So why did Pero not play last night?

[Admin]

It didn't look like anything was wrong with him and I didn't hear them address it during the game. 

[IheartVolt]

Well from the response I received he apparently required an oil change

[Admin]

Well from the response I received he apparently required an oil change

 

Yeah I saw that and it was funny until nobody else had a clue about the real reason, hence creating a thread just for it. 

[ersimo2889]

Maybe Coach Bud pulled a "DNP - Old" like Pop used to? Regardless, it was a good opportunity to get Elton some minutes for the first time in a week.

 

[Admin]

Rest.

Also, the Hawks have Dan Rosenbaum who was/is big on plus minus. There's a statistical problem with plus minus measurement if you keep a constant rotation throughout the year. So by having a somewhat random disbursement of rest and playing time you can help understand the value of players.

 

Oh good grief I hope it's just for rest and not some silly +/- stat. 

 

And I don't know how I feel about DNP-Old for guys who are younger than me!! 

[Admin]

How is it silly? Seems like the team sacrifices maybe a 1% chance in winning for a better understanding of how valuable your players are. That understanding probably raises the Hawks chances of winning in the future (playoffs).

 

Because I think it's a dumb stat. And I don't know you arrive at the maybe 1% chance number. We damn near lost last night as Scott and Brand combined to go 2-10 and I believe Pero could have helped there and on defense. 

[Plainview1981]

Because I think it's a dumb stat. And I don't know you arrive at the maybe 1% chance number. We damn near lost last night as Scott and Brand combined to go 2-10 and I believe Pero could have helped there and on defense. 

 

I blame Scott. The dude contributes nothing at this point. 

[Admin]

I blame Scott. The dude contributes nothing at this point. 

 

I think he was 2-7 and he hit that 1st three and it looked like he was going to have a big night. Then he fell flat. 

[Admin]

OK {/sarcasm}

 

I know you won't be able to discuss the stat without getting angry so I'll take the sarcastic OK.  That's just the one stat that really bugs me as I just don't believe you can simply rate players like that. If nothing else changed beside the player being on the court vs off then I would agree with it but there's too many variables for it to mean anything to me. 

[Colin]

I blame Scott. The dude contributes nothing at this point. 

 

I still like Mike Scott. Obviously he is a chucker, but when he is on he is a top-tier bench floor spacer. 7.3 PPG on 45/39/83 (in only 14 minutes is definitely contributing more than nothing...

[MaceCase]

Go back and retype that, you're missing more than a few s

[Admin]

I didn't misunderstand what you're saying. I'm saying there's already enough variance in +/- without the need for adding more variables. Especially with all the combinations of our lineups that we run. Im not claiming to be on your level with stars but when you're talking about a star that measures the impact of a player on the court vs off that seems simple to me that it introduces a bunch of variables built into that assumption.

[Guest]

All this because Pero had his maintenance done last night, lol.

[Guest]

I didn't misunderstand what you're saying. I'm saying there's already enough variance in +/- without the need for adding more variables. Especially with all the combinations of our lineups that we run. Im not claiming to be on your level with stars but when you're talking about a star that measures the impact of a player on the court vs off that seems simple to me that it introduces a bunch of variables built into that assumption.

Your previous post, I thought you understood. This post makes me concerned that you don't because you've flipped on the understanding of +/-. There is a big problem with the +/- and it is precisely because there is not enough variance in the lineups. So when you go through and try to grasp an estimate of the value for a particular player, the errors associated with that estimate are quite large and you lose precision in your estimate. A big part of this is the multicollinearity (or co-linearity as discussed here: http://www.goldenstateofmind.com/2011/12/6/2602153/advanced-stats-primer) that comes from your lineups not having enough variation in them. For example, if Jeffy only played with Kyle then you have a problem separating Jeffy's value from Kyle. Same story works if Jeffy and Dennis are only substituted for themselves. You've talked about this, and it's true that there is a problem.

I think (not sure) that your understanding of my claim that Pero is (pseudo-)randomly rested is that this adds on another variable to your estimation of the true value of a player. This isn't right, it actually works in the other direction. By (pseudo-)randomly resting players throughout the year, you are introducing variation but not a variable that you need to control for because resting a particular player is not correlated with anything else. If it was the case that resting the player was correlated with, I don't know maybe having recently been blown out or playing a game the previous night, then this would be a concern. But as far as I can tell by looking at the rest/play decision of Pero and Brand in particular, sometimes they are rested on the first of a back-to-back and sometimes it's the second. There does not appear to be a particular reasoning for the play/rest decision.*

I think this translates over to the Spurs as well. Popovich does not appear to have a particular pattern of play/rest for Duncan. If they go on a back-to-back, then sometimes Duncan plays the first but not the second. Other times it can be resting the first and playing the second. Sometimes Duncan appears to rest with Ginobili, then other times one plays and the other rests.

The idea of (pseudo-)randomly resting players helps add in variance to +/- lineups (because now you have a larger # of combinations) which gives you a more precise estimate, not a less precise or confounding effect. It seems that you flipped on this issue. You were right before, but not right in this post.

*even if the decision is not (pseudo-)random, if the Hawks know the reason for why someone plays/rest and this is quantifiable then you can exploit this in a beneficial manner. This would be a case where you are adding in an additional variable to your model of +/-, but since you control for this factor you have increased the variance in +/- lineups which helps out your precision in estimation. The added variable does not harm estimation so long as it is quantifiable...this is the more statistically complicated reasoning/answer that I'll give. Yes, I'm a douche and a nerd.

[ATLien_]

Your previous post, I thought you understood. This post makes me concerned that you don't because you've flipped on the understanding of +/-. There is a big problem with the +/- and it is precisely because there is not enough variance in the lineups. So when you go through and try to grasp an estimate of the value for a particular player, the errors associated with that estimate are quite large and you lose precision in your estimate. A big part of this is the multicollinearity (or co-linearity as discussed here: http://www.goldenstateofmind.com/2011/12/6/2602153/advanced-stats-primer) that comes from your lineups not having enough variation in them. For example, if Jeffy only played with Kyle then you have a problem separating Jeffy's value from Kyle. Same story works if Jeffy and Dennis are only substituted for themselves. You've talked about this, and it's true that there is a problem.

I think (not sure) that your understanding of my claim that Pero is (pseudo-)randomly rested is that this adds on another variable to your estimation of the true value of a player. This isn't right, it actually works in the other direction. By (pseudo-)randomly resting players throughout the year, you are introducing variation but not a variable that you need to control for because resting a particular player is not correlated with anything else. If it was the case that resting the player was correlated with, I don't know maybe having recently been blown out or playing a game the previous night, then this would be a concern. But as far as I can tell by looking at the rest/play decision of Pero and Brand in particular, sometimes they are rested on the first of a back-to-back and sometimes it's the second. There does not appear to be a particular reasoning for the play/rest decision.*

I think this translates over to the Spurs as well. Popovich does not appear to have a particular pattern of play/rest for Duncan. If they go on a back-to-back, then sometimes Duncan plays the first but not the second. Other times it can be resting the first and playing the second. Sometimes Duncan appears to rest with Ginobili, then other times one plays and the other rests.

The idea of (pseudo-)randomly resting players helps add in variance to +/- lineups (because now you have a larger # of combinations) which gives you a more precise estimate, not a less precise or confounding effect. It seems that you flipped on this issue. You were right before, but not right in this post.

*even if the decision is not (pseudo-)random, if the Hawks know the reason for why someone plays/rest and this is quantifiable then you can exploit this in a beneficial manner. This would be a case where you are adding in an additional variable to your model of +/-, but since you control for this factor you have increased the variance in +/- lineups which helps out your precision in estimation. The added variable does not harm estimation so long as it is quantifiable...this is the more statistically complicated reasoning/answer that I'll give. Yes, I'm a douche and a nerd.

Hawksfanatic are you a data scientist by chance? Your understanding of statistics, data and data bias is really impressive. I could use someone like you to work on some business intelligence projects.

The fact that Hawks are potentially doing this on purpose shows they truly understand these metrics. They're not just eyeballing them, they are using them to create a competitive advantage.

I would imagine they are studying other teams too. Coupled with the knowledge of how coaches react to certain Hawks lineups, you could strategically influence your opponent to play an unfavorable lineup a disproportionate amount of time.

Edited January 12, 2015 by ATLien_

[macdaddy]

As much as I want to win every game, I think we have to start letting some guys have the night off here and there.   Seems like we had a light November and now we're playing 5 times a week.   

[Guest]

Hawksfanatic are you a data scientist by chance? Your understanding of statistics, data and data bias is really impressive. I could use someone like you to work on some business intelligence projects.

The fact that Hawks are potentially doing this on purpose shows they truly understand these metrics. They're not just eyeballing them, they are using them to create a competitive advantage.

I would imagine they are studying other teams too. Coupled with the knowledge of how coaches react to certain Hawks lineups, you could strategically influence your opponent to play an unfavorable lineup a disproportionate amount of time.

I wouldn't call myself a data scientist, but that's a probably an accurate description.

Now I don't know if the Hawks are resting Pero because of the ideas I have mentioned. They could have other reasons. But I have a hunch that the Hawks are thinking about similar issues to what I've described. Especially if Rosenbaum has continued to work with +/-.

[Guest]

As much as I want to win every game, I think we have to start letting some guys have the night off here and there. Seems like we had a light November and now we're playing 5 times a week.

Our starters are averaging low 30's in mpg, if they were in the high 30's I would be concerned. I think it's too early to start resting players we're not even at the midway point of the season yet.

Al and Sap still needs to work on their cohesion, etc.

Besides when you are blowing out teams, the starters are getting quite a rest, lol. These are the minutes for our starters last game: 28, 32, 29, 25, 30.

Edited January 12, 2015 by JayBirdHawk

[Duff_Man]

Our starters are averaging low 30's in mpg, if they were in the high 30's I would be concerned. I think it's too early to start resting players we're not even at the midway point of the season yet.

Al and Sap still needs to work on their cohesion, etc.

Besides when you are blowing out teams, the starters are getting quite a rest, lol. These are the minutes for our starters last game: 28, 32, 29, 25, 30.

 

Pero is 32 and takes a beating every night. If Bud wants to give him a night off every 2 weeks, No worries.

[macdaddy]

Our starters are averaging low 30's in mpg, if they were in the high 30's I would be concerned. I think it's too early to start resting players we're not even at the midway point of the season yet.

Al and Sap still needs to work on their cohesion, etc.

Besides when you are blowing out teams, the starters are getting quite a rest, lol. These are the minutes for our starters last game: 28, 32, 29, 25, 30.

 

I hear you but even with similar averages last year we hit injuries to MIllsap, Antić, DMC and I think even TEague and Korver were out some in Jan/February.   Not that they were worn down or anything but you never know.    To me a night off is a world different than just less minutes.   

 

I agree with you though.  Bud is fantastic about managing minutes.  Last I checked MIllsap led the team at around 32 which is like 30th in the league.