Wrapping up Sloan 2013: Larry Sanders, Stan Van Gundy, playing with leads, looking around, links

After four days spent in Boston to cover the 2013 Sloan Sports Analytics Conference, I came away with a bunch of things to write about, like the difficulty analysts (and writers) have in communicating their findings in a way that will get other people on-board, the problems with banking too much too soon on the incredibly promising optical tracking data making its way through NBA circles like nobody's business, and the advancements being made in analyzing player's physical, mental and team-relationship health. But there were still a few other things in the notebook that I needed to empty out so I could properly wrap up this year's conference — without further ado, here they are.


People love Larry Sanders, which seems reasonable.

The Milwaukee Bucks big man has become something of a favorite among League Pass diehards in his third year in the league for an amazingly aggressive style of defense that's led him to the top of the league in both blocks per game and block percentage (a measurement of how many out of 100 shots you reject while you're on the floor; Sanders leads the NBA at 8.9 percent, according to Basketball-Reference.com). Steve Von Horn at Bucks blog Brew Hoop has been celebrating the awe-inspiring nature of his stuffs; Grantland's Zach Lowe had the VCU product among his Defensive Player of the Year and Most Improved Player honorable mentions; CBSSports.com's Matt Moore sat Sanders down for the kind of deep dive on shot-blocking that's typically reserved solely for a master of his craft. Everybody, it seems, loves Larry.

Perhaps the largest development here, though, was Sanders' appearance in "The Dwight Effect," the research paper co-written by Kirk Goldsberry and Eric Weiss and presented at Sloan in the service of trying to establish a new visual and spatial framework for evaluating interior defense in the NBA. By analyzing a a database of 76,000 shots taken in the NBA over the past two seasons compiled by STATS' SportVU optical tracking technology, Goldsberry and Weiss aimed to answer two questions: Who is the best interior defender in the NBA, and which metrics would you use to find out/support that?

Through a pair of case studies intended to measure how well NBA big men prevented shots from being taken near the basket (the only area on the court from which NBA players shoot better than 39 percent), how well they reduced opponents' field goal percentage on close-in shots they did allow, how often NBA bigs got close enough to a shot attempt to contest it, and how opponents shot when said bigs were able to contest. In their analysis — which, they've granted, has its problems — Sanders grades out as the best interior defender in the league overall.

Sanders and the Indiana Pacers' Roy Hibbert each held opponents to a league-low 38 percent on shots inside 5 feet, which is nearly way worse than the 49.7 percent the league shoots from that distance with somebody in the neighborhood on average. He also allowed the lowest field-goal percentage (34.9 percent) when within 5 feet of an opponent's shot, which, again, is way worse than the 45.6 percent the league shot on average with a qualifying defender that close.

He didn't rank especially high on the measurement that gave the paper its name, though — while opponents tend to fear former Orlando Magic/Los Angeles Lakers center Dwight Howard enough to shy away from even attempting shots at the rim (only 48.2 percent of opponents' shots came near the rim with him out there, as opposed to 57.2 percent for the league on average), nearly 62 percent of opponents' shots when Sanders is out on the floor have come near the rim. That might not last long, though; during his presentation, Goldsberry said that representatives from the Bucks in attendance at the Sloan conference told him they're already seeing a bit of a reduction in the frequency with which opponents drive and attack the rim when Sanders is in the game. If the result of that learned aversion is more low-percentage contested mid-range jumpers, that sounds like a pretty good outcome for a Bucks defense that already ranks ninth in the league in points allowed per possession and is tied for 12th in opponents' field goal percentage, according to NBA.com's stat tool.

Plus, he draws nice pictures, even after a long layoff.

What's not to like?

(The real key, of course, is if Goldsberry and Weiss' analysis of which bigs are best at reducing efficiency up close and getting out to contest a high volume of shots winds up becoming something NBA teams look at more closely come contract time. I'm looking at you, Kosta Koufos.)


Stan Van Gundy's an analytics guy, whether he thinks he is or not.

During Saturday afternoon's basketball analytics panel, the former coach of the Miami Heat and Orlando Magic made a point of noting that his organizations didn't do a whole lot of analytics work while he was their head coach, that he found the predilection of "you analytics guys" toward pretending the NBA was like "playing video games" somewhat distasteful, and that he didn't see himself as someone particularly interested in analytics.

Methinks the coach doth protest too much. Also, methinks he should get the chance to coach again, like, yesterday.


NBA teams with leads tend to change their play in the wrong ways (maybe).

I've given one of the NBA-related research papers presented at Sloan short shrift thus far — "Live by the Three, Die by the Three? The Price of Risk in the NBA," co-written by University of California, San Diego, economics professor Matthew Goldman and Microsoft researcher Justin Rao. In part, that's because it was the only one of the four NBA-specific papers that didn't work off the SportVU data, and it was a bit easier to discuss the big-picture optical tracking issue; in part, it was because I didn't feel like I totally got some of the "econometric" and "optimal team function" stuff in their paper, which I read on the way up to Boston, and their presentation, which I watched twice. (This is, I am 100 percent certain, a "me" problem.)

Basically, though, Goldman and Rao tried to figure out whether it was a better choice for NBA teams to shoot 2-pointers (which they hit more frequently but are worth fewer points) or 3-pointers (which are harder to make but more valuable each time you make them) at particular points in games. The decision-making process is, to a large extent, indicative of how teams measure the risk/reward of a 2 vs. a 3 at different times in the game. Three-pointers are more valuable to a team that's trying to erase a deficit, because they help you get closer faster, whereas 2-pointers tend to seem safer and more valuable to a team trying to hold on to a big lead, because they don't need to be bombing away ... except that, according to Goldman and Rao's research, they really should be.

Ever wonder why every announcer in the history of televised basketball has said some variation of the phrase, "It's the NBA — everyone makes a run?" It's all about the relative "win value" of the shot — meaning, how much more likely are we to win if I make this shot? For a team that's behind by 10 or more points, the answer is simple: You're more likely to win if you make 3s to come back than if you make 2s, because it doesn't take as many. For a team that's ahead by 10 or more points, the answer should be simple in the opposite way — you should want to take safer shots to maintain your lead, but instead you continue to fire away from long range, because you're far enough ahead to not worry that it could all go up in smoke. ("They should be risk averse, but they're going to a casino because they have a 10-point lead," Rao said.)

Goldman and Rao's most interesting takeaway deals with teams that have small leads. Again, going back to that central "win value" question, if you're only up a few points late in a game, the likelihood of you winning is greater if you make a 3-point basket than if you make a 2-point shot. But because most coaches fear losing what they have more than they value gaining an advantage, teams more often tend to tighten, becoming risk-averse when they should be something more along the lines of risk-neutral. As a result, the free-wheeling team trying to come back tends to be more successful in closing the gap, while the tighter leading squad tends to fall back toward the pack a bit.

In theory, at least. I guess it sort of boils down to, "Whether you're up big, up little, down big or down little, continue to run your stuff, like the Houston Rockets did when they tied a 3-point record against the Golden State Warriors a few weeks ago." (Provided, of course, your stuff is "looking to shoot 3-pointers rather than 2-pointers," which is true for more teams now than it used to be, but still isn't the case for everyone.)

Portland Trail Blazers general manager Neil Olshey seemed to like the idea when it was presented to him by Goldman and TrueHoop's Henry Abbott, but didn't necessarily seem totally sold on the idea that — independent of data on specific personnel on the floor, shot location of the 2 vs. the 3, etc. — this was the right idea all the time. Then again, it's probably not true that anything's the right idea all the time, except continuing to look for better answers. (Oh, and that if you can get a good 3-point shooter an open look, you should, no matter what time of the game it is.)


Let's start paying attention to which players look around a lot.

This wasn't basketball-specific, but one of my favorite things from the weekend was a presentation by Geir Jordet, the director of psychology at the Norwegian Centre of Football Excellence (which, apparently, is really a thing), of a paper he co-wrote with researchers Jonathan Bloomfield and Johan Heijmerikx about the field vision of soccer players — specifically, what happens in the period before a player actually gets to touch the ball, and how that influences what happens once he gets hold of it.

In his presentation, Jordet referenced a quote from Arsene Wenger, manager of Arsenal of the English Premier League (EPL): "There are some special players who always find openings. […] Those who see will get you wins. But there are not many players like this." And yet, there hadn't been many studies produced on players' vision in the context of live play; rather, most work on soccer players' ability to search their surroundings visually had been done in labs with very specific sets of stimuli. Jordet and his colleagues considered the "visual exploratory behaviors" of EPL players — basically, when they moved their body and head to better see their surroundings — prior to getting the ball, and how it related to their performance.

As it turns out, from the researchers' sample — 64 EPL games, 118 players and 1,279 relevant situations from Sky Sport's split-screen PlayerCam broadcasts — players who register more "explorations per second" tend to be more successful and more heavily decorated stars than those don't look around quite as much. We're talking about major differences in how often the high-frequency players connected on their pass attempts (about 81 percent) compared to low-exploration players (about 64 percent), with even larger margins when you're talking about forward passes and forward passes in the opponent's half of the field. According to Jordet and company, the more often you look around while moving without the ball, the more likely you are to do something awesome once you get it.

After hearing that, all I could think about was: Who are the "high-exploring" players in the NBA? Ricky Rubio and Manu Ginobili jumped into my head, but that might be some latent guys-from-soccer-playing-nations stuff. What's Chris Paul's explorations-per-second like? What about Steve Nash (more soccer bias), Rajon Rondo or LeBron James? And conversely, who are the "low-exploring" players in the NBA? For some reason, I can't shake the image of J.J. Hickson throwing this pass, but that's not really fair. Given what seems like a reasonable relationship between field vision on the pitch and court vision on the hardwood, though, I'd really love to see some enterprising analyst try to adapt Jordet's study to an NBA data set.


Some stuff to grow on.

A million people wrote a million smart things this weekend; this is not intended to be a comprehensive list of links. That in place, here's some stuff worth checking out:

SB Nation's Paul Flannery on how we can actually start using advanced stats in our regular NBA chatter without everyone sounding like they're talking down to everyone else.

TrueHoop's Kevin Arnovitz on how the least cool place on the planet became the place to be in just seven short years.

SI.com's Rob Mahoney on the importance of being patient with SportVU, even though we all want our flying cars now.

The Grantland Channel features short videos/podcast segments from a bunch of basketball analytics folks, plus Zach Lowe chatting with Daryl Morey, among other things.

The Brooklyn Game's Devin Kharpertian on how Brooklyn Nets center Brook Lopez fits into "The Dwight Effect" structure of measuring interior defense.

The Wages of Wins Journal's Andres Alvarez on, well, just about everything from the conference. (The shirt shots, especially, are a nice touch.)

NBA.com's John Schuhmann chats with ESPN writer-turned-Memphis Grizzlies executive John Hollinger about, among other things, how his approach to analytics has changed after moving from the press box to the front office.

More from the 2013 MIT Sloan Sports Analytics Conference:

Needles, haystacks and failures to communicate: The challenges of advanced stats
Optical tracking data has us asking better questions, but answers remain elusive
Next big thing in NBA analytics? Moving from what we can see to what we can't