Like most of us, I have viewed the SportVU Stats with interest, but also skepticism. There is no public history for these metrics and hence little understanding of how much value they add. Though not about SportVU, I introduced a basic concept a couple of weeks ago; correlate various measures with RAPM and see what “why” items correspond with the “what” of RAPM (hat tip to @HickoryHigh for the “what” and “why” discussion).
As I initiated that, the SportVu stats were a big part of the interest. Sure, it is only one season, so sample size is not ideal…but people need to start figuring out what these numbers are good for sometime. Right? Today’s article begins by looking at the Defensive Impact stats at NBA.com, particularly the new ones:
- Opponent Field Goal Makes at Rim per Game – Basically any made shot by the opponent when the particular defender was five or less feet from the basket, and within five feet of the offensive player.
- Opponent Field Goal Attempts at Rim per Game – Any shot attempt by the opponent when the particular defender was five or less feet from the basket, and within five feet of the offensive player.
- Opponent Field Goal Percentage at Rim – As would be expected, the first bullet divided by the second (referred to as “Opponent FG%” below).
This article’s intent is not to quibble over the potential flaws in each of these measures. I am sure that has already been done. The goal is to gauge how useful any of this information is. To do that, each metric was correlated against Defensive RAPM from 2014, for a sample of big men that have played over 500 minutes, as well as separated groups of “power forwards” and “centers”. For reference, I also added the more traditional advanced stat of “block rate”.
Regardless of how the sample is sliced, certainly for this one season, “Opponent FG% (at Rim)” bests block rate as an indicator of big man defensive performance; chalk one up for SportVU. Anthony Davis is an example of player performing exceptionally according to block rate (1st of the 133 players), but he ranks 48th of the group for “Opponent FG%”, while also only contesting the 36th most shots per game; comprising part of a package that is slightly below average according to defensive RAPM when compared to the entire set of big men.
A suprising result is that for power forwards, basket protection has relative importance towards strong defensive RAPM when looking at “Opponent FG%”, but has near zero correlation according to block rate. Interestingly, several power forwards post great block rates, and very marginal “Opponent FG%”. Brandon Wright, Jan Vesely, Terrence Jones, and Mason Plumlee each: have block rates between 3.7 and 4.3 (85th percentile or better of PF’s); opponent FG% between 51 and 55 (25th to 50th percentile); and below average defensive RAPM. I can’t speak in depth to each of these players, but they may be too aggressive, frequently blocking shots, but often also fouling, or biting on pump fakes.
Clearly one of the biggest problems with this data (and my analysis), is the “per game” nature of the first two items. With varying minutes played and team paces, per game stats are always bound to be misleading. Plus RAPM is per 100 possessions, which pretty well leaves the correlation with the first two measures (makes and attempts per game) “apples and oranges” comparisons.
Luckily, Seth Partnow (@whroffnshppns) of whereoffensehappens.com, created a variety of supplemental metrics that manipulate and combine the basic SportVU data for Basket Protection. I will particularly focus on his “Adjusted Points Saved” per game and per 36 minutes. These numbers combine the frequency with which players contest shots at the rim but also their effectiveness at stopping the potential scorer according to “Opponent FG%”. Finally, an adjustment is made for pace, while also controlling for whether the player’s team is particularly good at limiting shots near the basket. Correlated with Defensive RAPM, those metrics provide the following:
First conclusion…nice job, Seth. For this season at least, your “Points Saved” stats are an upgrade over the SportVU “Opponent FG%” and a large improvement over “Block Rate”. The second conclusion is that rim protection is more likely to be indicative of better power forward defense than for centers. Here is a graph of the 0.43 correlation:
First, the disclaimer. Centers are a bigger defensive force than Power Forwards; minute weighted average for DRAPM for the Centers was +0.32 and Adjusted Point Saved per 36 is 4.9 points. For Power Forwards, it is -0.37 and 3.4. For the power forwards with more than 500 minutes though, of the bottom 25 players for “Adjusted Points Saved per Game”, 23 are below average defensive power forwards according to RAPM. The four Power Forwards with the highest “Adjusted Points Saved Per Game” are Serge Ibaka, Tim Duncan, Paul Millsap and Amir Johnson, all players with very strong defensive RAPM. Millsap and Johnson, both previously featured as APM All-Stars at gotbuckets, look much better as basket defenders according to this metric than they do according to block rate.
Like so much in life, these “answers” lead to more questions. Why were the Power Forward correlations stronger than for Centers? It is probably because a large proportion of Centers are strong rim protectors and are employed because of that; it isn’t necessarily what seperates them as defenders. For the Power Forwards, adding a second basket protector to a defense is a bit more of a luxury.
And also, how much should any trait (in this case, basket protection) be valued relative to other skills? The first article in this series dealt with big man shooting range, with all correlations at 0.3 or less. For basket protection, the correlations extended to over 0.4. Clearly it is not as simple as taking that and saying basket protection is more important for a power forward than floor spacing & shooting range…but that could be a general outcome. Per pace-adjusted 36 minutes, Amir Johnson ranked in the top ten percent of all power forwards for basket protection; RAPM loves him and the SportVU stats may expose a new means to expand the understanding of his impact on the court. This Information could be valuable when evaluating free agents, or in viewing video to analyze how Amir impacts so many shots (the next level of “why”). The video could inform important skills to teach other big men.
I am beginning to ramble, but this somewhat comes full circle to the first articles I wrote at gotbuckets, about the undervaluing of defense. I never really completed that concept. Maybe this summer, a nice bow can be tied on all of this with a big regression, and I can say something really meaningful. Probably not though. For now, I’ll keep randomly correlating a variety of new metrics (SportVU, Synergy, etc) with RAPM.
Until next time…