Introducing SportVU On-Off Data: A Look at the MVP Candidates

Mar 22, 2015; Los Angeles, CA, USA; Los Angeles Clippers guard Chris Paul (3) tries to dribble past New Orleans Pelicans guard Norris Cole (30) during the first quarter at Staples Center. The Clippers won 107-100. Mandatory Credit: Robert Hanashiro-USA TODAY Sports

With the help of Darryl Blackport,[1. Whatever awesome interesting data you see on Nylon Calculus, Darryl was likely the one behind it. He’s absolutely invaluable to the site and I think all my fellow writers would agree with this.] I’ve been able to take an On-Off to some of the public SportVU data.[2. Darryl merged the public SportVU data logs with public play-by-play data. Then I ran all of the queries you’ll see in this article using R.] Specifically, every offenses’ goal is to create open shots[3. Though not all open shots are created equal and we’ll take a look at this later in the article.] and so we’ll look at how the MVP candidates create these shots for their teammates and what happens when you take them off the court[4. An open shot is defined as either Open or Wide Open based on SportVu’s definitions which can be found on player pages. Open rate is simply Open shots / all shots.]

So without further ado, let’s take a look[5. On-all includes shots taken by the player. On-teammates’ only includes shots only taken by the players’ teammates. The players’ own shots are removed in this case. The difference column is simply On- teammates only minus Off.]:

Open rate on all shots  | Create infographics

There’s a lot to look at in this table.[6. As with all On-Off stats, a big issue is going to be sorting out the entanglement of teammates’ who play a ton of minutes together- this is why we have metrics like RAPM. This analysis isn’t accounting for this effect and that’s why we have to keep this in mind and be careful with interpreting the results.] First, note the massive difference in CP3’s teammates’ open shot rate when he’s on the court versus off the court. CP3 has remained on the fringes of the MVP race with most overlooking him in favor of Curry and Westbrook, but if anything shows how important he is to the Clippers, it’s this graphic.[7. Although as we’ll see in a bit, this graphic is slightly misleading for CP3.]

How about Anthony Davis though? His teammates’ are actually getting a higher rate of open shots with him off the court. That’s drop-your-jaw level surprising. Is this an indictment on Anthony Davis? Perhaps but maybe there’s a bit more to this story that we’ll see later.

Additionally, we see that when Lebron, Curry and Harden sit, their respective teams have a below-average frequency of open shots but when they’re on the court, they all improve the teams’ open shot frequency to above-average rates.

Perhaps most interesting is the case of Russell Westbrook, who when you include him in the team totals is actually lowering the teams’ open rate compared to when he’s off the court. However, you’ll notice that he does give a slight boost to his teammates’ open rate when he’s on the court compared to when he’s sitting. Still, given how many contested shots Westbrook takes, we’d expect the difference in his teammates’ open frequency when on the court compared to off to be a bit larger.

Of course one of the biggest issues with the graphic above is that we’re giving open midrange shots equal weighting as open shots near the rim or open threes. Treating open shots from those locations as equal in value is simply not accurate, no matter what you may think Randy Wittman. So let’s look at Open Moreyball rate[8. Specifically: (Open 3s + Open Restricted Area shots) / All shots]:

Open Threes and Open Restricted Area rate | Create infographics

Let’s go back to Chris Paul, who suddenly doesn’t seem to be generating many more open shots for his teammates compared to when he’s off the court. Here we’re only looking at Moreyball shots. So it appears CP3 is generating a lot of open midrange shots for his teammates[9. I assumed that a majority of these were going to Blake and after doing some digging, I found out of the 711 Open midrange shots the Clippers have taken when CP3 is on the court but not the shooter, 396 were from Blake with 97 of those being assisted by CP3.] but not many more open threes or open shots near the basket. The Clippers Open Moreyball rate with CP3 off the court is only minimally worse then when he’s on the court with his teammate’s shooting. And in fact, if we add in CP3’s own shots, we see the Clippers are actually taking more Moreyball shots with him off the court. This certainly diminishes CP3’s impact from the first graphic a bit.

Earlier, we mentioned there was more to AD’s story than the first graphic showed and we see that here. Davis is now on the plus side when looking at his teammates’ Open Moreyball rate with him on the court compared to when he sits. However, the difference isn’t too large and we still see that once we factor in Davis’ own shots, the Pelicans are taking less Open Moreyball shots compared to when Davis sits.

How about Steph Curry though? What a huge difference he makes. The Warriors are getting a below average Open Moreyball rate when Steph sits[10. And it’s not just below average, it’s three percentage points below average, which in this case is fairly large given the scale.] but when Steph is on the court, his teammates see an increase of eight percentage point in Open Moreyball shots. And when factoring in Curry’s own shots, that jump is up to 10 percentage points. If we’re looking for reasons to give Steph the MVP, this is certainly one of them.

Lastly, we’ve been talking about Moreyball Open rate this whole time but neglected to mention the team who gives those shots the Moreyball name and that teams’ star player — James Harden. The splits for Harden are higher than any other player and by a decent amount no. This is clearly a function of the philosophy employed by the Rockets as we see that even when Harden sits, the Rockets are still getting a higher percentage of Open Moreyball shots then any of the other teams. But Harden does make a difference and we can see that his teammates’ are getting about four percentage points more Open Moreyball shots with Harden on the court compared to when he sits. And even when factoring Harden’s own shots in, that rate is still higher than when Harden is sitting[11. Most stars’ are taking a lower % of open shots because they are well “stars”. And so that’s why you generally see the On- All lower than the On- teammates’ only category.].

Finally, I thought I’d wrap it up with some words about the data itself. This is clearly a fascinating dataset where we can look at all sorts of cool things but we also need to be careful about applying and interpreting On-Off data. This is no different from looking at the On-Off data you see on the excellent NBA WOWY. While powerful, this data still suffers from the issue of collinearity. And this is why potential RAPM[12. Or any other advanced modeling techniques.] models of this dataset could be interesting.[13. Full Disclosure: I do have a RAPM Rim/Threes Protection and RAPM Rim/Threes Deterrence metric that I’ve been working on and with the addition of this kind of data, it could make the metric much better.]