Warning: This is math-intensive...

So as promised, here is the thread

I thought as I examined the 5-man units at 82games.com, how could I correlate winning with which player was out on the court?  So I came up with this idea: plot the winning percentage against the players, using "+1" as the player on the court and "-1" as the player on the bench. 

In an ideal situation, the Wolves win 100% of the time when player X plays and 0% of the time when he's on the bench.  This would look like a nice 45 degree line, kinda like this (just imagine all of the axes are labeled differently):

Recall that "winning percentage" means the times that the 5-man unit outscored the opposition divided by the total number of times the unit played.

Here is a table of the 5-man raw data, sorted by win percentage (apologies for the small text):

 

 When +1's and -1's are assigned, I broke up the resulting graphs into two for ease of viewing.

Here's the first, with McCants, Jaric, Foye, and Telfair:

 

 And here's the second, with Snyder, Brewer, Gomes, and Smith:

 

So as a result, the most positive correlation with winning is actually Kirk Snyder.  Second is McCants, followed by Foye and Smith.  Jaric, Telfair, Brewer and Gomes are all correlated negatively with winning; Jaric the most.

This seems to suggest that the Wolves play McCants, Snyder, Foye and Smith, and avoid playing Jaric, Telfair, Brewer, and Gomes.

 

Take all of this with a grain of salt, because 1) the entire team was bad, especially early 2) it depends on who the 5-man units were up against 3) the R-square values are very low here.

 

A couple of more things to note:

• I didn't bother with Al, because well, he's Al, and because he appears on every 5-man unit except 2.  This would probably lead us to believe that Jefferson is correlated negatively with winning.
• Snyder is helped by the fact that he arrived later on when the rest of the team was doing well
• Brewer is the type of player that won't always look well when analyzing statistics.  His slope is only slightly negative and the R-value is extremely low, so he's probably more even than negative.
• This whole thing may be hogwash.  If anyone has any suggestions on a better method, I'd welcome them.
  

Good lord, I feel like I just walked into math class!!!!!!!  

But very cool stuff Matty, but sometimes these kinds of statistics can get skewed and not always accurate. 

"There are three kinds of lies: lies, damned lies and statistics."
 

-Mark Twain-

Really impressive and interesting to think about.

 Wow.... my head hurts

Ah, you guys, what's the problem with a little math?  You think this is confusing, try reading up on real stat analysis by the pros... in fact I thought I was being much too simplistic in my analysis.

 

Oh, well.  I edited the original to include warnings for all of you faint of heart...
 

MattyP:

Ah, you guys, what's the problem with a little math?  You think this is confusing, try reading up on real stat analysis by the pros... in fact I thought I was being much too simplistic in my analysis.

 

Oh, well.  I edited the original to include warnings for all of you faint of heart...
 

 

I feel like Boxter, my head just exploded...... 

Also I just registered where the die-hards post at APBRMetrics.  When I get a chance I'll run this by them to see if it passes the "smirk" test. 

I did a similar analysis (well different but similar in nature...mine dealt with play pair combinations) that had a nearly identical outcome. 

 

Interesting.  Wait, all of us that actually watch the team understand this more than the coaching staf...

 Nice analysis.  I wouldn't trust the Snyder results any farther than I can throw Snyder, but those weren't surprising.  It clearly shows that a Foye/McCants backcourt is our best backcourt.  Who can argue with that honestly?

Yelo:

 Nice analysis.  I wouldn't trust the Snyder results any farther than I can throw Snyder, but those weren't surprising.  It clearly shows that a Foye/McCants backcourt is our best backcourt.  Who can argue with that honestly?

 

You really can't!!! LOL

But yeah, we have discussed this subject quite a bit in the past, but I never understood how in the HELL Wittman didn't start both Foye and McCants. I mean, come on! By doing that, we would have had a better gauge on what we have in terms of their on court chemistry.  Chalk that issue up for another disappointing part of this season.

Vendetta11:

Yelo:

 Nice analysis.  I wouldn't trust the Snyder results any farther than I can throw Snyder, but those weren't surprising.  It clearly shows that a Foye/McCants backcourt is our best backcourt.  Who can argue with that honestly?

 

You really can't!!! LOL

But yeah, we have discussed this subject quite a bit in the past, but I never understood how in the HELL Wittman didn't start both Foye and McCants. I mean, come on! By doing that, we would have had a better gauge on what we have in terms of their on court chemistry.  Chalk that issue up for another disappointing part of this season.

Exactly. And it would've been more court time for them to learn to play together and develop that chemistry. This is exactly why I think that Wittless failed this year.

As for Snyder, he came over after the team started playing much better (or coincidentally right as they started playing better), so I think his numbers are skewed too. He wasn't around for the first 1/2 of the season when they mustered 6 total wins. That's a perfect case of how stats can be misleading. 

Um, can you repeat the part where you said all the stuff about the......things?

 

Seriously, you lost me about here:  "Warning: This is math-intensive..." 

 

You know, for all of the flak that McCants gets for his defense, the numbers suggest that his offense is too valuable to bench him.  On a team that needs outside shooting, his skills are too important.  I'm all for starting him next season, unless something drastic changes on the roster.

 

And most of us thought/knew that Jaric wasn't all that great, but these numbers (negative correlation with winning percentage) help to support our opinions. 

CC: Wittman: FOR CRIPES SAKE, PLAY MCCANTS AND BENCH JARIC.

 

Another thing that I am seeing in various statistical analyses is that the best hope for the Wolves seems to be in outscoring or out-efficiency-ing their opponent, not stopping them with defense.  This method is best done with McCants out on the floor.

 

MattyP:

You know, for all of the flak that McCants gets for his defense, the numbers suggest that his offense is too valuable to bench him.  On a team that needs outside shooting, his skills are too important.  I'm all for starting him next season, unless something drastic changes on the roster.
 

HELLO????? This is what I've been saying all year! You don't bench Vince Carter because his defense isn't that great. When you have an offensive talent like McCants... you play him! Yeah, you try to coach up his defense in the process, but you don't bench him because he's not great at everything.

If you play Corey Brewer because of his defense, even though his offense was horrendous most of the year... then why wouldn't the reverse be true for McCants. Plus, I'd contend that McCants defense wasn't as bad as Brewers offense.

This coaching staff just won't let him be great at one thing and average at another. They are demanding perfection out of him and it's ridiculous! Let him play. It's been proven time and time again that they are a better team when he is on the floor and especially when he gets major minutes and can really impact the game. What he gives on the offensive end FAR outweighs any deficiencies he might have on the defensive end. And for a team that is trying to develop it's players, you think that playing him, building his confidence, and coaching him up would be exactly what they would want to do.... but not Wittless.

Have I mentioned how much I dislike Wittless? 

MattyP:

You know, for all of the flak that McCants gets for his defense, the numbers suggest that his offense is too valuable to bench him.  On a team that needs outside shooting, his skills are too important.  I'm all for starting him next season, unless something drastic changes on the roster.

 

And most of us thought/knew that Jaric wasn't all that great, but these numbers (negative correlation with winning percentage) help to support our opinions. 

CC: Wittman: FOR CRIPES SAKE, PLAY MCCANTS AND BENCH JARIC.

 

Another thing that I am seeing in various statistical analyses is that the best hope for the Wolves seems to be in outscoring or out-efficiency-ing their opponent, not stopping them with defense.  This method is best done with McCants out on the floor.

 

This is exactly right.  The thing that drives me nuts about it is that its not like looking at different numbers gives you a different answer.  Any way you want to answer the question of who should be playing at SG, the answer always comes back McCants unless your Wittmann or a hater.  Personally I think its obvious that the team is better with Shaddy on the floor, and no statistical analysis is necessary, but in the event they wanted to know all they'd have to do is look at the numbers and they'd have their answer.

 McCants is well above average offensively.  He's a top ten SG in per minute point production.  He's got a lot of room to grow in other aspects of his game, but his scoring ability is valuable enough that he should be getting over 30 a night.
 

Man, I am probably the biggest hater on McCants here. Don't get me wrong, I really believe that he has awesome talent, but he just gets too frustrated so quickly that it not only hurts his game, but the team at times. I do agree however that having Jaric in front of McCants is just absurd.  

Vendetta11:

Man, I am probably the biggest blind hater on McCants here. Don't get me wrong, I really believe that he has awesome talent, but he just gets too frustrated so quickly that it not only hurts his game, but the team at times. I do agree however that having Jaric in front of McCants is just absurd.  

  Fixed that for ya

Vendetta11:

Man, I am probably the biggest hater on McCants here. Don't get me wrong, I really believe that he has awesome talent, but he just gets too frustrated so quickly that it not only hurts his game, but the team at times. I do agree however that having Jaric in front of McCants is just absurd.  

I disagree on this.  I think he's a little pouty and crabby, but so are many of the best players in the league (see T Duncan and K Bryant), so you can't say that hurts him or the team without making the same argument for them.  I think he lacks focus at times, and thats why other parts of his game sometimes suffer, particularly his defense.  He needs that focus to turn into an elite player in the league, but he's good enough to be a role player in this league right now, even a starting role player.  He's got the talent and desire to be a very good player (Ray Allen like ceiling, IMO), but he's got to put it all together to get there.  I think he's already almost as good as Ben Gordon and players like that.

Great job Matty! I actually started teaching statistics last year so I have a pretty good idea of what you're saying. You're right - there are some caveats with this kind of analysis - and you know what they say, correlation is not always causation - but it certainly back up a lot of what we've been saying about this squad. Namely, that we NEED to get rid of Jaric!

LabRat:

Great job Matty! I actually started teaching statistics last year so I have a pretty good idea of what you're saying. You're right - there are some caveats with this kind of analysis - and you know what they say, correlation is not always causation - but it certainly back up a lot of what we've been saying about this squad. Namely, that we NEED to get rid of Jaric!

 

Okay, this a question for Stephen, is there any chance that the Wolves would want to buy out Jaric's contract? It would probably be too expensive, but IMO he is stunting the growth of the team.....

If you are teaching statistics, you'd better have more than a pretty good idea of what he's doing here. It's just a simple regression analysis. But I do have to say that an R-squared value less than about 0.7 is pretty worthless. Some of these values are really, really low. Perhaps you should just go with a simple correlation (r instead of r-squared) between minutes played for each player and the result of that game (win vs. loss). My guess is that there are so many different variables happening in a basketball game that you'll never get a finding of statistical significance with any test you use, but it would be interesting to see.

boognish:

If you are teaching statistics, you'd better have more than a pretty good idea of what he's doing here. It's just a simple regression analysis. But I do have to say that an R-squared value less than about 0.7 is pretty worthless. Some of these values are really, really low. Perhaps you should just go with a simple correlation (r vs r-squared) between minutes played for each player and the result of that game (win vs. loss). My guess is that there are so many different variables happening in a basketball game that you'll never get a finding of statistical significance with any test you use, but it would be interesting to see.

Did you learn statistics in the physical sciences?  .7 is a very high R, much less Rsquare in any type of complex dynamic system.  Usually a lot of variables combine for the total output of any equation, so I'd argue your standards are too high in the context of modelling basketball production, or any other type of human performance measure.  But perhaps you were pointing that out in your last sentence.