“The Dwight Effect” and Kevin Garnett
Posted by Brendan Jackson on Mar 6, 2013
We all know Kevin Garnett is an elite defensive player. Upon his arrival in 2007, the Celtics jumped from the 16th best team in defensive efficiency for the 2006-2007 season to 1st at the close of the 2007-2008 season. While that shift to the upper echelon was due in large part to the arrival of defensive guru Tom Thibodeau that same year, Garnett’s instantaneous effect on the defensive culture in Boston can not be overstated.
Flash forward six years. The Celtics have had five successive seasons of finishing with a defense that is top five in efficiency (6th so far this season). Three years prior, Thibodeau had bolted Boston for a much deserved head coaching gig in Chicago. In the three years sans Thibs, Doc Rivers has continued to preach the defense-first mentality, something engrained in the roster holdovers from that championship season. After Thibs’ departure, the Celtics even hired another defensive maestro in Lawrence Frank to prevent any defensive drop-offs. It all worked. However, the Celtics’ sustained defensive prowess has only had one real constant aside from Doc Rivers: Kevin Garnett.
Garnett’s on-court quarterbacking has become stuff of legend. His defensive rotations are sharper than Santoku Knives. His pick-and-roll defense is lethal. All anecdotes of Garnett’s defensive abilities will be recorded in the annals of NBA history, to be retrieved 50 years from now when codgers like me want to educate the young, uninitiated in dank Boston bars about what it truly means to dominate defensively. It’s a good thing advanced statistics exist. When I’m regaling those young basketball fans of the legend of Garnett as a Boston Celtic – telling them about how LeBron James never scored on him– I’ll be able to utilize this quantitative analysis to substantiate anything I haven’t exaggerated or embellished.
Defense is a difficult aspect of the game to properly evaluate. We have things like defensive ratings, on-court/off-court numbers, defensive rebound rates, and block rates to name of few go-to statistics that help shape a particular player’s defensive narrative. In the future, we will be able to deconstruct these numbers even further. If you were to believe Kirk Goldsberry and Eric Weiss, the future is now.
I love things like defensive rating and on-court/off-court numbers, but they lack the depth necessary to really properly evaluate a player’s defensive ability. Goldsberry and Weiss have shoved a Z-axis up this two-dimensional plane through their usage of spatial analysis. At this years’ MIT Sloan Sports Analytics Conference, they presented a paper entitled “The Dwight Effect: A New Ensemble of Interior Defense Analytics for the NBA.” The paper’s goal was to, well, I’ll just let them tell you:
The core objectives of this paper are 1) to improve the characterization and understanding of interior defense in the NBA, and 2) expose key challenges associated with measuring defense as new forms of performance data emerge. We present case studies that 1) use spatial analyses to extract new defensive metrics from optically tracked game data (SportVu data) and 2) use visual analytics to present results.
Makes sense? No? From what I understand about really smart endeavors – which, is admittedly very little – Goldsberry and Weiss want to expand the limits of defensive metrics and provide context to these numbers through the usage of spatial analysis. These two nerds want to add two components to measuring defense: 1) how close the defender is to his offensive counterpart; and 2) where they are on the court in relationship to the basket. I call them “nerds” both affectionately and with a hint of jealously. This stuff is just so FREAKING cool.
The findings of the paper were even more interesting. Even though the paper was titled “The Dwight Effect,” Larry Sanders proved to be the best interior defender in the NBA. This is not so hard to digest given Sanders’ length, athleticism, and energy. The real Scooby Doo moment came when Goldsberry and Weiss revealed that Andrea Bargnani was one of the top five rim protectors. Since they knew that to be false, Goldsberry and Weiss surmised that Bargnani was a beneficiary of their proximity variable. The data was based on defenders being within five feet of their opponent and there’s a good chance that Bargnani was often too far away from his man and thus reduced his sample size (I could also see the Steve Novak effect being a possible explanation). Whatever the reason, Bargnani sucks.
Speaking of sucking, David Lee registered so low in this study Goldsberry and Weiss made a video tribute of his bad defense. This was the real benefit of being in the audience at Sloan. The highlight of Day 1 was watching the inexplicable defensive “strategies” employed by Lee, against a counterpoint of Goldsberry’s pleasant, matter-of-fact monotone narration. Unfortunately, I couldn’t find that video and it would be without narration if I could so you’ll have to settle for the closest thing I could find to David Lee playing defense.
Absent in all of the analysis is Kevin Garnett. While not specifically known for being a rim protector (he’s more of a master-of-all-trades defender) I was still surprised when Garnett didn’t make the top-five. He actually didn’t even make the top-ten. Based on Goldsberry and Weiss’ research, here is where Garnett ranked:
If I’m interpreting this data set correctly, 54.4% of the shots Garnett defends are in close proximity to the rim and his opponents are shooting 48.3% on shots within five feet of the basket. Not bad ranks at all, but still not the elite level you associate with Garnett. It’s worth mentioning that these numbers are based on Garnett being within five feet of his defender. Based on the Celtics’ semi-complicated rotational schemes, Garnett could be a victim of his less-talented teammates blowing an assignment and Garnett having to retreat to within five feet of an opponent’s easy hoop. This is speculative, but when Garnett is counting on guys like Brandon Bass and Chris Wilcox to make their rotations, it’s definitely in the realm of possibility.
This could also mean that Garnett is not as effective a rim protector as he once was. The data for this study is from only the last two seasons and that’s hardly enough to make any definitive statements about someone’s career. It’s reasonable to think that Garnett has declined over the past two years, given his advancing basketball age, the tread on his tires, and his decrease and athleticism.
Here’s what we do know: Garnett is still an elite defensive player within the Celtics’ system.
Here’s how we know it: When Garnett is on the bench, the Celtics give up 109 points per 100 possessions. When he’s playing, that number comes down to 100 points per 100 possessions. According to mySynergySports, Garnett’s opponents’ overall FG% is 35.5%. According to my memory bank, Garnett has yelled, “by yourself,” “get him,” and “get that [stuff] out of here” 9,457 times.
Add (figuratively!) the two together (Goldsberry/Weiss and traditional defensive metrics), and you would have a great candidate for the NBA’s Defense Secretary.
It’s possible that people smarter than I can poke holes in Goldsberry and Weiss’ analysis (like this guy) but it certainly adds a new dimension to the evaluation of individual defenders. There are plenty of things I left out so if you have the time, check out the paper and/or watch this video. Both are well worth your time.