The 2018 MIT Sloan Sports Analytics Conference is being held in Boston Friday and Saturday. Several Nylon Calculus contributors are there covering the event and weāll be sharing our thoughts and reactions to panels and research papers. Check back throughout the day for updates.
Draft Day Analytics
While the NBA draft has traditionally been the province of basketball scouts, analytics has equipped front offices with new tools to identify and evaluate prospects. Models have their limitations, as Houston Rockets general manager Daryl Morey himself concedes in Michael Lewisās āThe Undoing Project,ā so how can they be employed alongside scouting reports to improve the entire process?
Moderator Cristina Daglas of ESPN kicked the panel off with this very question. A wide-ranging discussion ensued.
Austin Ainge, director of player personnel for the Boston Celtics, described a multidisciplinary process that involves not just analytics but also the scouting, sports science and other factors. He specifically highlighted player health as an important area ā something thatās seldom discussed in the public sphere due to medical privacy rules.
Gersson Rosas, executive vice president of basketball operations for the Rockets, elaborated on all of the information that must be processed. He noted that this kind of situation requires strong decision-making systems. You can see how teams that have clear philosophies and minimal internal silos can have advantages despite the fact that, as Ainge mentions, the āsame 30 guysā are āin every gym.ā
Jonathan Givony of ESPN and Draft Express fame wondered why NBA teams are precluded from scouting at lower levels like AAU. He said that it would help if for no other reason than by expanding the sample sizes with which they can evaluate players. Certainly, it would provide greater opportunities for front offices to understand how a specific prospect grows over time. In a similar vein, he identified the ānext frontierā as the āmental stuffā ā i.e., truly understanding the factors that help a player adjust to NBA life.
The panel ended with some discussion about how changes in one-and-done rules would affect teams. Rosas indicated that such changes would align college prospect evaluation work closer to the European paradigm. Givony said that many teams actually tend to prefer the status quo, since the alternative would essentially open them up to greater risk exposure. Regardless of the conditions, the panelists agreed that the draft preparation is an inherently difficult enterprise.
ā Positive Residual, @presidual
Bhostgusters: Realtime Interactive Play Sketching With Synthesized NBA Defenses
Player tracking data has become ubiquitous in NBA analytics but the deep applications continue to be generally restricted to actions that are taken before and after games. Using player tracking data for deep applications during a game continues to be something of an analytic holy grail.
Bhostgusters, the product of a research paper presentation by Thomas Seidl and Aditya Cherukumudi, could be a step in that direction. In the words of Cherukumudi, āBhostgusters is an intuitive sketching program that allows a coach to sketch out a play just like they would on a clipboard.ā From there the ādigitalā clipboard takes over ā it can animate the play, and demonstrate how individual defenders are likely to respond to the structure of the play.
This is specifically designed for an in-game application. Imagine a coach during a timeout, setting up a play for a last-second shot and getting to see how the opposing teamās defense is likely to respond to different sets. The system also generates an expected point value incorporating the offensive play drawn as well as the expected defensive reaction.
Obviously, the system is only as strong as the defensive modeling. Seidl explained just how much work has gone into making sure a myriad of factors are accounted for. āIf you only train your model on good defensive teams, you get good defensive āghostsā,ā said Seidl. Their defensive models also accounted for the specific defensive team and their personnel, defensive roles (a deeper assignation than just position), as well as player-specific (fouls, fatigue, etc.) and game-specific (score, time remaining, etc.) contexts in projecting how the ghost defenders are likely to react.
Given the underlying work it seems plausible, that the system could one day even do the work of designing an optimal offensive play as well, considering similar factors.
Next step ā robot coaches. Get ready for an NBA Finals featuring Bot Rivers against Mike DāAndroid.
ā Ian Levy, @HickoryHigh
High-Resolution Shot Capture Reveals Systematic Biases and an Improved Method for Shooter Evaluation
Four years ago, Darryl Blackport estimated that it took roughly 750 shots for a playerās 3-point percentage to stabilize. Itās an awfully high number, especially when you consider that only Stephen Curry and James Harden exceeded this mark last season. But it goes to show how difficult it is to get large enough samples that can truly help separate the signal from the noise and evaluate shooting skills.
This issue is at the heart of Rachel Martyās work. In her research paper, she endeavors to address it by using over 22 millions shotsā worth of data that were captured in high resolution by the Noah Shooting System. A few intuitive yet critical findings result.
First, on average, right corner 3-point shots are biased about 2.34 inches left of the hoop center. Marty estimates that this bias costs roughly 4 percentage points. A similar bias exists with left corner 3-point shots, although to a lesser degree: 1.05 inches right of the hoop center, costing the average shooter about 2 percentage points.
Marty believes that these biases exist partly because players are trying to avoid the backboard. There are, however, other biases that exist around the court (for example, shots that tend to go too long or too short, or have ādepthā issues), so they still have to be weighed. On the flip side, despite the large sample size, shooter handedness is currently not a variable thatās collected ā a gap that Marty acknowledges.
Potential practical applications abound. If teams have access to greater volumes of data, they can better evaluate shooting skills. They can tailor player development at a more granular level than they presently can. Shooting practice data may be better leveraged, albeit with the caveat that conditions differ from real games. Hereās yet another illustration of how technology can enhance our understanding of very minute parts of the sport.
ā Positive Residual, @presidual
NBA 2.0: New Rules to Transform the Game
At the 2018 MIT Sloan Sports Analytics Conference, the NBA 2.0: New Rules to Transform the Game panel reviewed a greatest hits collection of issues the league and its fans have publicly wrestled with the past few years ā intentional fouling, playoff seeding, draft lottery reform, and the increasing prevalence of 3-pointers.
ESPNās Kevin Arnovitz moderated the panel, featuring Ben Falk of Cleaning the Glass, Mike Zarren of the Boston Celtics, Rafael Stone of the Houston Rockets and Evan Wasch, Senior Vice President of Basketball Strategy and Analytics for the NBA. The group spent time winding through each issue some of the proposals the league has looked at to address them, as well as ideas that have come from fans and media members as well.
The NBA's Evan Wasch discussed the proposed change from conference-based to 1-16 playoff seeding. One other idea: keep the East/West seedings until the final four teams are set, when you can re-seed across conferences. #SSAC18 #NewNBA
ā Positive Residual (@presidual) February 23, 2018
Pausing for a few minutes on the issue of intentional fouls, game length and flow, Wasch shared specific numbers around how the league has defined the issue and how they have evaluated possible solutions.
Wasch discussed the idea of having players shoot a single free throw, to earn two points or three points depending on the shooting foul. His group estimated that could shave as many as seven minutes of real time off the length of a game. However, the potential value of such a move didnāt seem to make sense to the league.
āWe donāt think we have a length issue. Our average games are down to two-and-a-half hours this season and weāre comfortable with that,ā said Wasch. The implication being itās not just about length itās about improving the flow and the in-game experience for fans. In Waschās words, āItās about optimizing the time people spend engaging with your content.ā
To the issue of 3-pointers becoming the defining shot in the NBA, no one on the panel seemed particularly concerned with the trend. The straw man argument about 3-pointers is that homogenization makes the game boring but the panel seemed to unanimously disagree, with Falk summing up their points.
āYou have to defend the rim and the points that are geographically farthest from the rim. Strategically, you have to do something thatās very difficult and you have to make all these interesting tradeoffs.ā
This was a fun conversation, veering into all sorts of hypotheticals about how the league could look dramatically different on and off the court. But, in the end, Zarren had the most important statement of the panel.
āLife is really good in our league and drastic changes may not make sense.ā
ā Ian Levy, @HickoryHigh
No Stone Unturned: Keeping up with the State of the Art in the Public Domain
One of the defining elements of the analytics movement has been the growing chasm between work done in the public and private sphere. Teams, consultants, third-party companies, and tech start-ups often have access to proprietary data sources and more infrastructure. The insights they glean are kept private, used for the betterment of the organization or team.
At the 2018 MIT Sloan Sports Analytics Conference, Eric Tulsky, Manager of Analytics for the Carolina Hurricanes talked about the other side of that coin ā work done in the public sphere and how that work can be used to augment whatās being done in private.
Lately, a lot of public analytics works, both academic and non-academic, can be seen as an audition for jobs in the private sphere ā a key to gain access to whatās being done behind the scenes. However, there are some significant advantages that public sphere analysts have, including strength in numbers.
āItās really hard to sit off on your own and come up with something thatās better than anyone else,ā said Tulsky. The implication being that while quality matters when it comes to ideas, quantity helps make sure that there are enough quality ideas that are worth exploring.
āThis running conversation between hundreds of people on Twitter is fertile ground for new ideas, but itās not always the best place for advanced metrics and rigorous methods.ā
Tulsky also talked a it about the objectiveness that outside analysts are able to bring to bear on evaluating a teamās decisions. While bloggers and media members have a reputation for biases analysis the simple fact that they arenāt personally, financially and professionally invested in the outcome of the decisions. It may not appear that team bloggers would have more distance from the issues of an organization but simply being one level removed from those inside the organization allows for additional perspective. For example, the ability to focus on different issues than a team needs to be absorbed with. As Tulsky pointed out,Ā āThere are very few metrics that separate āwhat are we doingā from āwhy are we doing it.'ā
I would guess that almost anyone working in the public sphere would trade these advantages for the access, infrastructure and influence that comes from working inside an organization. Still, in thinking about analytics as a community of people working on similar problems, maybe itās worth taking a more holistic view of the symbiotic relationship between public and private, even if one side holds all the responsibility for implementation.
ā Ian Levy, @HickoryHigh
Previewing the 2018 MIT Sloan Sports Analytics Conference
āEarlier this week, The Ringerās Bill Simmons posted a July 2017Ā podcastĀ with Houston Rockets general manager Daryl Morey to commemorate the 12-year run of the MIT Sloan Sports Analytics Conference. The wide-ranging interview served as a great preface to the upcoming data extravaganza. Although it was recorded before James Harden and Chris Paul stepped on the court as teammates, it covered many of the topics that ultimately landed on theĀ conference agendaĀ ā perhaps apropos in light of Moreyās influence on both the event and the industry.
āCorrectionā was a term that featured prominently in the podcast. Morey relied on economic parlance to describe what happens in the competitive NBA ecosystem whenever a team exploits a market inefficiency and the rest of the league eventually follows suit. He also cited examples when teams āovercorrected,ā or went so far to pursue an advantage that the costs exceeded the benefits. He painted a picture of a dynamic landscape; in many ways, every front officeās job is to establish processes that allow it to make big bets when the time, place and conditions are optimal ā and to exercise enough discipline to withdraw when the opportunity has passed.ā
ā¦
ā Positive Residual, @presidual