Is there trouble in Houston?
Over the past few days, Houston has added to its list of worrying losses to teams like the Mavericks and Wizards. This far into the season, if you were a hopeful contender and you’re struggling to stay above 0.500, the panic button has already been pressed. We all expected some dip in performance this season given the loss of Trevor Ariza, but the drop has been suspiciously severe. What happened?
Firstly, the team has actually been healthy, and outside of the brief Carmelo Anthony experiment, there’s been a high degree of roster continuity. The team has only had minor injuries to its key players so far. They even got Jeff Bzdelik, their defensive guru from the past couple seasons, back and they still haven’t regained their form.
Turning to the stats, the Rockets have had a remarkable collapse across the board. Their defense has changed from borderline elite to well below average, and their offense is no longer good enough to cover up any mistakes. You can see their basic stats in the bar graphs below. Their shotmaking, in general, has taken a dip, which is more important than offensive rebounding — their one big improvement. However, their defense has clearly fallen apart. Oddly, they’ve gone from being one of the best teams at limiting free throws to the absolute worst. That’s not a change you can wave away with the loss of Trevor Ariza.
When the problem is this severe, there are likely multiple causes. The Rockets core has gotten older, and the team is thin on its depth chart. Eric Gordon, for instance, is a player with a troubled injury history and a small guard about to hit 30-years-old, and Nene Hilario can no longer add much value to the team. With the loss of two rotation players from last season, the failed addition of Carmelo Anthony and the various modes of regression from age and other factors have left the Rockets in dire straits. They will likely rebound and play a bit better — Eric Gordon won’t keep shooting 31 percent from behind the arc, for example — but the roster has worn thin and the cracks are real.
The gift of Luka
This far into the season, I think we can safely say Luka Doncic will not become a bust. He’s been a revelation for the Mavericks, and not only does he look like a future star he’s having a truly valuable rookie season. He’s genuinely helping the Mavericks push for the playoffs, and he’s coming up with highlights like these two blocks of LeBron James at the rim. Halleluka.
Players of the Week: Paul Millsap and Kawhi Leonard
I thought I’d highlight this: Paul Millsap and Kawhi Leonard were the players of the week last week, and after both suffered through ailments last season I think this is a very positive sign. Leonard played well enough to defeat (the admittingly shaky) Warriors, while Millsap helped trounce the Lakers. Millsap isn’t playing heavy minutes, but it’s good to see him play well and not miss any games so far. As for Kawhi, I’m glad to see him on a contender and away from the slow-burning disaster of his situation with the Spurs. Sometimes you see an out-of-left-field winner for Player of the Week like Pascal Siakam, but they’re usually all-stars. It’s an encouraging sign, and let’s hope they remain healthy.
Derrick Rose for Sixth Man of the Year
Since I’ve been dismissive of Derrick Rose’s recovery in the past, I thought I’d comment on his continued success this season. I would classify myself as an empiricist. This is why I wanted to wait until well after his 50-point game to make any kind of declaration. Although I doubt he’s going to continue shooting 50 percent from behind the arc for the rest of the season, his percentages are up across the board and we’re going farther and farther into the season. What’s interesting is that he’s not doing anything appreciably different on the court or on the stat sheet, besides the higher percentages. He’s even taking the same proportion of shots he’s been taking these past few years.
But I think back to Grant Hill and his slow recovery. After his major injuries, he had trouble staying on the court and playing effectively, and few people honestly thought he would stick in the league for much longer. But he found a stable role for the Phoenix Suns in his mid-30’s, and even played at least 80 games for three seasons in a row. Sometimes it takes a while for the body to recover, and perhaps that’s what is happening here. I wouldn’t pencil him in for Sixth Man of the Year yet — his defense should hurt his case, but of course, that’s often dismissed for this award — but that may help us understand his season and his future in the league. Sometimes all you need is time.
CarmELO projections now with added nougat (and minutes)
Given that I’m currently writing a critique on 538’s methods, I thought I’d commend them for a positive change. They started factoring in actual player minute distributions and estimations. This is something I’ve been calling for in the past, and I’m glad a major site is finally doing it. It’s silly to consider that a major star could go down for the season and most team metrics don’t even change to reflect that. This is the data-ball era. We can do better. So let’s hope this was implemented well. Hopefully, someone will find a way to test their changes, because, and I know they’d agree, all good systems need to be evaluated by third parties.
Mitchell Robinson, from the air up there
I’ve written about a few rookies already and how they’ve lived up to their draft hype, but I should talk about a second-round steal who’s a dark horse to the Rookie of the Year/Luka Doncic runner-up award: Mitchell Robinson. Taken by the Knicks early in the second round, he was actually a higher-level recruit back in high school but was deemed ineligible to play in the NCAA because he enrolled at Western Kentucky and left early in the summer before the fall program started. Thus, he ended up not playing in any organized league for an entire year — it was a tough situation to judge and slipped in the draft.
After setting records in the summer league for blocks and offensive boards, we should have all been prepared for a solid rookie season, however. He’s leading the league in blocks per possession, and he’s hitting a high percentage of his shots. He’s the prototypical high-flying, rim-running big man in the mold of Tyson Chandler. But he does have a few unique qualities.
First of all, Robinson is an inevitable dunk contest entrant. He can get up as high as anyone, and he’s a nimble runner who can pick up speed quickly and has dunked from just inside the free-throw line with one dribble from half-court. But he’s an extremely low usage player who can only score close to the rim. His offensive rebounding is merely okay for someone of his type, and he can’t hit his free throws either. However, his vertical threat — lobs can be thrown over a defender’s head — does put stress on the defense.
On the other side of the court, Robinson is an impressive shot-blocker. That’s the obvious part; dude blocked nine shots earlier this season, and that included poor defenders away from the rim too. And look at the clip below. It’s not the amount of blocks that’s necessarily impressive but the amount of ground he can cover. For the negatives, he has a high foul rate, which is common for both rookies and defenders of his ilk. He has a tiny defensive rebound rate too. That’s mainly caused by his tendency to chase shots and playing next to rebound gobblers in Enes Kanter and Noah Vonleh. Remember that not grabbing a rebound doesn’t always translate to poor team rebounding — and so far the Knicks have been a little better with him on the court.
Overall, Mitchell Robinson is a player of extremes but he’s the type of role player who could become very valuable for a contender. He’s a dangerous finishing option off a pick-and-roll or a loose board, and his speed and vertical leaping ability combine for a tremendous defensive potential. For someone who never played in college or a professional league for a season, his progress has been amazing and it will be fascinating to see how he could fit in with a future Knicks squad with a healthy Kristaps Porzingis.
From the corner: Are preferences real?
With the rise of massive amounts of data in the NBA, there are many tools to visualize the game and player tendencies. One of the most popular is the shot chart, which is now ubiquitous. You can see a player’s range, how efficient he is, and where he’s most proficient. Back when these charts were popularized with the debut of SportVU data, you saw articles explaining that, for example, Klay Thompson was pure fire from the right corner and average from the other corner. But was that a property of the actual player or was it randomness? Was there value in seeing the right-and-left splits of a shot chart?
While most charts and articles don’t explicitly state that a player is better and will be better in a particular spot, it is at least implied and if not, what use is the information? I know there is value in statistics that summarize events or a series of events, but when a stat has no predictive power it’s an indication the stat does not accurately explain the underlying mechanics. It’s just a summary of what happened. You test the validity of ideas — i.e. is Klay Thompson shoot better from the right corner — by predicting with a different set of data. It’s okay to say a player performed better from a particular spot on the court, but going so far as to say the player is better is another question.
Here’s the obvious real-world application: when you scout an opposing team, you have access to reams of data. That may include right-versus-left shooting. If you present this data to your team, are you truly giving them useful information on how to guard them if there’s no predictive power those percentage differences will hold? Think about telling the team whenever I bring my blue water bottle the other team misses their free throws, and whenever I bring the red one they hit them. You could find some strange correlations if you look hard enough, but it doesn’t validate them just because you can quantify them.
Let’s set up the test I’ve been hinting at. I’m going to make this simple since this is a week in review piece, and it’s a first pass at looking at this question. I’m going to grab every player who took corner 3-pointers from the 2015 and 2016 seasons, calculate their differences from right and left, and then see if that correlates to the 2017 and 2018 seasons. If this right-left preference is inherent to the player, we should see a correlation.
For some simple filters, I’m only looking at players who attempted at least 50 shots from both corners during both time frames. It was still a hefty set of players — 49, to be exact. Also, this data is taken from stats.NBA.com, and as it goes without saying, shot data is often imperfect but with a big enough sample I’m confident in the results.
Look at the table below with the players who had the most extreme splits from 2015 to 2016. The only solid pattern is that most players became right-dominant from 2017 to 2018 with the exception of Harrison Barnes, who was actually right-dominant the previous two seasons. Look at Tim Hardaway Jr. for instance. He shot 44 percent before 2017 from the left corner on 68 attempts, and a paltry 28 percent from the right corner. What happened in 2017 and 2018? He hit 39 percent and 46 percent, respectively. The “preference” inverted.
Table: Corner 3PT% differences
Taking all the numbers together, the correlation coefficient was a weak -0.14. Not only is that mediocre, but it’s a negative relationship, meaning shooting better from the right meant it was more common to shoot better from the left in the next two seasons.
Using the t-test (two-tailed, equal variances), the P-value was 81%. For those not familiar with the stat, the usual significance levels at 5% and 1% — you have to be under those marks to qualify. There’s a lot of rightful criticism at the use of a P-value, but this is a case where the results are so lopsided that I’m certain with the results. The differences in right and left corner percentages appear to be noise.
A smart counter would note that 3-point percentages are inherently noisy, so as a sanity check I looked at the correlation of right corner 3-point percentages alone, not the difference. The P-value dropped to 8%, and it became significant once you factored in overall 3-point percentages.
Another possible issue is that I’m only looking at players with enough attempts from both players, and a true right-left preference shooter would likely not have enough attempts from one side. That does sound like a selection bias, but it’s harder to test that because with a low number of attempts you get noisier numbers. It does warrant future investigation, however. And perhaps with a larger sample of players I could tease out more edge cases who do have that preference.
As it stands though, whenever you see stats showing right-left shooting splits on the court, take them with a grain of salt. They may not be showing what you think they are because the vast majority do not retain those split percentages from one season to the next. Klay Thompson in that aforementioned article shot 50 percent from right corner and 39 percent from the left (you can see that here.) Has that trend continued? Well, he did shoot 54 percent in 2018… from the left corner. And he’s already switching that this season, shooting a little better from the right corner again. We can concoct stories all we want about these stats, about why these player shoot better and what it means, but it’s like seeing a die roll a six twice in a row and calling the third roll with certainty. We’re smarter than this.
Just don’t leave Klay Thompson open, whether it’s on the left or the right.