NBA Week in Review 20: MIT Sloan Sports Analytics Conference

Over the past weekend, a few thousand people congregated in Boston to discuss analytics in sports at the MIT Sloan Sports Analytics Conference. It was the tenth such conference, and at this point it’s tough for critics to deny the role and influence these advanced numbers and analysis have on sports. This very website had several people attend. The revolution is over, and even the Lakers are embracing the new movement. So now that the league is listening, what do we say? The next step will be effectively communicating best practices and improving the league in several tangible ways, not just “shoot more three’s.” And with that, let’s take a look back at the past week and use some numbers wisely in dissecting the NBA.

Dinosaurs in Miami

Since Chris Bosh has been out with an injury, the Miami Heat have been seemingly unaffected. They’ve gone 9-3 from February 19 to March 12, outscoring opponents by 6.7 points per game. One loss was to Golden State and another was to Toronto in overtime. Looking at their team stats before Bosh’s absence, they were actually getting slightly outscored by the opposition on average. They had few other changes either; it’s mostly a major subtraction from the team. So how have they been performing significantly better over these past few weeks?

Usually when I do these investigations, I can latch onto a hypothesis and find some reasonable argument for why a number is a mirage — or that it’s perhaps a sign of real change. But what I’ve found is madness. The Heat haven’t been boosted by a Mickey Mouse schedule — the average opposing team SRS during that time period was roughly -1, which is weak but only slightly. There are no anomalous stats, like extreme three-point percentages. Instead Miami has been killing teams whenever … Amare Stoudemire and Joe Johnson are on the court. Amare replaced Bosh in the starting lineup, and there’s nothing special about his individual numbers over that stretch except for his plus-minus. Joe Johnson, meanwhile, has been shooting 57/63/90 splits — yes, over 60% from behind the arc. Joe Johnson has been starting too, and when the two are on the court the Heat have destroyed opponents with a net rating of roughly 24. For reference, that’s about the twice the net rating of the greatest teams of all-time.

What’s strange is that Miami’s three-point attempts have gone down and the percentage has essentially stayed the same, despite Joe Johnson’s torrid shooting. Where they’ve really improved is with their rebounding. This isn’t odd because Bosh is a nonentity on the offensive boards and he’s not a strong defensive rebounder either; Amare at least grabs offensive boards. A few more minutes from Hassan Whiteside probably helped too. But there is one sane, non-Amare/JJ explanation: Goran Dragic has played better since the break, scoring and assisting more often, even though his three-point percentage has dropped. With his improved play and Joe Johnson’s surprising shooting, Miami’s offense has survived Bosh’s absence, but I don’t expect the same of their defense. The Amare-Johnson duo is probably a small blip on Miami’s season, shaped by Dragic’s strong play and a few random factors. At least that’s what I hope — otherwise this is pure madness.

Sloan: Biometrics Talk

Andy Glockner held a talk at the Sloan Conference about the future of biometric data in the NBA. There are ethical concerns here, and it’s obvious why: tracking a player’s movements on the court is one thing. It’s entirely another to track how a player moves, sleeps, experiences stress, etc. off the court and out of the arena too. With how expensive players are and how much infrastructure is needed for a few hours of their performance, or just minutes, every week,  it’s easy to understand how players could be treated less than people and more as multi-million dollar pieces of equipment. A well-run organization will probably avoid this issue because the players will feel respected and a part of the team, but look for the day where a controversy occurs because a player does something he shouldn’t have done and was caught using biometric data. It could be relatively soon.

Brook Lopez: Nets Superhero

As I feared before the season started, the Nets season is bleak and uninteresting except for Brook Lopez, a lone shining beacon of slinky, ground-bound post moves and Simpsons fanaticism. Brooklyn is playing its way out of the bottom-three lottery spots, much to Boston’s chagrin, and it’s partly because Lopez is quietly having (arguably) the best season of his career. His stats are fairly in line with his career averages, but he’s been healthy and scoring well. In the past, his per game stats haven’t matched his plus-minus stats. In fact, during one of his best seasons, his second year, he had a multi-year RAPM of -3.52, which is downright atrocious. His offense hasn’t translated well to the team level due to his passing and lane-clogging ways, and his defense has always been an issue. This season, however, his plus-minus stats have been kinder, though they’re still far from star-level. His slow feet limit his defensive effectiveness, but he’s huge with long arms and his rim protection numbers have been solid to above average.

Brook Lopez had an excellent game against the Raptors, which the Nets lost but by only five points. He’s a skilled low-post player, of course, armed with slow but effective hook shots and other moves. He’s surprisingly good off the dribble too, and I never tire of watching him crouch down and drive inside at five mph to finish under the rim even though he’s seven-feet tall. He just has a great knack for nailing awkward looking shots, like this bank shot off the dribble while leaning in. Like Marc Gasol, he has pretty good range too even though he has more of a set shot than a jump shot. He has few defensive highlights, but he blocked three shots thanks to his length — few players have a higher standing reach than 9′ 5″. Brooklyn isn’t on national television much anymore, and the team isn’t high on the league pass rankings, but he’s an entertaining, unique player who’s having one of his best seasons in near absolute darkness.

Sloan: A Panel on How Not to Use Modern Data

Which one of these things is not like the other? There are a lot of smart, talented people who attend Sloan and hold panels, but a few seemed out of place: Vinny Del Negro, Mike Brown, and Scott Brooks. No offense to these basketball coaches, but I do not think of them as analytics experts. In fact, one comment they had was particularly troubling. Brooks openly said he would ignore the biometrics data for Russell Westbrook because he just seemed like he could bring it every night and had boundless energy. That’s the kind of flippant, ignore the information attitude that derails the best efforts of modern analysis. Human intuition isn’t a terrible thing and I’m not saying we should replace all our coaches with AlphaGo, but it is terrible to selectively ignore crucial data just because you feel like a guy has a lot of energy.

Jabari Parker’s Defense

Jabari Parker was not drafted for his defense, but it can’t be ignored either — he’s arguably one of the worst defenders in the league right now. Going through the basic stats, he doesn’t net a lot of steals, blocks or defensive boards for a big forward. He doesn’t do anything else that helps his team on defense as well, unlike a no-stats defensive star like Joe Dumars or Jason Collins. His plus-minus stats are significantly negative. However, digging into the video, he looks even worse on the actual court.

Usually when I try to find clips of poor individual defense, it takes a little while because many mistakes are too minor to show, or it’s a team-wide problem and crediting the blame is unclear. If you want to show a guy’s shooting stroke, just search for his jump shots — that’s easy. However, it was surprisingly easy to find examples of Jabari’s defensive problems even though he was usually hidden on a weaker defender. For instance, in this clip he leaves Jared Cunningham completely open in the corner. At the very least, it’s bad communication. In this next one, he lets Alonzo Gee go around him and finish at the rim; Jabari gives up halfway through and just watches him score. Guarding Ryan Anderson is difficult, and Jabari makes it more difficult here: he jumps in front of Luke Babbitt in a pick-and-roll even though he’s already covered, leaving Ryan open. Khris Middleton has to run at light speed to cover the gap and Jabari gets back a split second after. Ryan doesn’t shoot, but he dribbles inside the arc and pump fakes, completely throwing Jabari off. Now Jabari is behind Ryan Anderson and watches him pass and move inside, hanging out at the three-point line and guarding no one. Eventually he moves, only to have Alonzo Gee dunk over him at the rim.

The defensive system the Bucks use is not suited to Jabari Parker’s skills. He’s too slow and reacts as if he’s trapped in maple syrup. The team is probably allowing him to play through his mistakes because of his draft pedigree and his scoring potential. But he’s playing next to Greg Monroe at center quite often and it’s tough to cover his defensive mistakes; you have to hide him on poor scorers too. It’s no surprise the Bucks were drastically better on defense when he was off the court last year, and it’s perhaps why they’ve been such a disappointment this season: he’s been healthy and playing.

A Breakable Record: Single Game Blocks

On a recent episode of the Dunc’d on podcast, a listener asked which traditional single game individual record could be conceivably broken. One of those stats, rebounds, is completely unbreakable without a major change in the NBA — Wilt Chamberlain had 55, boosted by the high pace of his era and the large amount of misses. Wilt has the points record too at 100, which doesn’t seem reachable. The assist record is 30 — Scott Skiles — and it’s rare we see a player go past 20. But on the podcast, they singled out the blocks record as something that’s attainable[1. Unfortunately, Danny Leroux stated the the record was a mere 12 blocks, which was already accomplished by Hassan Whiteside in 2015. If it were 12, then yes, that’s very beatable.]. However, the record is 17 from Elmore Smith back in 1974. With Hassan Whiteside’s prodigious shot-blocking, is it even possible?

Looking at some of Whiteside’s totals from the past, and it’s easy to see why he’s the best candidate. He’s had multiple double-digit block games, and as can be seen from the histograms below he has a large spread in his single game block totals — that’s important for breaking a record. He’s been more impressive this season, however, with a higher average and multiple games with blocked shots of 8 or more.

For estimating the probability of an event, you need an appropriate discrete (read: whole number) statistical distribution. Basically, you need something that matches the frequency of Whiteside’s block totals. For people with some background in statistics, the inclination here is to use a Poisson distribution because it’s widely used. However, it’s too simple and has one unrealistic condition[2. The mean equals the variance for Poisson, which was not the case with Whiteside’s blocks. The variance is likely higher because blocks depend on minutes too and there are other factors like a differing strength of schedule.] that causes a terrible fit. Instead I’m using a gamma distribution with one clever tweak[3. Gamma distributions don’t use zero, so I just added one to every block total. It’s not a perfect fit since the distribution is for a continuous variable and not a discrete one, but the results appear to be solid with reasonable estimates even at the extremes. For example, the odds of a game with blocked shots are 3.1*10^-6% — that’s virtually zero.]. For people adverse to numbers, it’s just a distribution for positive numbers that’s a bit more adjustable. The fit isn’t ideal for the 2015 season, but Whiteside played fewer minutes and was more of a deeper bench player for the first part of the season.

Using the aforementioned gamma model, one can now find probability estimates for a certain number of blocks. With the record at 17 blocks, given Whiteside’s proclivity in 2016 over a 70 game sample he’d have a 1.0% chance at breaking the record. That’s non-zero, at least, but it’s not likely either. He’d also have a 1.9% chance of tying or breaking the record and a 3.6% chance of beating the post-merger record of 15, held by Shaquille O’Neal and Manute Bol. Using his 2015 stats, the odds are even more remote (nearly cut in half.) Given an optimistic outlook, Whiteside could have several prodigious shot-blocking seasons — perhaps he gets signed by a lesser team that wants him to start and play more often — in a row and maybe his overall odds at breaking the record could hit 10%. But he’s already 26, and I don’t trust his longevity.

There’s a question of how good my estimation is. After all, he’s had four double-digit block games already, and the model thinks that’s unlikely. The relationship is also complicated because his minutes are variable yet games are usually capped at 48. But he’s under-performed on blocks of 6 to 9 in a game, and this is a small case of a selection bias. I chose him because of his extreme stats, and when you pick guys at the extreme end they tend to regress to the mean — in other words, it’s possible he doesn’t have another 10 block game for a long time. But this is the best data I have and having odds greater than 1% for a record like this is pretty remarkable. In fact, if he does remain healthy and keeps swatting shots at this rate, he’d have a very believable chance at breaking the post-merger record with 16 blocks.

Framing the numbers a different way, he’d be expected to have 18 blocks every 7262 games and 16 blocks every 1932 games. Those are preposterous numbers, but no one else remotely has a chance. Anthony Davis is second to Whiteside in high block totals in recent seasons, but it’s not a close second. You can see his distribution in the graph below — his odds shrink drastically around 9 or 10 blocks. Over a 70 game season, given his block rates over the past three seasons, he’d have a 0.06% chance of breaking the record. Even with a 1000 games the odds only improve to roughly 1% — or, by basically, he’d be expected to break the record every 117,541 games. Just imagine the odds for a lesser shot-blocker.

I would not say the record of 17 blocks from Elmore Smith will likely be broken by Hassan Whiteside by any stretch of the imagination, but it’s at least possible.

Bonus: What Are the Odds of Isaiah Thomas Breaking the Record?

Isaiah Thomas is currently the shortest player in the league, and, not surprisingly, he has one of the worst block rates in the league over the past seasons (somehow he’s not first and Arron Afflalo has a worse rate). But he’d probably be the most shocking player to break a block record, so I was curious if the method would even work for him and what the odds would be. Using the last three seasons, you can see the distribution below. He’s only had one game with at least two blocks and the vast majority of his games have zero blocks. His odds at breaking the record over a 70 game sample are 0.00000002% or once in every 4,341,278,307 games. I wouldn’t say that’s likely, but it’s wondrous to imagine.