Continued Thoughts on Rim Protection in Isolation (more numbers)
It's been two weeks since I first used SportVU data to try and isolate "rim protection value" provided by various NBA bigs. I want to revisit this periodically to see how it holds up over the course of the season. Has anyone improved or regressed greatly, and do these changes mean anything? Certainly teams have gone on streaks, which usually entails "an improved commitment to defense", or so we're usually told.
A quick refresher on the metric - basically I took the SportVU defensive impact numbers from NBA.com, and calculated the points "saved" by each player in comparing the shots contested and percentage of those shots made with the approximate FG% on uncontested shots of < 5 feet. However, this points "saved" is relatively meaningless in a vacuum, so I went a little further and found the average percentage on contested shots as well as the average number of contests per 36 minutes for NBA big men. I then compared the raw "points saved" with the expected number for a league average big playing the same number of minutes as the player in question.
I realized I was probably undervaluing shot prevention, and probably overvaluing players on bad defensive teams who get more opportunities to "save" points due to the high number of shots at the rim allowed (based in large part on how well Chris Kaman scored, though even after a team adjustment, Kaman does quite well because of his combination of high contest rate and low FG% on shots he contests). So I added in a team adjustment based on the number of < 5 ft FGA/GM allowed by each team as compared with league average. Finally, please note all the numbers in this post are through the Christmas day games but do not have Wed-Sat's games yet, so keep that in mind.
For this iteration, I expanded the parameters slightly to include more players both because I was curious how some more backups rated (for example, Joel Freeland is obviously worse than Robin Lopez in this department, but I wanted to get a sense of how much worse), and to make sure I wasn't biasing the "average" against which players are being measured by including only starters, so I looked at players who have played at least 15 games and contested at least 4 shots per game at the rim. This left me with 85 players, but I eliminated Shawn Marion as he has played the majority of his minutes at SF (I probably SHOULD eliminate Josh Smith as well, but I didn't intend to make it a hard and fast rule, not to mention Smith is contesting 7.2 shots per game at the rim, so he's still defending the paint plenty).
The net main change is that the average "points saved" per 46 dropped slightly from just under 4.5 to around 4.2 pts/36. Without deep diving into it, hard to say whether this is because the league just made a few more shots at the rim as a whole, or the inclusion of the extra 20 some bench (and thus presumably lower skill) players brought the average down some. These are all relative to a league average of contesting around 8.5 shots per 36, leading to a savings of about 4.2 points/36 over what would be expected with no rim protection whatsoever. So, here are the updated rankings (team adjusted) in terms of points saved over the expectation of that "average NBA big's" per game, per 36 and total points saved:
As a reminder the team adjustment is meant to "penalize" players who through pace and/or team defensive frailties have more opportunities to contest shots at the rim, allowing them to "save" points on rim attempts a better defensive and/or slower paced team would not allow.
Looking at the per game list, the first thing to note is all 3 Timberwolves rotation bigs are in the bottom 12, allowing a total of 5.29 points per game above what would be expected with league average bigs in terms of rim defending. It is unsurprising, then that the Wolves continue to allow the highest percentage in the league at the rim.
On the other end, it shouldn't be a shock to see Indiana's historic defense at the top of this category, and the effectiveness of Ian Mahinmi in terms of providing a quality replacement for Roy Hibbert shows that it's not just Big Roy who does the dirty work.
Other points of interest include the wide disparity between Timofey Mozgov and J.J. Hickson in Detroit, with the latter giving up almost 4pts/36 more at the rim than the former, which could go a ways towards explaining why Denver routinely falls behind before the 6 minute timeout in the first quarter.
I also wanted to briefly look at the degree to which the rankings held steady, below I've listed the ten most improved as well as biggest drops in terms of per/36. (Note these are non-adjusted as I didn't want changes in team adjustment possibly caused by rotation changes to confuse the issue):
On the other end, it shouldn't be a shock to see Indiana's historic defense at the top of this category, and the effectiveness of Ian Mahinmi in terms of providing a quality replacement for Roy Hibbert shows that it's not just Big Roy who does the dirty work.
Other points of interest include the wide disparity between Timofey Mozgov and J.J. Hickson in Detroit, with the latter giving up almost 4pts/36 more at the rim than the former, which could go a ways towards explaining why Denver routinely falls behind before the 6 minute timeout in the first quarter.
I also wanted to briefly look at the degree to which the rankings held steady, below I've listed the ten most improved as well as biggest drops in terms of per/36. (Note these are non-adjusted as I didn't want changes in team adjustment possibly caused by rotation changes to confuse the issue):
| Ten Largest Improvements in SOA/36
| Ten Largest Drops in SOA/36
|
The "gains" appear to outweigh the losses, simply because of the drop in the "average" against which players were being measured. If I had to guess, I'd say the inclusion of more backups is the likely culprit. Regardless, the absolute size of the change matters less than the change relative to the rest of the league. Obviously, this is a 5-7 game sample for most players, so it's hard to put too much value in the small sample size, but I want to take one particular example.
Up until their controversial Christmas day loss, the Clippers had been on something of a role. Part of the narrative in the media had been the improved play, especially defensively, of DeAndre Jordan. Backing the numbers out, over the Clippers first 24 games, DAJ had contested 8.5 shots per game, allowing 56.4% shooting on those shots. Over the next 6, his contests jumped to 11.5 and opponents' dropped to 47.83%. Now it's nearly impossible to say how much of this improvement was good luck over the course of the sample and how much was actual improved rim defense from Jordan, but the fact remains that there was almost a point and a half difference per game.
Some of the other changes match existing eyetest and/or narrative developments - Valanciunas has been much better since the Gay trade, apparently at both ends. Sam Dalembert has lost time in Dallas as his effectiveness decreases (he appears to have very little lift at the moment). Noah is wearing out under the crushing oppression of Thib's tyrannical allotment of minutes.
In the near future, I plan to take a qualitative look at rim protection, as a number of people have been discussing the difference between selling out for shot blocks and focusing on preventing interior scoring more holistically. My understanding is that SportVU has yet to incorporate the vertical axis, so it will be hard to quantify, but like obscenity, I think we know "bad" contests when we seen them.
Some of the other changes match existing eyetest and/or narrative developments - Valanciunas has been much better since the Gay trade, apparently at both ends. Sam Dalembert has lost time in Dallas as his effectiveness decreases (he appears to have very little lift at the moment). Noah is wearing out under the crushing oppression of Thib's tyrannical allotment of minutes.
In the near future, I plan to take a qualitative look at rim protection, as a number of people have been discussing the difference between selling out for shot blocks and focusing on preventing interior scoring more holistically. My understanding is that SportVU has yet to incorporate the vertical axis, so it will be hard to quantify, but like obscenity, I think we know "bad" contests when we seen them.
RSS Feed