Doctor MJ wrote:To your point here: I'm actually rejecting your starting point. I'm saying it's leading you to err on one side that leaves any defense of your opinion in the face of incomplete data easily rebutted. Now, perhaps that's not the concern for you that it is for me and you want to continue in that, but I come to the opinions I have by playing internal chess against myself.
Ok, what is there to reject?
The Spurs have a dominant defense? Check.
Tim Duncan was the anchor of that defense? Check.
The defense is at its best when Duncan is still in his prime? Check.
The defense declines right when Duncan declines? Check.
Doctor MJ wrote:Well, it's '03 & '04 a lot , and I guess that doesn't sound like much. Perhaps I make too much of it because of my historical background here. +/- stats came out and Garnett was WAY ahead of Duncan in on/off at the exact time where the Duncan vs Garnett debate was really the debate for best player of the game.
At the time, the only defense for Duncan was the ceiling effect you describe, and then Minny fell off and the consensus swung to Duncan never to really get doubted again. Some of us though have long memories, and when Garnett shattered through the supposed ceiling on the Celtics, it made us completely re-evaluate majority opinion. If Garnett could slaughter Duncan by that metric on slightly weaker team, and he could keep it up on a great team, then how important is this ceiling effect anyway?
Of course there are other factors here:
-There is a clear potential ceiling effect on on/off data, but in APM, it's not so clear cut. And in APM, Garnett's lead remained. Ilardi came out with his 6-year APM estimates, and Garnett was totally off the charts.
-"The ceiling effect" ties directly into the "portability" phrase that ElGee's doing a great job of introducing into the discussion now, and of course when we see Garnett's ability to adapt his game ridiculously in Boston, it's pretty clear that Garnett's portability is far better than most stars, and hence a ceiling isn't going to apply to him that much.
I'll take a moment to touch on RAPM, and that it is the state of the art, and while it typically favors Garnett (not just the non-Minny nadir years, but in multi-year studies), it's not as extreme as APM or on/off. That's probably the thing to go by more than the other stuff, but I will say RAPM is systematically design to be skeptical of outlier data - the really out there stuff that's unlikely to repeat itself. Well, the stuff that happened to Garnett's teammates, that was one of the most amazing runs of "unlikely to repeat itself" stuff I've seen.
RAPM may have been correct in essentially labeling things as luck, but in the court of public opinion, they weren't being labeled as luck, or as a positive for Garnett. They were being used against Garnett unknowingly to end the Duncan vs Garnett debate prematurely.
Ok, nothing to really disagree with here. Garnett has had a big lead in APM and on/off (probably because he was a superstar on a bad team, so he had more lift to give). I have a problem with the usefulness of those statistics, and with RAPM, Duncan has the advantage. It seems pretty selective to use only 03, 04, and 08 onwards for Garnett and just chalk up 05-07 as the "dark days". Not to mention that 05 wasn't even that bad of a year. And that Garnett only had small advantages in 03 (with a worse team) and 08.
And I'm all for giving Garnett plenty of credit for maintaining huge impact with a good team like he has in Boston...but hasn't Duncan done the same thing in San Antonio? Like in 07 for example? Using RAPM, Duncan in 07 was +1.5 over the 2nd best player, compared to Garnett who was +1.6 over the 2nd best player in 08. Very comparable impact, with two very good teams. Garnett gets a lot of credit for not having a ceiling in terms of his impact, but Duncan doesn't get the same credit?
Doctor MJ wrote:Oh wow, you're really struggling with the logic here. Sorry that sounds condescending, but you're struggling. I'm having trouble thinking of how I can really get at the root of it for you so I'll just stay literal:
Productivity and Offensive RAPM are not measuring the same thing. These are not meters and feet where you can simply convert from one to the other. Hence, you cannot say "Well productivity says the offense is alright, so by definition whether RAPM says about offense is officially 'alright', and I'll just judge defense by using Offensive vs Defensive RAPM proportions.".
Productivity, to the extent it is measuring offense, in 2003 it's telling you that Duncan is the best in the league. Offensive RAPM on the other hand is telling you that Duncan is 34th in the league. Right there, you have to stop and recognize that the two metrics are disagreeing with each other VERY strongly. They can't both be right, so what you need to apportion credibility. Either you side with one, or the other, or you keep using both but with not a ton of confidence on their own.
You're still using both, so that means you should be acting cautious with them. Instead, you're using them as a foundation for your analysis as if they are allies. That's a problem. It makes it so that, for example, I can rebut your arguments before you even get started. I can't imagine that's what you want.
My bad that I didn't get the logic, but it seems to me that you're not exactly understanding what the SPM is telling us. As far as I know, SPM is a box-score based stat and assigns a value based on the player's box score production. Not only do stats like rebounds and blocks (which don't have much to do with offense and are two of Duncan's strongest areas in terms of box score production) affect SPM, it's also very possible to have the best production in the league but not be the best offensive player in the league. All SPM is telling us is that Duncan in those years that I described has the best box-score stats in the league, aka "production". This is supported by other box-score metrics such as PER and WS/48, when Duncan was 2nd in both in 02, and 4th/3rd in both in 03.
What the SPM basically tells us is: "Duncan is a decent volume scorer with very good efficiency. He keeps his TOs down, and he gets a decent amount of assists. He's an excellent rebounder and shot blocker. All in all, nobody is giving the same box score production that he is, so he's going to rank at the top of the list". Although the box score is biased towards offense, that doesn't translate to "This is essentially Duncan's offensive value", which we would then hope that RAPM agrees with, and if it doesn't, that exposes some flaw. IMO, they tell us separate things. To me, SPM is a quicker way to just compare box-score stat lines against each other, like is very common. What is also very common is to back the stat lines up with RAPM data, or any kind of impact data, to see how well those box score stats are helping the team. That's all I'm really doing.
With Duncan, I see elite box score production (SPM/PER/whatever)...as in, NOT offense, but simply box-score production. Then I go to RAPM or on/off or with/without to check his impact, and I see someone having monster impact. I really don't see the disconnect here.
