note: Chris Paul went from one of the worst defensive +/- players in 07-08 to one of the best defensive +/- players in 08-09. He pretty much didn't do anything different.

[mike-]

Adj. +/- is far far too noisy to make definitive judgments in the vast majority of cases.

[jinxed]

Explain to me how the backup effects the starters adj +/- score in the regression analysis...I'm not saying it doesn't, i just don't understand how. This is a plus/minus stat, not an on/off stat. Two different things.

And Azureus, can you please name a better stat?

[jinxed]

Explain to me how the backup effects the starters adj +/- score in the regression analysis...I'm not saying it doesn't, i just don't understand how. This is a plus/minus stat, not an on/off stat. Two different things.

And Azureus, can you please name a better stat?

[azuresou1]

Explain to me how the backup effects the starters adj +/- score in the regression analysis...I'm not saying it doesn't, i just don't understand how. This is a plus/minus stat, not an on/off stat. Two different things.

And Azureus, can you please name a better stat?

Sure, let me explain. We'll use the Atlanta Hawks as an example, since I'm a Hawks fan.

In 07-08, Mike Bibby had a regularized adjusted offensive +/- of 1.737. The next season, in 08-09, despite being healthier and posting more efficient numbers all around, and the Hawks being a significantly better team, this figure dropped to 1.079. What happened?

Flip Murray happened. In 07-08, Bibby's backup was either Acie Law or Salim Stoudamire. Salim Stoudamire is no longer in the league, and Acie Law likely won't be either in a season or two. In 08-09, Bibby's backup (and they played almost mutually exclusively) was Flip Murray, who also happened to post career bests on the Hawks where he was mostly allowed free reign of the ball.

Thus, despite Mike Bibby shooting 2.1% better in FG% and 3PT% and boosting his AST/TO ratio from 2.6 to 3.125 and everything else pretty much staying the same, and ATL winning more games, his offensive +/- dropped .7 points - because Flip Murray is a significant upgrade over Acie Law.

I don't have 09-10 numbers yet, but I'm going to suspect that Jamal Crawford will drop Joe Johnson's +/- numbers; not sure about Bibby, who is increasingly beginning to be relegated to a 3pt shooter or a spacing threat.

Stats taken from http://www.hoopnumbers.com/teamAnalysis ... is=offRAPM


As for your question, there IS no single stat to determine the 'best' player, on either offense or defense, or as a whole. You need to look at a range of stats to paint a complete picture. This is true even in sports like baseball, which is individual data is far easier to extrapolate. Say a starting pitcher has an ERA of 2.25. Simply stellar. If just given this fact alone, you'd say he should win the Cy Young. But what if you break it down just one level, and you see that his ERA against righties is 1.00 and his ERA against lefties is 18.00? Now this paints a different picture: your pitcher is useless. Sure, he torches righties all day, but when a left handed batter comes to the plate, if there's anyone on base, you'll need to pull him from the mound because he's going to bleed runs. Now, I pulled the numbers out of my ass, and admittedly I don't watch much baseball, but I hope you get the point about how just one stat doesn't tell you much. This is especially true in basketball, which is a team sport, and there is so much clutter.

If you must know, here are the stats I primarily look at when judging a player: TS%, PPG, TRB%, AST%, STL/BLK%, TOV%, Opponent Counterpart 48-Minute PER, and MPG. This lets me know that a player shoots efficiently, can maintain that efficiency while taking a large number of shot attempts, rebounds, passes well, plays adequate lane/help defense, doesn't turn over the ball much, and actually gets enough playing time that I can feel comfortable with the sample size. I'm not a big fan of PER, but there's really no great defensive stat, which is why I use Opponent Counterpart 48-Minute PER to judge, as it shows me Shane Battier holds opposing SGs to 12.0 PER and SFs to 13.9 PER, while David Lee lets opposing PFs get a 21.8 PER and Cs a 19.7 PER.

Does this all make sense?

[jinxed]

You still haven't explained how your backup effects your plus/minus ranking. It still seems to me that you are confusing plus/minus and on/off stats. Let's use Mike Bibby as an example.

If Bibby goes in the game and the Hawks go up +7...then Bibby's Plus/minus ranking (unadjusted) is +7 for that time period. If he then goes out of the game and Acie Law comes in and the Hawks lose that period by -4, then Bibby's on/off score (given equal minutes) would be +11. However, his plus/minus would still be + 7. Plus/minus is a statistic that measures what happens when you are in the game, NOT when your out of the game.

As for how Mike Bibby's adjusted plus/minus could go down despite his other numbers going up? For many reasons, maybe his defense slacked a bit..you know..half the game of basketball that your stats don't measure. However, this may all be mute because..

Bibby's adj. plus minus in 2007-2008 according to Ilardi was -0.09 and in 2008-2009 it was 0.91..so according to his stats they actually went up..(although ilardi uses previous years to help measure and bring down the error score), basketballvalue also has higher numbers for bibby in 09 then 08.

[azuresou1]

You still haven't explained how your backup effects your plus/minus ranking. It still seems to me that you are confusing plus/minus and on/off stats. Let's use Mike Bibby as an example.

If Bibby goes in the game and the Hawks go up +7...then Bibby's Plus/minus ranking (unadjusted) is +7 for that time period. If he then goes out of the game and Acie Law comes in and the Hawks lose that period by -4, then Bibby's on/off score (given equal minutes) would be +11. However, his plus/minus would still be + 7. Plus/minus is a statistic that measures what happens when you are in the game, NOT when your out of the game.

As for how Mike Bibby's adjusted plus/minus could go down despite his other numbers going up? For many reasons, maybe his defense slacked a bit..you know..half the game of basketball that your stats don't measure. However, this may all be mute because..

Bibby's adj. plus minus in 2007-2008 according to Ilardi was -0.09 and in 2008-2009 it was 0.91..so according to his stats they actually went up..(although ilardi uses previous years to help measure and bring down the error score), basketballvalue also has higher numbers for bibby in 09 then 08.

Defense didn't factor in, since like I said, I posted the adjusted OFFENSIVE +/- stats.

Also, I'm getting different numbers for Bibby's overall adjusted +/-: -.910 in 07-08 and -.935 in 08-09. Source: http://www.hoopnumbers.com/teamAnalysis ... lysis=RAPM


To answer your point
Say we have Bibby/Flip. Hawks go up 30-18 to end the 1st quarter; Bibby gets a +12. Awesome. Start of the second quarter, Flip Murray comes in, and the Hawks stay pretty even. Bibby comes back with the Hawks up 50-35 with 4 minutes left; he's still at +12. They finish the half at 60-41. Bibby finishes the half with a +16.

Now we have Bibby/Law. Hawks go up 30-18 to end the 1st quarter; Bibby gets a +12. Start of the second quarter, Acie Law comes in, and the Hawks get demolished. Bibby comes back Hawks down 40-33 with 6 minutes left; he's still at +12. They finish the half down 49-48. Bibby is now at a +18 for the half.

In both cases, when Bibby is in the rate of scoring is exactly the same: 2.5 PPM for the Hawks, 1.5 PPM for the opposing team. However, because Flip Murray is better than Acie Law, Bibby plays less time when Flip is his backup versus when Acie is his backup. Thus, he has less time to build up (or lose) his adjusted +/-. Obviously, it doesn't always happen like this, but it's an example of how +/- is a cloudy stat.


TL:DR
Against bad teams, if my backup is good enough that he can maintain or build on a lead, then I can sit more on the bench, which is nice for my oft-tired body, but worse for my adjusted +/-, since then I don't get to stat-pad and build the lead even more. If my backup sucks, then games are a lot tighter and I get more playing time to crush runs from bad teams, building my +/-.

Conversely, if we're facing a great team and we're down some, if my backup is good-to-decent, we will be down but still in the game, which forces me to play, and since they're a better team, my adjusted +/- drops. On the other hand, if my backup sucks, then in the time he's in we're going to get utterly demolished, and there's no point in my coach bringing me back in when we're in an insurmountable hole and tiring me out on a lost game, which preserves my adjusted +/-.

Does this all make sense?

[jinxed]

In both cases, when Bibby is in the rate of scoring is exactly the same: 2.5 PPM for the Hawks, 1.5 PPM for the opposing team. However, because Flip Murray is better than Acie Law, Bibby plays less time when Flip is his backup versus when Acie is his backup. Thus, he has less time to build up (or lose) his adjusted +/-. Obviously, it doesn't always happen like this, but it's an example of how +/- is a cloudy stat.

But APM is adjusted for minutes played, so as long as Bibby is in there for the 2.5 PPM, it doesn't matter if he is in there for 5 minutes or 40.

TL:DR
Against bad teams, if my backup is good enough that he can maintain or build on a lead, then I can sit more on the bench, which is nice for my oft-tired body, but worse for my adjusted +/-, since then I don't get to stat-pad and build the lead even more. If my backup sucks, then games are a lot tighter and I get more playing time to crush runs from bad teams, building my +/-.

Conversely, if we're facing a great team and we're down some, if my backup is good-to-decent, we will be down but still in the game, which forces me to play, and since they're a better team, my adjusted +/- drops. On the other hand, if my backup sucks, then in the time he's in we're going to get utterly demolished, and there's no point in my coach bringing me back in when we're in an insurmountable hole and tiring me out on a lost game, which preserves my adjusted +/-.

Does this all make sense?

Your APM won't drop if you are playing against a better team. Adjusted means that it is adjusted for who you are playing with and who you are playing against. You would be right if we were talking about pure plus/minus, but we aren't. Adj..takes into account who is on the court with you and who is on the court against you and your minutes played.

[azuresou1]

I'm under the impression that APM only adjusts for quality of opposition and teammates and not minutes? Could you point out where it says otherwise?

And of course APM drops when you're playing a better team that's beating you and goes up when you beat a bad team: it's you losing or beating a team that makes you better or worse, right? The NBA isn't like a tournament everyone plays against one set team and then derives their APM based off how they did against that one team. It's adjusted, yes, but ultimately if you play only bad teams and crush them, your APM will be higher than if you played amazing teams and got repeatedly destroyed.

TL:DR; the quality of your opponents is inversely proportional to your own quality as a player/team. The Nets would have insane APMs if they were a NCAA team, even though they're the same players.

In any case, you didn't address how I solely used offensive APM for my initial Bibby post, and how it dropped .7 after getting Flip Murray instead as a backup, despite having better stats across the board and Atlanta winning more games and by a higher average margin. Unfortunately, I can't find the average margin of wins stats anywhere, but I will provide the following:

Better stats in bold and larger (just bold is kinda hard to see)

07-08: 14.9 PER, .526 TS, .498 eFG, 0.5 ORB, 32.9 AST%, 15.5 TOV%, 2.6 AST/TO, 21.5 USG%, 14.1 PPG, 1.737 OAPM

08-09: 16.3 PER, .544 TS, .518 eFG, 1.6 ORB, 24.2 AST%, 10.6 TOV%, 3.125 AST/TO, 20.6 USG%, 14.9 PPG, 1.079 OAPM

How do you play more efficient ball across the board, have your team win more games by more points than before, and then have a LOWER offensive adjusted +/-?

[mysticbb]

I'm under the impression that APM only adjusts for quality of opposition and teammates and not minutes? Could you point out where it says otherwise?

Iliadi is using per 40 minutes numbers. Just read his article about that on 82games.com. ;)

And of course APM drops when you're playing a better team that's beating you and goes up when you beat a bad team: it's you losing or beating a team that makes you better or worse, right? The NBA isn't like a tournament everyone plays against one set team and then derives their APM based off how they did against that one team. It's adjusted, yes, but ultimately if you play only bad teams and crush them, your APM will be higher than if you played amazing teams and got repeatedly destroyed.

Well, that explains why Wade lead the league in adjusted +/- last season ... wait ..., anyway, I doubt that you have any idea what you are talking about. Please, read those links before you answer again.