In a week, the voting for the starters for the All-Star Game will be complete — the season is already crumbling away. That acceleration in time passing feels the same as the one we experience when aging, and it’s probably not a coincidence. Researchers postulated that life “goes faster” when you’re older because there are fewer new experiences, thus there are more routine actions that are quickly forgotten or discarded.
That happens during the season — the first couple weeks are exciting as we see all these new combinations, new players, and the stats are all wild and unsettled from low sample sizes. By the time we get to January, there are fewer surprises, and it appears as though, in retrospect, that the season is moving faster and faster. We’ll get to the end of the season soon enough, but enjoy those moments in between and find new experiences for watching, new players and teams to explore — it’ll last longer.
The James Harden injury
The big news item of the week was an unfortunate injury to an MVP candidate: James Harden will be out for at least two weeks with a hamstring strain, a notoriously chronic issue that can linger for longer periods of time than first estimated. Of course, few teams are this well equipped to survive without their MVP-caliber player — Golden State is one of the few in history who qualify, as the Stephen Curry injury was a mere blip in their season — as Chris Paul will play full-time orchestra leader on offense for now. As we’ve seen in previous seasons with the Clippers, Paul and his team do well when another star gets injured, like when they went 30-15 without Blake Griffin in 2016.
Evaluating how well the Rockets will do without Harden is a little tricky, however. They stormed the first part of the season with the league’s best record and point differential going deep into December, but they’ve been slumping lately and that was actually before Harden got injured, which you can see in the graph below. They’ve been reeling since that ugly Lakers loss on the Dec. 20. That snapped their win streak, and they haven’t been the same since.
Note that the numbers above are adjusted for opponent strength and homecourt advantage. Sometimes wild swings happen within a season, and it’s without a rhyme or reason. Both Harden and Paul missed time recently, sure, but even with both guys they still had embarrassing losses to the Wizards and Lakers. Their defense has fallen back to Earth, which does make sense — they aren’t built like an elite squad on that end of the court. If anything, this should give you pause before you make grand proclamations on a player based on a small stretch of games when he’s out. There are a lot of variables out there on the court, and it’s hard to isolate just one.
The NBA and AI (no, not that AI)
As the NBA embraces the databall era by digging into player tracking data and hoisting up an incredible number of 3-point shots, much of the scientific/mathematical world is transfixed by the power of artificial intelligence in increasingly varied settings. Notably, a new chess algorithm is king (for now), AlphaZero, and it was built with no prior human knowledge. In other words, its intelligence in chess was created by playing itself and learning — no heuristics or input required. And, quite interestingly, AlphaZero was described as having an “alien” style of play.
What influence will AI have in the NBA? We can’t pit NBA players against a computer directly, but more and more decision making could in theory be made by artificial intelligence. What strategies could an AI conjure? What will the future look like? Some of us are still pondering whether or not it’s good to shoot lots of 3-pointers, but progress can be quick. Imagine plays conjured by a computer program, ones with no antecedent in basketball history. The future could get weird.
Another week, another argument for Nikola Jokic
The Denver Nuggets have had a relatively quiet season. Outside of die-hard fans and people in Colorado, few people talk about them. They’re doing okay enough to be in the playoff race in the West, but not well enough to make waves. I’m afraid that’ll keep Nikola Jokic from making his first All-Star team, even though he absolutely deserves it. He doesn’t have the immediately obvious stats, like points per game or flashy blocks numbers. Instead his numerical dominance is subtle as a big man with extraordinary passing, elite efficiency with range past the 3-point line, and a surprisingly good defensive impact. Every year his teams do much better on defense and rebounding (and offense of course) when he’s on the court; this is a real pattern. So let’s please make sure he’s ultimately selected for the All-Star Game — look at those plays below. Wouldn’t you want him?
The defensive transfiguration of Karl-Anthony Towns
Karl-Anthony Towns crashed onto the scene at full speed as a rookie, but his development has been lacking, and a lot of that was due to his defense. It’s okay if defense isn’t your strength when you’re as talented a scorer as he is, but the results were awful and it was disappointing given his skill and agility for his size. However, he has improved over the last few games, and the signs are encouraging. He’s always been blessed with great lateral quickness — remember his vaunted defense against Stephen Curry at the height of the MVP’s powers? — so it’s encouraging to see him use it more. For example, here’s two blocks he had against Kyrie Irving, that trick-artist of a scorer, using his exceptional length, foot speed, and timing.
Blocking a shot isn’t the best measure of defense, of couse; it only accounts for a small fraction of all basketball plays. Here’s an example of a play where Towns affects Kyrie’s shot, forcing him into a tough floater. He also had a recent game against the twin towers of DeMarcus Cousins and Anthony Davis. He was mostly matched up against Davis — you can see him defend Cousins well here though –and he held his own, as Davis had 16 points on below average efficiency. If Towns can keep using his length well, and if he continues to stay active and engaged on defense, he can legitimately be a plus on that end of the court. It would do wonders to catapult him from a promising star to the superstar we all hope he becomes.
Lakers exceptionalism
After a hot start, the Lakers have settled into being the team we (minus LA fans) all thought they were: a cellar-dweller. That’s fine if you’re a rebuilding team with a stash of young talent, but the team once again assumes it’ll be a major player in free agency this summer and can bypass the painstaking, patient work of rebuilding. This “Lakers exceptionalism” thinking was born out of decades of success, but it’s dangerous when they no longer possess the kind of competitive advantage they had in the past.
The league has gotten smart, and stars are fine playing in other markets. I think they’re also mistaking the savvy and luck they had in the past, like the trade for a number one pick that ended up being James Worthy that was so bad the NBA created a new rule to help prevent it — this also involved the Lakers winning a coin flip to pick first over the Clippers — or when they fleeced the Magic of Shaquille O’Neal, for pre-destiny. They need some objectivity.
My totally useful All-Star voting
The All-Star voting for the starters ends this week, so I thought I’d highlight a few things and put in my two cents while there’s still time. I’m under no illusion I’ll tip the scales, but if there’s anyone still on the fence about something, I’m here. First, let’s start in the East and look at the guards. You only need to pick two guys, and according to the voters it’s obvious: Kyrie Irving and DeMar DeRozan. They’re the two top scorers of two of the elite teams, so it adds up. But I gotta break from the pack here. Although DeRozan is finally changing his shot types and scoring more efficiently, there’s plenty of evidence to suggest that Kyle Lowry is still the most valuable player, which is the story every season and it’s one I’m comfortable committing to. There are some decent arguments for others too, fun ones for Spencer Dinwiddie and Victor Oladipo, who will definitely not keep up his scoring and 3-point shooting percentages, but Kyrie Irving is still a safe bet. No one else is overwhelming.
For the frontcourt players, two are spectacularly easy: LeBron James and Giannis Antetokounmpo. For the third player, it gets tricky. There are a number of decent options at center, and most’ll probably be invited regardless: Joel Embiid, the current leader in votes; Al Horford, the guy on the leading team; Kevin Love; Kristaps Porzingis, the scorer; and Andre Drummond. Going only by the stats, Al Horford stands out — unless you want to go weird and vote for Robert Covington here. Joel Embiid is fun to watch, and he’s probably my favorite player to watch right now, but he’s played fewer total minutes than most stars and I don’t think his offense is quite as great as most think — he takes commanding control of the ball on offense, for better or worse sometimes. Porzingis has a similar problem. Horford is the safest pick, and I don’t mind going with him.
In the West, the guards are obvious, despite their injuries: Stephen Curry and James Harden. The only other guard who’s played as well this season is Chris Paul, who missed a large stretch of games. There’s a case for Russell Westbrook, but his efficiency has been sub-standard and most metrics are overrating his defense — he’s probably getting lucky on those on-off stats, and his box-score stats overstate his value there. I’ll throw in some love for Jimmy Butler too, who should be considered for small forward as well, and therefore “frontcourt” players — he could have slid in with the next group.
As for forward, it gets trickier. I’m comfortable going with Kevin Durant here with no deep thoughts necessary. After that, it’s a crap-shoot. DeMarcus Cousins is having a sensational season by the numbers, even if the Peicans aren’t. Anthony Davis is Anthony Davis. Draymond Green is getting overlooked, but he’s still a world-class defender who’s a plus on offense. You can argue for Kawhi Leonard, despite the missed games, because of how good he is and how he should be represented. Paul George is the only thing not wrong with the Thunder, and he’s been excellent on both sides of the backcourt line. Both Karl-Anthony Towns and Nikola Jokic were mentioned above — they’re worthy. LaMarcus Aldridge is getting some attention, but only because the Spurs are doing well and he’s the easiest guy to identify on a deep, parity-laden team. (They’re doing well because of multiple players, not because Aldridge has been better than, say, Jokic.) By the tiniest of margins, I’m going with Draymond, a versatile engine in a juggernaut, and Anthony Davis, who does appear to be more valuable for the Pelicans than Cousins and does a little bit more than Paul George.
My picks:
East
Kyrie Irving G
Kyle Lowry G
LeBron James F
Giannis Antetokounmpo F
Al Horford F
West
Stephen Curry G
James Harden G
Kevin Durant F
Draymond Green F
Anthony Davis F
Playing up, or down, to your competition
Myths persist because it’s easier to repeat them than to do some research and refute them. You may know about the infamous myth about dropping a penny from the Empire State building. We all accepted it: the penny would be dropped from such a great height that it would kill someone below. The mechanics had some logic to it — it’s metal and the Empire State building is 1250 feet tall. But few people have the time, and fewer still the ability, to refute the claim. Some people even still believe it, but it’s definitely not true as Mythbusters fans will attest. These beliefs persist until we have the collective power to overturn them.
There are plenty of myths in the NBA that are still lingering around. We’ve all heard a certain ex-power forward loudly state that jump shooting teams don’t win championships — and then we saw the Golden State Warriors take multiple titles. Then there’s the one that defense wins championships. You may be able to argue defense is more important, but both defense and offense win championships together. There are all sorts of beliefs about title teams, and I’ll focus on one of those today: does it matter when a winning team has a relatively worst record against other winning teams and beats up on lesser ones?
I can understand why people want to go beyond mere point differential and win percentage — and even beyond basic strength of schedule adjustments. That data definitely doesn’t tell the full story, and it’s fun digging into the numbers and finding interesting slices yet undiscovered. For instance, when criticizing a contending team, you may see someone cite their record against the other top competition. This suggests, supposedly, that this team is a false elite team and will crumble in the playoffs. The same is true vice versa; you’ll see similar thoughts on teams with weak records who “play up” to their competition. Many people use these numbers, even ones with a statistical background (although you do see some skepticism noted.) Even I’ve done it. But is there any truth in those numbers? Can you significantly play up or down to your competition?
First of all, while most people cite these kind of numbers with win/loss records, like Denver having a 6-10 record against playoff teams, I will go in another direction. Putting data into two bins, or maybe three or four, is inaccurate and rough. Instead I’ll make this data continuous with a different technique. I look at the expected point differential of a game based on a team’s adjusted rating versus the actual point differential, and then I look at how that correlates to how good the opponent is. Basically, if your team plays better against the best teams, then that team will have a positive correlation, and if they play worse than expected (according to the strengths of the two teams and homecourt advantage), then the correlation will be negative.
I’ve done something similar before, but I focused mainly on individual players and how they fared as the competition increased. I used the team “v-stat” as a way to denote how any stat, like scoring or team point differential, can vary with the opponent. It’s just a way to summarize an attribute in a number that’s more nuanced and granular than “20-10 versus playoff teams.” To make things simpler, from here on out I’ll call my new variable the team v-stat, and it’s roughly on a scale of -1 to 1 — teams with perfect inverse relationships to teams with perfect 1-to-1 correlation. That is, with a v-stat of 1, for every 1 point your competition increases by, that’s how much better you get.
My data-set included the 2005 to 2017 seasons, excluding the lockout season in 2012. To test whether or not teams playing “up” or “down” to their opponents I chopped seasons in half so I could see if a team’s coefficient changed or stayed the same from one half to the next. If there’s a strong correlation, then it’s indeed a skill. I considered expanding the data-set — I just need more code to adjust for fewer teams and lockout seasons — but the results were so strong right away I did not deem that necessary.
Moving onto the results, the correlation between one half of the season to the next was 0.0059 (on a scale of -1 to 1.) In other words, there was virtually no correlation. In fact, if you set a regression to predict the v-stat from one half of the season to the next you’d get a coefficient of 0.0012 with a standard error of 0.053 — it’s not great when your standard error is about 50 times larger.
Table: top and bottom teams in team v-stats
| V-stats | First half | Second half |
| Top ten | 0.323 | 0.002 |
| Bottom ten | -0.362 | -0.020 |
You can see how you can get a p-value of 98% with the table above, which shows the most extreme teams in the first half of the season and how they fared in the second half. A p-value of, say, 5% is typically the cutoff for significance and 1% is preferred; 98% is not ideal. This is just noise.
Next: Nylon Calculus -- Defining and calculating luck-adjusted ratings for the NBA
By the way, who had the best team rating “v-stat” in my sample period? The 2016 Golden State Warriors had the highest coefficient in a half-season — the first half — and it partly explains their record. They were good enough to mop up poor teams playing in second gear, and they ramped things up against the best ones. But their coefficient fell to near zero in the second half of the season, and they famously lost in the finals to the Cavaliers. How a team plays against the best teams has no predictive power, and, as Neil Paine also found with a different method, it’s not an indicator of how someone will fare in the playoffs.
This one is a myth — teams don’t play “up” to their competition. They just drive our narratives.