RealGM

Much has been said recently about Larry Sander's contract vindicating some of the statistical analysis about his interior defensive capabilities which don't necessarily translate into the normal stats.

http://www.sloansportsconference.com/wp ... %20NBA.pdf

While I generally agree with Goldsberry's premise, an analysis that seems to imply Bargnani is the #2 interior defender is questionable (see top of page 6).

Nanogeek wrote:Much has been said recently about Larry Sander's contract vindicating some of the statistical analysis about his interior defensive capabilities which don't necessarily translate into the normal stats.

http://www.sloansportsconference.com/wp ... %20NBA.pdf

While I generally agree with Goldsberry's premise, an analysis that seems to imply Bargnani is the #2 interior defender is questionable (see top of page 6).

2 possible explanations (at least my hypotheses):
1. While he seems to have a low opponent fg percentage, if you see appendix 2, he's not w/in 5 feet of the shooter as often as Sanders, which could both be negative in his defensive positioning and also skew his opp. shot data

2. About 1/2 of his minutes in this time span were at PF, whereas around 25 percent of Sander's were. Just the fact that Bargs would be a 7 foot PF would mean he's going against potentially much smaller players, which could also change shots without actually meaning he's a better defender.

Don't know if either are true, but thoughts.

Nanogeek wrote:Much has been said recently about Larry Sander's contract vindicating some of the statistical analysis about his interior defensive capabilities which don't necessarily translate into the normal stats.

http://www.sloansportsconference.com/wp ... %20NBA.pdf

While I generally agree with Goldsberry's premise, an analysis that seems to imply Bargnani is the #2 interior defender is questionable (see top of page 6).

I've found this study and how it's reported confusing as well. Varajeo is almost always described as a good defender and he is worse than Lee here, while there is little to no difference between Lee and Kevin Love. Not that Lee is good, but not sure why he gets all the flak. Asik, Noah and Chandler for example, are all middle of the pack (all in reference to the final table).

I can think of several possible "problems" with the analysis. Biggest one being that the real defensive skill is getting close to the offensive player, and that whether the offensive player makes or misses the shot is mostly random. In other words, challenging shots is the skill, but there's no significant and consistent differentiation between guys when it comes to actually making them miss. I'd have to re-read to be sure, but maybe someone knows -- has Goldsberry looked at year-to-year consistency in this data? Has he looked at statistical significance between defender results?

Nivek wrote:I can think of several possible "problems" with the analysis. Biggest one being that the real defensive skill is getting close to the offensive player, and that whether the offensive player makes or misses the shot is mostly random. In other words, challenging shots is the skill, but there's no significant and consistent differentiation between guys when it comes to actually making them miss. I'd have to re-read to be sure, but maybe someone knows -- has Goldsberry looked at year-to-year consistency in this data? Has he looked at statistical significance between defender results?

Good points. He does address the issue of getting near the player, I believe that is the reasoning behind partitioning the results by distance from the defender when the shot is released, 1' vs 3' vs 5'. I'm not so sure that simply the distance accounts for everything though, as you say, as there are various reasons why the distiance is what it is. It doesn't account for the difference in, say, biting on a pump fake and being left nowhere near a play giving up a hoop, vs a guy who is contesting his own man's dunk attempt after stopping a dishing penetrator.

There was two years of data, the last two years since they were using the SportVu cameras, that's all they've got. But I didn't see any mention of consistency between years for individuals, so it is possible there is a significant change from season to season.

Though with cameras going into all the arenas this year, they should double the data set this year alone, so it will be interesting to see what comes out of it.