ElGee wrote:Well, you're right, but that's some Holy-Grail stuff. There is a way to get there -- but I don't know how to get all the way there on defense.
One thing to do is track things like "correct" rotation and "deterrence." I find these things extremely difficult to quantify -- even moreso than anything I do, which has its own ambiguities at times -- but that would be a way to start quantifying some of what you're talking about. For example:
Duncan needs to rotate off the ball because of a breakdown on the weakside. If he does so in a manner that is better than 99% of centers, he can simultaneously be aware of his man he is leaving and *deter* a pass to the man he is helping on. Other players may not do that and end up with a shot taken against them, adding to the defensive usage. Duncan gets nothing there. (Note: In your example, Duncan should have a very high usage because he team is choosing to funnel toward him.)
So, the question is, how do we account for stuff like that on defense? I find deterrence to be nearly impossible to quantify because of the continuum of physical space in relation to the choice of the offensive player. In English, how close/obstructing does a defender have to be to deter an offensive player from making a pass or taking a shot? It's nearly impossible to gauge without fMRI readouts of the offensive player.
My next generation of defensive usage (which I will post in the near future) will incorporate more help/team concepts that can be quantified. That will (justly) raise DUSG percentages. But it's still nearly impossible to encroach on the maximum usage offensive players. My offensive usage stat has players over 50%. Defensively, even bumping up the number WITH stuff like deterrence, it's still not going to get there. There are simply too many possessions involving screens, fast breaks and fouls. Offenses can force that stuff by going to one or two guys, but a single defender registers 0 usage every time one of those events occur away from him, even with all team concepts and interactions in play. (Basketball, by nature, is a game designed to cater to the offense. It's just the way the rules are set up, and why we see 200 points a night and not 40.)
Btw, you love the +/- family and the standard deviations are always smaller for defense than offense there, suggesting individual offense has a wider-ranged impact. (I see the same thing in my own offense/defense ratings, fwiw).
I definitely look forward to your next iteration and find it very informative, but I'm still not quite ready to come over to the dark side just yet. I'll grant you that the offense gets a certain amount of decision-making advantage over defense, but I'm not sure you're giving full credence to all of the elements of team defense. Perhaps you are, as I said I haven't come down off my fence either way, but I'm not convinced yet.
For example, there are very, very few half-court possessions for which a dominant team defender (especially a big) should truly get a '0' for defensive usage. I don't care if Wade and LeBron just pick each other's man and take jumpers all game against the Magic, Howard still looms large on every defensive possession. In fact, perhaps the reason that (hypothetical) Wade and LeBron would be running those continuous picks-and-pops outside of the lane is expressly BECAUSE Howard is there as a deterrent and that might make an open jumper a higher percentage shot for them than a Howard-contested drive to the rim. Now, that's something that even holy grail couldn't necessarily pin down in a repeatable stat (offense-changed-from-what-they-might-want-to-do?), it would never go in a defensive usage stat, but over the long haul it affects opposing offenses.
In other words, while an offense has the autonomy to initiate the play, a defense can be built in such a way that it forces offensive players to do what they wouldn't ordinarily want to do. And the greater the individual defensive talent, the more things they can take away from the offense. And whether it could show up in a usage-type stat or not, I'm not ready to say that such a defensive impact isn't (in conjunction with the things that are measurable) not just as valuable to a team as an individual's offensive contribution.
You mentioned the +/- stats, and that's a reasonable place to leave this. Because you're right, in my perusal of the different APM calculations I've looked at there is a bit more bang for the buck at the top on offense. But it's not a huge difference. For instance, in the B-R win shares calculation offensive win shares are counted about twice as much as defensive win shares with the thought that individual offense is that much more important. But that's not what I see in the APM calculations.
In the 2005 - 10 calculation I've mentioned a few times lately, an arbitrary cut-off of +2 seems to capture the better offensive and defensive threats of the past 5 years reasonably well. On offense there are 37 folks above that threshold, on defense there are 54 (though probably only around 30 or so that I'd classify as playing starters minutes). Now, the offensive players in this range rank pretty continuously from +2 to about 4, then you start seeing gaps between varying outliers (Dirk at 4.3, jump to LeBron/Billups/Kobe around 5, then jump to Wade at 5.9 then huge jump to Nash by himself at 8.05). On defense it also ranks continuously from 2 until about 4, then there's only 1 outlier at +6.3. On the flip side there are more offensive players below -2 (58 vs 21), but the extreme values again are similar (-3.9 on offense, -3.3 on defense).
My point? As you mention, the standard deviation on offense (1.29) is slightly larger than defense (1.14), and there is a slight difference in the best-of-best and worst-of-worst in each category, slightly in favor of the offense. So looking just at the APMs, I could imagine you could convince me that in general individual offense might be
slightly more important than individual defense. But I'm talking slightly, and only as a trend since the defensive outlier was having a similar impact on his end of the court as the offensive outliers. Nothing like the 2-to-1 that win shares would suggest, or even the roughly 60% difference you report among the top-5 usage leaders in last year's playoffs. If there is a slant towards individual offense, I think that degree is way too large.