mysticbb wrote:The issue is that in the 2nd case, the Timberwolves wouldn't have used Garnett against Duncan, like the Mavericks did not use Nowitzki for the majority of the possessions against Duncan. It is more likely that Joe Smith works on Duncan while Garnett switches to Robinson, which then makes it less likely that Duncan goes completely off like he did in 2003. Or the Timberwolves might even do what the Mavericks did for extended stretches, using Garnett as the SF while playing two big guys next to him. The issue is, the game dynamic would shift, which is essentially what the Mavericks had as an advantage with Nowitzki. You are basically eliminating that advantage, thus making a similar thing as other who are just looking at the worse scoring performance by Garnett in order to determine the value of the player.
The problems gets best illustrated by your assumption that the team would then lose by 10 points more, which is completely baseless.
I agree with your 2nd point, but for the Nowitzki game we should mention that the Mavericks were +10 with him on the court, while -7 in the 4min without him. And especially in the last couple of possessions Nowitzki did a pretty good job defensively against Duncan.
You are combining 2 things in this post, one of which I think is a really good area for more follow-up discussion and one that I was trying to get away from a bit as it muddied the waters. In the post that you quoted, I was purposefully no longer considering 2003 Nowitzki expressly because a one-game snapshot from Dirk's career isn't enough to characterize him one way or the other. Instead, I'm using the numbers in this example to get at the concept of whether a KG with high scoring volume/efficiency but poor defense/less ability to distribute would have been more effective or less effective to the team's results. So the points you make here defending Dirk or going into more detail about 2003 are actually shots fired on a battlefield that I already stepped off of, and I have no real rebuttal to make there.
On the other hand, I think you raise some interesting points regarding how the team dynamic might change were Garnett in super-scoring/worse-everything-else mode, and I'd love it if all the interested parties weighed in on this. Because this didn't happen in real life, we're forced to some level of informed conjecture about what the outcome would have been on the court. However, the entire purpose of the PC board is essentially informed conjecture about how we think one player compares vs another if their circumstances were hypothetically even. And if we play out the KG-that-was vs hypothetical-KG experiment from my post that you quoted, I think we can still make reasonable estimates of what the outcome might be based on our knowledge about the players involved (e.g. I don't think the speculation that the team loses by more than 10 points is at all baseless). So let's look at it, starting again with your quote:
"The issue is that in the 2nd case, the Timberwolves wouldn't have used Garnett against Duncan, like the Mavericks did not use Nowitzki for the majority of the possessions against Duncan. It is more likely that Joe Smith works on Duncan while Garnett switches to Robinson, which then makes it less likely that Duncan goes completely off like he did in 2003. Or the Timberwolves might even do what the Mavericks did for extended stretches, using Garnett as the SF while playing two big guys next to him. The issue is, the game dynamic would shift, which is essentially what the Mavericks had as an advantage with Nowitzki. "I don't think the assumptions you make here fit with the Wolves personnel. In real life Joe Smith (30 mpg), DEAN GARRETT!!! (23 mpg), rookie Rasho (10 mpg), and 35-year-old Sam Mitchell (33 MPG) essentially covered all of the non-KG frontcourt minutes. None of them were particularly strong on defense at that point in their careers...Smith was reasonable, but by no means even approaching All-D and he was too light in the tail to be very effective against 7-foot post-up centers. The hypothetical set up is that Garnett traded in his great defense (individual and team) as part of the package to become a dominant scorer.
Thus, I see no basis that any combination of that frontline was going to be a major deterrent to Duncan going off. In 3 of the next 9 post-Wolves games that postseason Duncan posted scoring lines of 37 points on 64% TS, 33 points on 82% TS, and 33 points on 65% TS. He definitely had the explosive scoring capability in 1999, and those bursts happened reasonably regularly post-Wolves series. So if Duncan was on against the Wolves frontline that no longer has arguably the best defender at his position of all-time I see no reason that he couldn't reasonably produce the statline that I suggested in the hypothetical.
The Dirk advantage that you mention here, I believe, is that he could shift to SF or allow better defenders to get more minutes as needed. But KG-that-was in '99 already was playing several minutes per game of SF, and there WERE no better defenders on that Wolves frontline to get more minutes as needed. As such, I think it very reasonable to project that on that '99 Wolves team, mega-scoring/average-D-non-facilitating KG would have given up more points at the defensive end to an elite Bigs combo of Duncan and Robinson than he scored (without even going too far into the effect of the loss of passing/facilitation ability).
mysticbb wrote:drza wrote:In the '99 playoffs Duncan averaged 18.8 points on 51.7% TS against KG
I'm not disputing an defensive effect by Garnett on Duncan's performance, but those 52 TS% actually are in agreement with my statement, that the 42 TS% in that particular game was rather the lower end, and thus variance played a role. The same goes for Duncan in 2003 in that game 1, because Duncan was not a 70 TS% scorer in average.
I know that we both know what variance means. I don't pretend that everyone reading this thread and hopefully getting some use out of it has the same grasp. And I think the way you're trying to make this point here may fit the letter of the (statistical) law, while somewhat obscuring the main idea. Another example of missing the forest for the trees by zooming in to far.
Because the point that I was making, using this one game (real and hypothetical) as an example, is that the odds were much greater that Garnett-that-was would suppress Duncan's scoring (volume AND efficiency) as opposed to the offensive-no-defense Garnett of hypothetical. My point was never that Garnett would hold Duncan exactly to 18 points on 42% TS in every game, any more than I would expect hyper-scoring-Garnett to score 38 points on 70% TS in every game. Those were both examples to put numbers to the concepts that I have a) explained qualitatively, b) explained in amalgum using info from KG's entire postseason career from 1999 - 2008, c) explained in more detail by highlighting five of the worst shooting postseason games of KG's career, and now d) have been trying to make clearer with a single game example.
And if the mean 18.8 points on 51.7% TS that Duncan averaged in 1999 against Garnett-that-was are taken to be the real expected values for Duncan, then a 16 point performance on 42% TS would be much, much closer to the expected value than a 35+ point/65% TS effort. Both are possible outcomes due to variance, but if you plot scoring volume vs scoring efficiency over the 100 trials that you mention in your later post, the 16-point/42% effort is going to be right near (if not within) the main cluster of data points whereas the high-30s/65% explosions are going to be far into the outlying regions...with Garnett that-actually-was.
If you ran that same experiment with no-D KG and the Timberwolves '99 frontline (again Joe Smith, DEAN GARRET!!!, 35-year old Sam Mitchell, rookie Rasho) the entire mean in both scoring volume and efficiency goes way up, the 16-point/42% possibility for Duncan becomes exceedingly unlikely of ever occurring (in '99 post KG, Duncan only had 1 game out of 13 with as few as 16 points OR a TS% as low as 42%, and they both happened in the same game, a blowout win in which he played only 20 minutes and took only 7 shots). On the other hand, the kind of scoring explosion that would have been an extreme outlier against real Garnett becomes a much more likely event with no-D KG (in '99 post KG, Duncan had 4 games out of 13 of 30+ points AND a TS% of 60% as part of his overall averages of 24.6 points and 59% TS).
Yes, variance allows for the possibility of Duncan having scoring outputs ranging from getting shut down to exploding under any circumstances. But with Garnett-that-was, the odds were MUCH better for the upper teens/below par shooting outcome and for No-D-KG the odds would be MUCH better for the Duncan scoring explosion. And I know you understand that. But if you just make the point as you did and chalk up the numbers I gave as "variance" as though both were equally likely under the scenario, I think you mislead the readers even if you're statistically legal to say it.
