The NBA is still going strong despite an avalanche of key injuries. Kevin Durant was injured a week ago, but the Warriors are as equipped to handle his loss as any team could be. For teams like the Raptors and the Clippers, the loss of their best player is enough to send their respective teams into a tailspin. It’s a war of attrition right now, and all the high seedings and awards will be given to the survivors.
Luckily, the league is deep enough now that it can absorb the losses without the year seeming like a total loss. Maybe the basketball gods are toying with us — Andrew Bogut breaking his leg one minute into his Cleveland career is funny in a morbid way but painfully frustrating for a Cavalier fan — but I hope that by the time the playoffs start, the chaos dissipates.
And with that, let’s look back at the last week in the NBA.
MIT Sloan Sports Analytics Conference
I imagine people reading this are familiar with the MIT Sloan Sports Analytics Conference. If you didn’t attend, you can check out the coverage here for both days. The research papers generate a lot of attention, but the conference is an event because it’s a convergence of some of the brightest minds in basketball, and it’s rare you can get them all in one place. Hence, it was a good opportunity to interview some NBA thinkers, like Nate Duncan, Chris Herring, and Neil Paine.
As for the specific papers, there were three about the NBA, and two of them I found particularly interesting, along with a few other posters as well. Firstly, there’s the possession sketching paper, where researchers used SportVU movement data to catalog NBA actions and play structures. This actually reminds me of a previous Nylon Calculus article. Then there’s the paper on NBA field goals and how players are “posed” when they take and make them. Once again, Stephen Curry is an outlier — who knew? You can see all the papers here, including the full papers for the presented posters. Even if you don’t think you have the ability to parse the research, you can at least look at the spiffy figures and read the summaries.
The Most Valuable Player candidates
If it weren’t for that horrible stretch of injuries, this season would have a solid argument for deepest MVP-race ever. From James Harden to Kyle Lowry, my metric HBox sees eleven decent candidates, if you ignored the injuries to Lowry, Kevin Durant, and Chris Paul. I know DeMarcus Cousins is a stretch, but his stats suggest a borderline MVP-candidate. It’s not often you see a center take 37 percent of all his team’s available shot attempts including free throws. Using my MVP Index stat, there are only two seasons where seven players had an index of 10 or higher, and interestingly they were two adjacent seasons: 1990 and 1991. The players were Charles Barkley, Magic Johnson, Michael Jordan, Karl Malone, Hakeem Olajuwon, David Robinson, and Stockton for 1990. Then you’ve got the same group for 1991 minus Charles Barkley and Hakeem Olajuwon, and with Kevin Johnson and Scottie Pippen instead.
We’ve already got five players this season — Cousins, Durant, James Harden, LeBron James, and Russell Westbrook — and there’s still a big chunk of the season left. It’s likely Giannis Antetokounmpo and Kawhi Leonard join the group. In other words, it’s not a good time to have a career season — there’s a big crowd at the top of the NBA landscape, and this race is going to be tough to call.
Kawhi Leonard’s defense and the search for an explanation
This MVP race is as wide open as it’s been in the NBA in recent memory, and one player’s candidacy rests on his defensive reputation: Kawhi Leonard. Since he’s become one of the best scorers and overall offensive players in the league, the two-time Defensive Player of the Year winner should surely be the best player overall. But there’s a reason he’s not the front-runner for that defensive award right now. San Antonio’s defensive rating is at its worst with him on the court, ignoring a couple of low minute players. This partly explains why most advanced metrics don’t have him first in the league, despite his defense.
I’ve seen various explanations as to why the Spurs are not defending well with Kawhi on the court, but the one that’s stuck to me the most is this article by Matt Moore with the premise that Leonard is so good at defense that he’s hurting his own team. Teams are supposedly freezing him out on defense, hiding the guy he’s defending in the corner and operating essentially 4-on-4 against the other players. However, I don’t buy that hypothesis. For one, it doesn’t appear Kawhi is unusually inactive on defense this season. Some of his stats like steals are what you’d expect and within career norms, and he’s contesting a normal amount of shots: not as many as Danny Green, but more than most of San Antonio’s other perimeter defenders. Nothing really changes if you only look at the games before the article was published either.
I also have this objection: if this is such a great strategy, then why would teams only use it against Kawhi? In theory this would work against any guy, and any reason it’s a better option for Leonard would have to involve him being an inferior defender (e.g. not being quick enough to offer help defense, which is obviously not true.) Plus, the starting lineup does have another above average defender in Danny Green with LaMarcus Aldridge not being bad either, and Leonard plays in many lineups with other plus defenders like Dewayne Dedmon.
It just doesn’t make sense that freezing out Kawhi is responsible for the outlier stat. Finally, let’s go to my favorite stat: opponents have their highest 3-point percentage when he’s on the court, and that’s something that’s mostly noise. I don’t think he’s the most valuable defender this season, partly because his efforts have shifted more to scoring, but he’s definitely not a liability and this weird on-off trend won’t continue much longer.
Impact: No longer just for bugs on a windshield
We have yet another tool over at stats.NBA.com, and it’s similar to NBAWOWY.com: impact. You choose a player on the right you want to see on/off splits for, and then you choose the players on the left you want to see put through that analysis. For example, here’s how the Warriors do with Stephen Curry, Klay Thompson, and Draymond Green, with and without Kevin Durant. You can see how opposing teams perform too. Here’s how LeBron James and the Cavaliers do against the Celtics when Jae Crowder is on the court versus off the court.
This doesn’t completely displace NBAWOWY.com, however; the stats.NBA.com version doesn’t share some of the same features and, most importantly, is only available for this season (at least currently.) If you’re wondering if some lineup combo’s trend is meaningful, it’s helpful to see what happened in past seasons. But it’s something else to use every now and then from a site providing an under-appreciated amount of information.
As we get deeper into the MVP race and our arguments solidify, I thought I’d mention a couple important points about one of the most widely used stats, Basketball-Reference’s BPM, and how it pertains to Russell Westbrook. He is going to topple the record for highest BPM in a season by a good distance — and the same goes for usage rate. His quest for averaging a triple-double is looking good too. All are related: BPM is built with interaction variables, which multiply stats like usage rate, rebound rate, and assist rate together. Thus, doing extraordinarily well in all three categories will super-charge your rating, even if your team is not particularly good.
What’s the most damning bit of evidence? Westbrook’s defensive BPM is currently third in the league, only behind the two main Defensive Player of the Year candidates in Rudy Gobert and Draymond Green. This is because Westbrook is rebounding on defense like a big man, and the metric BPM found that big men (and every player with a high rebound rate) with high assist rates were often better at defense than traditional stats suggested. But Russell is an outlier, and it’s stressing BPM to no end.
This brings me to my final point: although BPM is known as “Box Plus-Minus,” it does not use plus/minus in the actual ratings — plus-minus was only used to build the weights for every variable. So no, BPM is not accounting for Westbrook’s global effects on defense. Something like ESPN’s RPM does, however, and its view of his defense makes more sense: his defensive RPM is -0.09. This is corroborated by Stephen Shea’s defensive metric too, which aggregates all the easily countable defensive stats available now from deflections to shots “defended,” where Westbrook ranks 60th out of 244 players — even with defensive rebounds included. He’s still one of the best players in the league, but he’s not running away with the MVP award.
Dallas’ resurgence and a Yogi’s influence
The Dallas Mavericks have been on a tear lately, and many have subscribed their success to a fun and previously unknown small guard, Yogi Ferrell. But their inflection point for the season actually started before that. I picked Jan. 12 for that date — Yogi’s first game was on Jan. 29. Before that day, the Mavericks were 11-27 for a win percentage of .289, which was the 28th-best record in the league. But from Jan. 12 on, they’ve been 15-9 for a win percentage of .625, which is good for seventh in the league. That’s a stark contrast, and I understand the need to identify a more tangible cause — and one preferably with a face. However, I’ve discussed Dallas several times this season already, and one factor caught my attention: opponent 3-point percentage.
Let’s look at those same time frames again: during the first part of the season, opponents shot a blistering 40.4 percent from behind the 3-point line. From Jan. 12 to Mar. 5, that’s dropped to 34.1 percent. Their defense rating has greatly improved too, and a large chunk of that is because of opponent 3-point percentages: when the average team is taking 27 of those a game, every percentage point matters.
As I’ve discussed frequently, those percentages aren’t easily controllable by NBA defenses, and that 40.4 percent mark was unsustainable, even for a poor defense. I don’t think Dallas was ever truly that bad on defense; they’re well-coached and had useful defenders like Andrew Bogut in the middle (and now they have Nerlens Noel.) Unless, of course, you think Yogi is the reason for their renewed defensive vigor, but I think the explanations are more mundane — fluky shooting percentages and random chance.
If you’re the type of person who only cites a statistic when it supports your favorite player and your favorite player happens to be Kawhi Leonard, let me introduce you to Dredge — a statistical plus-minus model that includes miscellaneous play-by-play stats like offensive fouls drawn and goaltends. Kawhi is first there, and the two guys behind him are injured (Kevin Durant and Kyle Lowry.) Of course, there’s no agreement among the metrics this year, so don’t expect this to stop any arguments. For instance, in HBox, James Harden has quietly had the greatest season on record, which goes back to 1974. There are good arguments for a number of players — have fun.
The cry of the timberwolf
Once again, Minnesota’s season has been submarined by a weird inability to convert point differential into wins. Based on how they outscore opponents in an average game, you’d expect they’d be around 0.500 now. Instead they’re on a pace for about 33 wins. This is typical for Minnesota, however; they had that memorable season with Kevin Love where they had an adjusted point differential of 3.1 and still had a losing record — that’s a point differential you’d see on a weak title contender, not a team with only 40 wins. You can see how they’ve fared since 2011 in terms of win percentage versus margin of victory in the graph below. I’ve also plotted their antithetical team, the Memphis Grizzlies, as well.
That trend has persisted after Kevin Love left the team, and it was around before Ricky Rubio too. In fact, you have to go back to the days of Kevin Garnett to find a few seasons where they regularly outperformed their point differential. If there’s a fix to be had with their close game performance, it would do wonders with the perception of the team and the development of Karl-Anthony Towns, Andrew Wiggins, and others. Pre-season expectations had the Wolves clearing 40 wins, so many are thinking of this year as a lost season mired in failures. But the problem they have now doesn’t have to do with talent; it’s their old problem of winning close games and converting point differential into wins.
As a part of my ongoing efforts to provide miscellaneous stats, I thought I’d tackle the fabled chase-down block — which I declared years ago was a fool’s errand. For the uninitiated, it’s a stat associated with LeBron James because an Ohio sportscaster, Fred McLeod, coined the term when covering James. The impetus was actually the famous Tayshaun Prince block on Reggie Miller. The spirit behind the stat is about quantifying and identifying when LeBron snuffs out an easy transition bucket from seemingly out of nowhere — it’s about when he runs from baseline to baseline to block a shot right at the rim. At first glance, that does not appear to be an easy stat to create en masse, but with a few strict filters I think I have a decent method to generate the chase-down block.
Here are the conditions: a blocked shot within five feet of the rim and within five seconds of a possession change (i.e. turnover, made shot, made free throw, or defensive rebound.) That’s all you need, and it works decently well enough to be usable for every season with play-by-play data available. When the Cleveland was tracking LeBron’s chase-down blocks, they found he had 23 for 2009 and 20 for 2010. My method has 28 for 2009 and 21 for 2010 — it doesn’t perfectly align. I imagine their subjective stat-tracking was stricter because they weren’t counting plays where he started at the half-court line, but I don’t have access to that information. It’s an approximation, but at least it’s consistent.
I’ve collected all the results since 1997 up to Mar. 5 of this season, and the top seasons are listed below. So, yes, LeBron James does not have the best season — that belongs to Josh Smith in 2006. Maybe that’ll surprise people now, but back then he ran as smoothly as a deer and was a terror in transition. In fact, one of my most vivid basketball memories in person was watching him sprint down the court and block a shot at the rim versus the Blazers years ago. Bo Outlaw is more surprising to the average fan, but he was a fantastic shot-blocker and he was quite athletic. There are a number of other big men too, and it makes me wonder how many of these blocks occurred with a big man who was already at half-court when the fast break truly began.
Table: chase-down blocks, 1997-2017
If any one player should be associated with the stat, it’s actually Andrei Kirilenko, who’s the prototype of the guy who does well by this measure: an athletic and long forward (other examples include Lamar Odom and Nic Batum.) K.J. McDaniels sticks out though, and it’s not just because he was listed as a shooting guard. He had 20 of these in 2015 when he played only 1352 minutes, and per minute it’s the highest rate anyone’s had with at least that many minutes. Chris Andersen had a higher rate in 2004, but it was in 1029 minutes and his rate was barely higher.
For career totals, Josh Smith is in the lead too with 174, but that lead won’t last: LeBron’s at 167. For this season only, two players are tied through Mar. 5: John Wall and Giannis Antetokounmpo with eleven each. I was delighted to see both those names because Wall has definitely had a few memorable chase-down blocks, and Giannis is, well, Giannis.
The stat needs some fine-tuning. There are too many plodding big men with a few chase-down blocks. I believe the central issue is that I don’t know where the shot-blocker was when the possession change occurred; thus, many of these aren’t true “chase-down” blocks because chasing did not occur — the guy was already out in front. Here’s an example where Brook Lopez is simply ahead of the offense, and blocks a shot at the rim. But, hey, Brook deserves a little credit too, because even if he’s starting at the half-court line it can still be an impressive block.
With SportVU data, this would definitely be possible to track accurately. You just need the positions of the shot-blocker when the possession changed and the ball itself, and with some reasonable filters you’ll have fewer Lopez blocks and a better looking stat. I don’t have that kind of access, but for now I can provide this stat, along with a couple others, in this little google spreadsheet.
And LeBron James may not be the “best” chase-down artist, but he has the most memorable chase-down block in all of NBA history.