I decided to look at the SportVU defensive data for the top eight Rocket's players in minutes played. For your information SportVU data is the seminal analytical tool available to NBA teams. Even the neutered version offered to fans is the most precise measure available of many different metrics in the NBA. SportVU data offers a special look into the defensive prowess of NBA players. Normally we fans are limited to imprecise metrics such as xRAPM. SportVU offers a precise analysis of every play in the NBA, and distinguishes between the specific offensive player versus the specific defensive player. SportVU then rolls all of that data up into an easily digested format. In this article I will focus on the Rocket's defensive player's ability to reduce the offensive player's FG%. After all that is what NBA defense is all about. There are a number of caveats to this analysis. It does not take into account the Howard effect (The Dwight Effect:A New Ensemble of Interior Defense Analytics for the NBA, MIT Sloan Sports Conference, 2013). I have assumed that the SportVU data properly corrects for the quality of the offensive player. I believe it does, but I have no proof that Westbrook is treated differently than Canaan on offense. That issue is easily solvable. Unfortunately Jones does not have a statistically significant amount of data this season. For that reason I did not include his numbers. Papa is on the borderline, but I did include him. He has four times as many minutes this season than Jones. A player like Ariza may guard more dangerous opponents than someone like Harden. Because I can tease out the data easily here are the overall shooting percentages of the opponents for each player covered. As expected the Rockets players who guard mostly around the basket have their opponents make a higher percentage of their shots. The data presented in the following graphs is easily interpreted. The better defender will always hold their opponent to a lower FG%. Therefore the more negative the number the better the defense. This chart conveys the over riding theme of this article. Smith is really really good on defense. And D-Mo is really good. Otherwise it is a mixed bag that I would not have expected. Smith, despite who he guards around the rim is a great defender. Dropping the FG% around the rim from say, 54% to 44%, is actually a huge deal. That reduces an opponent like West to somewhere around the offensive force of a second string power forward like Plumlee. D-Mo, with around fifty percent of his court time at center, effectively reduces a Cousins to a Len. From a powerhouse to an average NBA center. Ariza, Bev and Brewer are obviously over matched around the rim. This graph documents each players defense from inside the arc to the basket. It is reminiscent of the previous rim protection graph. What this graph really describes is how quick a player is closing out on two point shots, and how effective they are on the close out. And again somehow we again see that Ariza is less effective than other Rockets defenders. A 7.6% over average FG% is huge. More on this later. But I would like to point out that besides D-Mo and Smith, the Rockets cannot effectively defend a player like LMA. It will become obvious later why Howard would not be a great choice to defend LMA. What really stands out to me in this graph is how darn good Papa and Harden are at defending the arc. Those two players basically convert a Curry into a D-Mo from beyond the arc. And look at D-Mo's numbers. I have little doubt that D-Mo is the fastest seven footer in the NBA. This graph illustrates why the Rockets team is the best in the NBA at defending the three point shot. A minus five percent in opponents three point percentage takes an offense from ~36% to ~31%. No surprises here. Except Papa being so good 15 feet and out from the basket. The eye test is certainly satisfied that Howard has a difficult time leaving the paint to close out on shots away from the basket. This means that whoever is playing power forward along side of Howard must be great at switching on defense so that they can cover Howard"s defensive outside responsibilities. Howard definitely won't be there 18 feet out. Less than 10 feet. The paint. Again it becomes obvious that Brewer, Bev and Ariza need help down there. Significant help. However Smith, Howard and D-Mo can obviously handle their defensive assignments down there with little support. I do find Harden's defense to be less robust than I would have expected. But then again it is approximately league average. And then the overall data. How important is the difference between Smith's -4% and Bev's +2.5%. About 13 points per 100 possessions. At 1.03 points per shot in the NBA. D-Mo's stats do pass the eye test. He is all over the floor, and is great at intimidating shots. Harden's numbers belie the MSM rhetoric that Harden is not a very good defender this season. Brewer, Ariza and Bev are not as good on the defensive side of the ball as my eye test. However their three point defense is obviously the league best. These numbers were much better earlier in the season. Which is why the Rockets are gradually falling in defensive ranking. A final caveat. Some statistical analysis has shown that Houston's defense benefits from an above league average of missed wide open shots.
So basically this confirms what everybody has seen this year - Beverley's defense has really slipped.
My take away from this analysis is that Bev is obviously playing hurt, Smith was a much better pickup than Brewer, Howard has been hobbled since his first knee injury, and Harden is playing third team all NBA defense this season.
Good stuff. Ariza has got to improve somewhere. Either side of the ball. I still believe he's a big part of our 3-point defense, he gets steals, but I feel like he's basically having a neutral impact at best at the end of the day.
Excellent point sir. I had not noticed that aspect of the data. But that is one reason that I post these analyses. So that I can get other peoples insights.
If your stats are showing Harden is a better defender that Ariza, you may want to stop and think deeper for some more facts.
The >15 ft chart can not be more misleading. Dwight does not have difficulties to roll out to defend mid range to long shots, the fact is, the Rockets does not want him to. From couple games against the Grizz, he was staring Gasol to shot the long 2 and 3 while just stay back and look for rebounds. That along tells you how much the Rockets are relying on him to protect the paint and rebounding.
These are SportVU stats. They are gathered from video analysis of every single play in the NBA this season. From six digital cameras surrounding every NBA court. These really are not stats, they are data. Facts. They show who the offensive player is on every play and who the defender is and then catalogs the outcome. No one in the NBA analytics community questions this data. This is the fourth year that Morey has used it.
It is not that Howard physically cannot defend beyond 10 feet. He psychologically cannot separate himself from the paint. If he could we would have seen him all over LMA last season in the playoffs. No one in the NBA is better equipped to shut down LMA physically than Dwight.
The Rockets defense will rotate depending on the ball handler a lots, especially on pick and roll. And because of that, you will see people keep rotating on their defense assignments. One of the biggest problem of Harden's defense is he is often missing his man and that resulting wide open shots and more than often, that will rely on other guard and wing defender to rush in and cover up for him. Not to mention Rockets is the worse team in TOV and when the other teams are cashing in those TOV, more than often, Ariza and Bev were the first to run back and of course the other teams were making very high percentage shots over their faces. Harden has improved his defense this earlier season but for the last 1-2 months, his defense is not that good at all. These are also facts.
Excellent points. In statistical analysis you have the data, you have the stats, and then you have the interpretation. SportVu provides the data. I did not do any statistical analysis of that data. But then there is the interpretation. That is where the rubber hits the road. Bad interpretation of data screws everything up. On Harden's defense. Harden is somewhat "protected" on defense. For instance the next time Houston plays Dallas I expect him to draw Rondo on defense. However NBA defenses are predicated on rotations these days. So it is impossible for me to tease out Harden's exact defensive performance from this limited subset of SportVU data. But overall SportVU provides a comprehensive view of NBA players defensive performance. It is not humanly possible to do a better job. So I rely on SportVU's data above stats like xRAPM, PER, ...
I think overall speaking, Bev's defense is taking a step back this year, whether his role had been changed or he is trying too hard on offense or maybe both could have an affect on his defense, but his speed and energy is not matching with what we have seen the last 2 years. As for Ariza, he is one of the worse offensive players in the whole NBA this year. I think couple games ago, there is a chart to show in the whole NBA league, with players taking more than 400 shots, the top 5 players with worse FG%, and Ariza is no. 2 with less than 37%. No. 1 is CJ Miles and 2 Lakers also were in the top 5 (Young and Kobe). Ariza's offense is bad, really bad. But on defense, he is still contributing and no doubt the best outside defender in the whole team. His playing time can speaking for himself. Morey and McHale both know about this, Harden is having an amazing offensive MVP season but he still needs help on the defense side to this team to win games. Hence Ariza still plays major minutes.
IMHO Bev has not been healthy since he missed time with the leg injury earlier in the year. I ran the numbers on 12/6/14 and his defensive numbers were exceptional. To get the best data one would have have to get the numbers for all small forwards in the league. SportVU does not make that easy for fans.
I would say that vouching for Harden playing 3rd team All-D is a very bold statement, especially considering he isn't arguably even the 3rd best perimeter defender on the team. Well, I guess it matters how dialed in he is... hence "arguably". I like the analysis, I really do. But as Morey said, analytics is a supplement to watching the tape. I am not sure the right conclusions are being made... but that might be more of a problem with the data. I think it is similar to economics: The Production Function, Utility, etc. are all nice and dandy, but are not a final say-all. But they can still explain a lot. Does that make sense?
I believe the data is correct. Well, as correct as any data set of thousands of hours of video analyzed by algorithms run on near super computers can be. My analysis however is opinion of course. I hope that it did not appear that it was anything other than my opinion. The Harden statement was half in jest though. For example ezPM, an analysis site that actually parses through all of the game logs Full Name Defense James Harden 1.98 Stephen Curry 0.68 Chris Paul -2.18 Anthony Davis 3.83 LeBron James 1.53 John Wall -0.17 For an explanation as to how this the defense data was derived see http://www.d3coder.com/thecity/2011...orporating-counterpart-defense-or-the-biggie/
Why don't you post this data for Goran, Lawson, RJackson, DCollison, Knight, DWilliams, Bledsoe,...and Rubio? Also, how about comparing Ariza, Brewer, Parsons, Tobias Harris, Tucker, Afflalo, Wilson Chandler, Jae Crowder, Jeremy Evans, Khris Middleton, and Jeff Taylor?
