Suns' players plus/minus--last 30 days

[F-Dog]

This is from www.82games.com, a neat little site with good stats and analysis. I suck at code, so try and bear with me:

Individual Player Floor Time statistics -- Mar. 6 to Apr. 4

Player Min    +/- Off48 Def48 Net48  W  L   Win% 
McDyess 34%   +14  96.3  93.6  2.7   5   5   50.0 
Barbosa 51%   -41  99.6  104.9  -5.3   7   7   50.0 
Voskuhl 13%   -11  93.1  98.7  -5.6   3   2   60.0 
Johnson 92%   -84  97.4  103.5  -6.1   5   9   35.7 
Stdmire 83%   -96  97.2  104.9  -7.7   3   12   20.0 
Marion 85%   -101  96.1  104.1  -7.9   4   9   30.8 
Jacobsn 57%   -77  92.5  101.5  -9.0   5   10   33.3 
Cabrkapa 22%   -32  89.0  98.6  -9.6   4   9   30.8 
Lampe  21%   -34  98.8  109.2  -10.4   1   11   8.3 
Harvey 14%   -28  96.8  109.8  -13.0   4   6   40.0 
White 10%   -24  90.2  105.0  -14.8   1   7   12.5 
Eisley 13%   -51  86.5  112.7  -26.3   0   6   0.0

http://www.82games.com/0304PHOR.HTM

I rearranged them according to Net48, the team's net points per 48 minutes when the player is on the floor.


Pretty much what you'd expect, I suppose. The thing that interests me is that playing time has seemingly been adjusted so that the higher the player is on this list, the more minutes he gets. I wonder how much the Suns' coaching staff uses +/- when they are making decisions about PT...

Anyway, just a little food for thought.

 

[George O'Brien]

The McDyess numbers certainly stick out. I knew he was doing well, but that is impressive.

But the relately good numbers by White don't just seem to lead to more minutes.

 

[F-Dog]

But the relately good numbers by White don't just seem to lead to more minutes.

The only player White's (+/- per 48 min) is better than is Howard Eisley.

Like Eisley, White is legitimately hurt right now, and I believe he's out for the year. Actually, it's possible that White's inconsistent minutes throughout the season have been partly due to nagging injuries.

 

[George O'Brien]

Like Eisley, White is legitimately hurt right now, and I believe he's out for the year. Actually, it's possible that White's inconsistent minutes throughout the season have been partly due to nagging injuries.

He hurt his thumb or something like that relatively soon after joining the Suns which might explain why he has so much trouble catching the ball. Unfortunately, he had a reputation for being injury prone in Washington.

I think that White could become a serviceable backup if he lost some weight and became quicker. He gets abused by quicker opponents, but he has had some impressive moments.

 

[Errntknght]

F-Dog, "I wonder how much the Suns' coaching staff uses +/- when they are making decisions about PT..."

Hopefully, not very much. +/- stats are quite simple and understandable when used to rate lineups, though like many conglomerate stats you need a large sample size to average out the variation in the opponents contribution. When you start to extract the contributions of the individual players from the lineup +/- it is a 'whole other ballgame'.
At the time I ran a test on that part of it, the method for extracting the players scores was to take a weighted average of the lineup's +/-, for the lineups the player was part of. (The weighting being the PT of each such lineup.) The results were very bad, with an average error just slightly less than the average +/- per48 differential. I was testing the mathematical method itself, using hypothetical data, and the method simply did not calculate anything close to the right answers. I can't tell from the website whether the 82games people are using this old math method or not but if they are the individual +/- figures should be taken with a very large grain of salt.

In case you are wondering, there is another mathematical method, which will at least get the right answers on hypothetical data... basically it is matrix inversion. Of course, no mathematical method will help with a flaw in the whole idea - namely that if two players are in exactly the same lineups then they will get exactly the same score, independent of what each contributes. If two players come close to that condition then their scores will be close to the same, regardless of what each contributes.

30 days worth of data, something like 12 games, is not a large sample size.

 

[Joe Mama]

Well, that might as well have been in Chinese for me.

Joe Mama

 

[George O'Brien]

To me the +/- is only marginally useful. A guy who starts should have a better +/- than a guy who usually plays with other backups. A starter who plays very long minutes like JJ should have a weaker +/- than one who plays fewer minutes, because more minutes are with backup players.

 

[Joe Mama]

To me the +/- is only marginally useful. A guy who starts should have a better +/- than a guy who usually plays with other backups. A starter who plays very long minutes like JJ should have a weaker +/- than one who plays fewer minutes, because more minutes are with backup players.

you seem to be forgetting that backups play a lot of their minutes against other backups also.

Joe Mama

 

[George O'Brien]

you seem to be forgetting that backups play a lot of their minutes against other backups also.

Joe Mama

Yeh, but their backups are better players.

Seriously, basketball is a team sport. Unless the player is the only reason for the +/-, it is likely to be distorted. It takes doing a multiple regression analysis which no one would understand anyway.

 

[F-Dog]

Well, that might as well have been in Chinese for me.

Joe Mama

I was following pretty well until I got to the part about 'matrix inversion'. I'm sure the folks at that site would love to discuss the math with Errntknight if he contacted them, though.


Obviously, there's no point in using +/- to compare Marion with Lampe. However, if the Suns' bench players have similar roles and usually aren't out there at the same time, so I believe these numbers say something about how effective they have been relative to each other. (With the tiny sample size, they're not much help in predicting the future, of course.)

 

[Errntknght]

"It takes doing a multiple regression analysis which no one would understand anyway."

Multiple regression analysis isn't required - matrix inversion is simple, straightforward and does the trick, as long as you have as many lineups as players.

I think high school algebraII covers inverting 2X2 and 3X3 matrices... some may recall the word 'determinant' - that is the same topic. It's also known as solving 2 equations in 2 unknowns, etc. It's not major league math...

I did visit a number of sites related to 82games and they seem to be rather proprietary about just what they are doing - not to mention having delusions of grandeur. I wouldn't dream of using a stat that the proponents of keep shrouded in mystery.

 

[Phill11]

"It takes doing a multiple regression analysis which no one would understand anyway."

Multiple regression analysis isn't required - matrix inversion is simple, straightforward and does the trick, as long as you have as many lineups as players.

I think high school algebraII covers inverting 2X2 and 3X3 matrices... some may recall the word 'determinant' - that is the same topic. It's also known as solving 2 equations in 2 unknowns, etc. It's not major league math...QUOTE]


Wow. It's Major League math for me! Nerds..........

 

[George O'Brien]

I'm way over may head here, but I think matrix approach over simplifies the problems.

1. Players are often used in matchup situations. For example, White is often used when opponents have really big opponents, yet he can actually dominate smaller guys like what happened when he played against the Lakers without Shaq.

2. Player combinations vary depending on foul situations in terms of when the player is being used and for how many minutes.

3. A large number of short time periods on the floor do not have the same meaning as a smaller number of longer stints.

4. Game situations when the player is used can vary. Players who are used when the team is being blown out is will have different statistics than when the team is ahead or in a close game.

So while lineup combinations will get a better view than a straight +/-, I am not convinced it is going to really tell much.

 

[Errntknght]

George, "So while lineup combinations will get a better view than a straight +/-, I am not convinced it is going to really tell much."

I don't have any idea what you are trying to say.

What do you mean by straight +/- ? For individuals or for lineups?

I see a value for lineup +/- stats ... it gives you some measure of how well a lineup meshes or the amount of lineup chemistry if you will. On both ends of the floor at once! Which is something very few stats measure. If two lineups are regularly used in very different situations then comparing their +/- stats is useless or close to it. Were I a coach and wanted to make use of +/- stats to compare two lineups, I might even go to the lengths to make sure they are used in comparable situations or perhaps use only those minutes that come close to matched situations.

So what do individual +/- stats measure? They can't measure the chemistry since it's only a player at a time. In fact, the concept of an individual player having a +/- means that he has a certain value whenever he steps on the floor, regardless of the lineup he is part of. Now if that were true then lineup chemistry wouldn't mean anything. Coaches and most of us believe lineup chemistry means quite a lot - after all the five positions of a team are staffed differently from each other in belief that a balance of skills is necessary to a lineup and that is the foundation of lineup chemistry.
It's not a good practice to collect a stat if you don't have some concept of what it might measure. The straight +/- for a player obviously is not a direct measurement of anything about an individual player, while it is a direct measurement about a lineup. You have to resort to statistical arguments about lineup effects cancelling out, plus all the other causes of variation also cancelling out to convince anyone the individual stat has any meaning. Now you might have a situation where a team used lots of lineups and shuffled people in an out a bunch where the lineup effects might nearly cancel but, in general, I wouldn't expect them to come close so I don't see that the stat can stand on it's own.

 

[George O'Brien]

George, "So while lineup combinations will get a better view than a straight +/-, I am not convinced it is going to really tell much."

I don't have any idea what you are trying to say.

What do you mean by straight +/- ? For individuals or for lineups?

I see a value for lineup +/- stats ... it gives you some measure of how well a lineup meshes or the amount of lineup chemistry if you will. On both ends of the floor at once! Which is something very few stats measure. If two lineups are regularly used in very different situations then comparing their +/- stats is useless or close to it. Were I a coach and wanted to make use of +/- stats to compare two lineups, I might even go to the lengths to make sure they are used in comparable situations or perhaps use only those minutes that come close to matched situations.

So what do individual +/- stats measure? They can't measure the chemistry since it's only a player at a time. In fact, the concept of an individual player having a +/- means that he has a certain value whenever he steps on the floor, regardless of the lineup he is part of. Now if that were true then lineup chemistry wouldn't mean anything. Coaches and most of us believe lineup chemistry means quite a lot - after all the five positions of a team are staffed differently from each other in belief that a balance of skills is necessary to a lineup and that is the foundation of lineup chemistry.
It's not a good practice to collect a stat if you don't have some concept of what it might measure. The straight +/- for a player obviously is not a direct measurement of anything about an individual player, while it is a direct measurement about a lineup. You have to resort to statistical arguments about lineup effects cancelling out, plus all the other causes of variation also cancelling out to convince anyone the individual stat has any meaning. Now you might have a situation where a team used lots of lineups and shuffled people in an out a bunch where the lineup effects might nearly cancel but, in general, I wouldn't expect them to come close so I don't see that the stat can stand on it's own.

If understod, I think I would agree.