Chasing Perfection: A Behind-the-Scenes Look at the High-Stakes Game of Creating an NBA Champion

by Andy Glockner

  Of course, it’s difficult to compare a guard (Bejarano) who is forced to be a primary ballhandler because of injuries and roster construction with a big man (Avila) who gets the ball in more favorable scoring positions (and also was a good passer from the post and off the bounce), and that is the kind of conversation Glover wanted to promote with his data e-mails to Eustachy and his assistants without making things too confrontational or complicated.

  “That would be a good argument or discussion to have in the coaches meeting,” Glover said about the Bejarano/Avila comparison. “Hey, how can we get the ball into JJ’s hands more so we can be a more efficient offense. Daniel, for example, he has the ball in his hands a lot, but the percentages are not adding up for the touches he had. [Backup point guard] John Gillon, he was up there with JJ; typically, when they have the ball, good things happen for our team compared to other guys, and that’s something that I may put in a report because, again, I don’t want to overwhelm him.”

  The 2014–15 Rams were an interesting test case for the introduction of SportVU because in certain ways, the way they played stepped well outside of Eustachy’s long-established coaching footprint. Pomeroy’s data goes back to the 2001–02 season and, prior to last season, a Eustachy team in his database had never taken more than 30 percent of its field goal attempts from behind the arc nor finished higher than 234th in the country in terms of the percentage of field goal attempts that were 3-pointers. That changed with the roster in question, as Colorado State took 36.6 percent of its shots from 3-point range and connected on a reasonable 34.8 percent of them. Four of their five primary 3-point shooters made at least 36 percent of their attempts, so the higher rate of shots from that distance, on the surface, made sense.

  Things got more complicated for Eustachy because of what happened at the point guard position. Grambling University transfer Antwan Scott was presumed to be the starter at that position heading into the season, but he suffered an early foot injury, was never healthy, and ultimately played in just four games before missing the rest of the year. As noted earlier, sliding Bejarano over to the point had its limitations because he was not a great passer and also, relatively speaking, was a volume shooter who didn’t shoot very well from inside the arc or get to the free throw line to bolster his decent 3-point marksmanship.

  That left Eustachy with the option of Gillon, a transfer from the University of Arkansas-Little Rock. Gillon had a solid assist-to-turnover ratio, may have been the league’s most effective penetrator off the dribble, shot 39 percent from the arc, and got to the free throw line a ton for a guy who only played around twenty minutes a game. Despite Gillon’s seemingly solid offensive contributions, Eustachy—whose reputation centers around rugged defense and rebounding—had a hard time trusting the diminutive Gillon (who was six inches shorter than Bejarano, who was an excellent rebounder for a guard) on the defensive end of the floor. This was a great situation into which to introduce some of the SportVU data and create a discussion.

  “As coaches, you tend to get things in your brain and you tend to see things, you see almost what your brain is telling you to see,” said assistant coach Ross Hodge. “Like, John Gillon for example. We always had it in our head that he was a poor defender, he’s a bad defender, but then you get into some of the SportVU stuff and some of the Synergy stuff, it’s like ‘He’s grading out pretty well.’ He may not be in a stance, jumping to the ball, closing out like crazy, but his man don’t score. Or he gets over the ball screen. It may not be what you want, or how you want it [but it works].”

  That said, the Rams’ staff also had to figure out how and when was the best time to send information to Eustachy for his consideration.

  “Coach is a timing guy, like most head coaches. Will will hit him with the stats. That’s the best way to do it, we feel like, because usually it will give him a chance to be at home,” Hodge added. “He can look at them, digest them a little bit, not feeling like it’s an argument. Not coming in, feeling like ‘I don’t give a [expletive] what you all say, John Gillon can’t guard.’ And then you’ll be like ‘Well, coach . . . ’ It just gives him some time to digest it. Sometimes, he’ll come in and won’t bring it up at all. Sometimes, we’ll be in there watching film as a staff, and twenty minutes later, out of nowhere, he’ll be like ‘Yeah, I saw that on Will’s stats.’”

  Hodge said the new data they had at their disposal also helped the assistant coaches make a case to Eustachy for a different way to defend ball screens last season. Instead of having the handler’s defender jam up and then go under the screen, or have the screener’s defender hedge the screen to try to slow down the dribbler, they went to more of a soft drop from the screener’s defender so they didn’t end up allowing as much penetration and an easy kick for an open three.

  They also were able to provide more context to some of Eustachy’s preferred rebounding stats and tactics by showing that certain players were actually going hard for rebounds despite less-than-impressive rebounding numbers. They were collecting a high percentage of their realistic opportunities and/or the Rams were getting the rebound a large portion of the times a player was within a few feet of the rebound.

  This is still a work in progress in Fort Collins, especially since Colorado State doesn’t pay for SportVU use for practices, so the data set is limited to the Rams’ home games, many of which come in non-conference play and can be somewhat lopsided affairs. But the Rams are bullish on what they were able to accomplish in their first year using the SportVU system; they say it has added just a little more clarity and insight to their staff discussions.

  “If you didn’t have the data, and it’s just arguing what your account of events is, that’s when it can get difficult,” Hodge said. “When any person sees it one way, and you see it another way, basically you’re just arguing what you both see, and you could both be seeing it right. And [SportVU clarifies] a little bit of the ‘maybe this is happening. Or maybe this isn’t happening. Maybe we’re seeing it wrong.’”

  Duke operates at a different level than Colorado State. The Blue Devils have the highest-paid college coach in the country in Mike Krzyzewski, one of the largest overall budgets in the country, and one of the most national presences of any college program in America. So when Duke was one of three schools to first dive in to SportVU in 2013, the Blue Devils also made sure it was installed in their practice facility. Through the end of the 2014–15 college season, they were the only team in the country with that privilege.

  As noted in the Colorado State section above, sample size is one of the principal issues in effectiveness of data mining and analysis at the collegiate level. Even if you have the system, it’s almost certain at this stage that your opponents will not, so teams that use SportVU for games are only able to collect data from their sixteen or seventeen home games a season. Since maybe half of those are against non-league competition, with a good percentage of those games being “guarantee games” against overmatched opponents grateful to receive a high five-figures check for playing, there is a lot of noise in the data provided, if you even have the proper staffing to analyze it.

  By having the system also available for their practices, the Blue Devils are vastly increasing the number of times it can be utilized within a season, even if the primary purpose of the analysis then becomes self-scouting from practice sessions rather than understanding what has worked better against opponents. As the competition in practice may often be better than what a blueblood program will often face in nonconference play, there’s additional merit to what’s being analyzed from practices.

  Krzyzewski was reluctant to discuss exactly how Duke is using SportVU and how the system has aided the Blue Devils, but his former assistant Wojciechowski was willing to provide a bit of detail on how the Blue Devils had started to explore its uses.

  “It’s one of those mechanisms that has tremendous potential to really take a look at the game from an analytical standpoint at the very highest level,” he said. “And I think the
challenge for us at Duke, and for anybody who uses that, [was] how do you use that data to generate things that are useful on a day-to-day basis to impacting the team? And I think that’s the challenge of it. Again, it’s in the infant stages, especially at the college level, of using it and figuring out, ‘How do we take this slice of the pie, the analytics, and make it work for us in a positive and meaningful way?’”

  Wojciechowski acknowledged that much of the value from the system came from breaking down the five-on-five sessions during practice, or self-scouting, since there are many more practices than games over the course of a season. The Blue Devils, per Wojciechowski, were able to use the SportVU data from practices for a variety of purposes.

  “I think some of the things that you look for, or you want to kind of reinforce what your eyes see with data, are like rebound chances,” he said. “How many times does a guy have a chance to rebound the ball? And then how many times they go after it, how many times do they get it? I think that’s one thing. Shot selection. Where are guys actually an efficient and effective scorer and what type of shots do they produce and score them at a high percentage, and what type of shots do they shoot at a low percentage and what areas of the floor they are? Those I would say are two examples that we used [in 2013–14].”

  A December 2014 column by Barry Jacobs in the Charlotte Observer noted that Duke was using SportVU data to break down the team’s in-game perimeter shot selection as the Blue Devils were learning to play with gifted freshman low-post scorer Jahlil Okafor. Like many teams, Duke tries to limit the number of lower-success shots it takes, and having a post player as dominant as Okafor made that mandate even more important for the 2014-15 Blue Devils.

  Per an EdTech article from March 2015, Duke ended up making 41 percent of 383 catch-and-shoot 3-point attempts during the 2014-15 regular season. Those shots generated 1.23 points per attempt, which was relatively close to the 1.33 points per attempt posted by Okafor (who made 279 of 420, or 66.4 percent, of his shots without attempting a 3-pointer). Add in the impact of floor spacing and optimizing Okafor’s scoring load, and those are good shots to be taking. Conversely, according to the EdTech column, Duke only made 33 percent of its threes attempted off the dribble. That is not a good enough percentage to be taking those shots over giving Okafor low-post touches (or finding better catch-and-shoot looks), so those were shots Duke wanted to try to avoid as much as possible.

  Per Wojciechowski (and numerous other college coaches), the ability not just to provide stats to the players, but to show specific breakdowns of where and how shots were coming from—paired with graphics or video to hammer home the point—is the best way to impact players.

  “How you package the data, especially when you’re dealing with college-aged kids, is important,” he said. “For us [at Duke], it was more of a verbal [thing], talking about verbally what we saw, and then backing up with evidence of the data. Whether it’s a ‘You’re not going to the boards enough’ or ‘You’re taking low percentage shots,’ it just gives another layer of proof that what you’re saying is true and why you’re saying those things to them. And I think in today’s day and age with kids, it’s why are you saying that and having as many layers as possible to kind of prove a point and to teach them.”

  Wojciechowski was more circumspect when asked about how he planned on using the Bradley Center access to enhance his Marquette program, but noted that because of last season’s roster turnover from the year before, it was almost like starting fresh with a new set of data since the stuff from the season before wouldn’t be particularly helpful. He also added that, as a newer coach taking over his first college program as a head coach, he would be more apt to revisit some of his decisions after being presented with the enhanced data SportVU can provide, in addition to using the output to back some of his predetermined thinking.

  “In Coach [Krzyzewski]’s case, he’s done it for so long. His feel and his eye for the game, when you have a guy that’s at a master level of coaching, mostly it’s used for validation,” Wojciechowski said. “But mostly for younger coaches who are trying to establish a system and get know new personnel, I’ll use it both ways. Hopefully, I’m validating the decisions I’m making, but I think there will be times more so than with Coach where you’re taking a step back and asking yourself, ‘Is this the right thing?’”

  When asked, though, whether he recalled a specific instance where SportVU had altered his thinking, Wojciechowski demurred.

  “Off the top of my head, from my experience with it, no, not to this point,” he said with a laugh. “I assume that that day is coming.”

  A program doesn’t even have to have SportVU, though, for it to have an analytics basis. And analytics don’t just have to be about on-court performance for them to be very effective.

  Buzz Williams is one of the more unique characters in the college head-coaching ranks. Williams himself is extremely comfortable with statistics and analysis, but he’s also a master motivator and a voracious information hound. Williams is a huge fan of NFL coaches like Jon Gruden (who turned to a career as an analyst for ESPN’s Monday Night Football) and Gus Bradley, the head coach of the NFL’s Jacksonville Jaguars, for their creativity in thinking and motivating. Williams also says he has offered to pay reporters for their leftover interview notes from feature stories he finds interesting.

  Williams, both at Marquette (where he did actually have a year of SportVU access through the Golden Eagles’ games at the home of the Milwaukee Bucks) and now at Virginia Tech, also tends to cull from a different pool of recruits than many other top-tier coaches. He often dips into the junior college ranks to grab one- or two-year contributors to fill holes in his roster, believing (a) that the growing number of transfers each season in college basketball makes building around freshmen a bit riskier, and (b) that he’s not going to get blue-chip recruits to come to Milwaukee or Blacksburg, Virginia, and needs to craft a team with a hard-nosed edge in order to compete, first in the old Big East and now the Atlantic Coast Conference.

  Putting together rosters that way also shortens the time that Williams has to make things cohesive. He doesn’t have the luxury of leaning on full recruiting classes that stay together for four seasons and increasingly understand what he wants from his team. Not that every year is a reboot, but there’s a significant learning curve involved with playing for him, and Williams is aware that different handling is required of today’s college players. Mostly gone are the days of coach as autonomy figure, so Williams needs to work harder to get his messages to sink in, and he likes using data and descriptive/visual presentation to help make his points.

  “I think that the world’s society, their attention span is much, much shorter than ever before,” he said. “So I think when you’re trying to teach, you have to do it with evidence. Because I think the way of how coaches from a generation above me once coached, I think that’s over in some regards. In that, ‘Well, the head coach said this, and that’s it.’ And everybody just takes it as if that’s the gospel. Not saying that’s completely gone—I think the position is still the position—but I think when you’re talking to your staff, I think when you’re talking to your team, there has to be some sort of evidence of why you’re saying what you’re saying.

  “These kids can’t listen to a diatribe for twelve minutes of ‘Hey, this is how we do it.’ You gotta explain why,” he added. “And anything you can use from data—and I understand you can twist it and skew it however you want, but it’s the best, in my opinion—it’s one of the best ways to teach. Because it’s something that kids can understand. It’s something that you can hold them accountable to. And it’s also something that not only can you get it as it relates to a stat, but you can you use video to show it, you can use stats to show it.

  “And so, I’m always looking for ways to take opinions out of things and say ‘These are facts. And from these facts, this is gonna determine how we should play, why we need to play this way, who we need to recruit, why we need to
recruit them, etc., etc.’”

  Williams loves to dive deep into advanced statistics, but understands that the meaning and usefulness of them can get lost in the message if it’s not delivered in an accessible way. “I think in order to magnify something, you gotta simplify it,” he said. And perhaps the most unique way that Williams shapes and simplifies his messages while building credibility with his players is through a weekly series of offseason chats he holds for his team. At Marquette, they were called Life Lessons. At Virginia Tech, they’re called Tech Talks. Williams focuses each talk on a specific topic, which run the gamut from life skills like personal finance and understanding health insurance, to the more academic, like how different people’s brains are wired differently.

  One purpose of the talks is pretty straightforward. Williams wants to help his players understand each other better and also prepare them for life beyond the insular nature of college, whether they go on to professional basketball careers or out in the general workplace. The talks are also geared to have a significant side effect, though: Williams loads them with information and data, so when the players show they’re starting to understand the Tech Talk lessons and show increased eagerness as each week brings a new one, Williams knows they are more likely to buy in during the season when he uses analytics to make his points about how he wants the team to play.

  “If I have taught them things that they want to know, that utilize numbers, then their trust in me is, ‘If coach is talking about numbers, [he] knows what he’s talking about . . . ,’” Williams said. “You know what I’m saying? So, it’s stuff that they need to learn, it’s stuff that’s healthy for their life, but it’s also stuff that I build equity in what I’m teaching them, because when I get to December, there’s never going to be, ‘Well, Buzz is writing all those numbers up there. He don’t know.’ Because they’re not going to view it like that.”