The Past, Present and Future of NBA Scouting
NBA advance scout Pat Zipfel told ESPN's Brian Windhorst that he once had to sleep in Penn Station because "a game in New York went to overtime and [he] missed the last train out of the city and all the hotel rooms were booked."
They may not get VIP treatment, but scouts are indispensable in today's NBA. Their impact is felt in all 30 front offices, coaching staffs and locker rooms.
Advance scouts like Zipfel enable the league's best coaches to make their game plans and adjustments. They pore through endless video and information to prepare coaches for the competition. Personnel scouts, on the other hand, keep general managers in business by evaluating draft targets, prospective free agents and players who may be acquired via trade.
Whether watching players or watching plays, scouts have become the NBA's information processors. And much like the computers upon which they rely, these information processors have become infinitely more sophisticated over the years.
But that shouldn't take anything away from the classics.
"Godfather of NBA Scouting" Marty Blake passed away at 86 this past April, effectively reminding the NBA how it got to where it is today.
To fully grasp modern scouting, there's no way around taking stock of Blake's foundational impact on the discipline. He wasn't just the first real scout; he was the best at it for a long time. The New York Times' Bruce Weber shared a number of incredible reviews from the people who knew Blake's work best, including former Atlanta Hawks GM Stan Kasten:
Marty was our human database before the word database was invented. ... Encyclopedic doesn’t begin to describe his knowledge of the available crop of basketball players, and it didn’t matter where they played. Every single team in the league leaned on him for generations.
But before Blake, no one was collecting data, much less robotically retaining it. As Weber put it, "Scouting meant calling up coaches around the country to ask for player recommendations."
Blake traveled extensively, turning firsthand player observation into a science that remains a staple of the profession today. Without more than basic statistics to supplement his findings, the "human database" turned up star after star where others just weren't looking.
While working for teams like the Hawks (and eventually creating his own scouting company), Blake was inventing the job even as he lived it out, blazing a trail that would quickly become commonplace around the league.
The list of stars (and international players before they were fashionable) Blake discovered would be astounding for any scout. Even one with access to modern tricks of the trade, a trade now defined by the kind of statistical information Blake probably hadn't imagined in the 1950s.
How the Math Craze Changed Things
They're already calling it the "Summer of Analytics" in Phoenix, even though it's probably at least one summer too late. There's nothing like a three-year, $18 million deal for Michael Beasley to make a front office start looking in the mirror.
The organization thought it was getting an untapped talent finally discovering his potential. When given over 32 minutes a game in the first of two seasons with the Minnesota Timberwolves, Beasley averaged 19.2 points per contest and made a respectable 45 percent of his field-goal attempts.
That has to count for something, right? Maybe, but probably not as much if you're counting the right things—like "NBA Geek" Patrick Minton, who ranked Beasley's signing as the third-worst contract of the 2012 offseason.
He's extremely turnover-prone, does not defend well, is terrible at getting to the line and doesn't pass well. Well, let's be fair, he just doesn't pass. And he loves to shoot the ball. He doesn't pay a lot of attention to where he's standing when he does, though. Have fun watching him shoot with just one foot behind the three-point line.
True to form, Beasley underwhelmed in his first season with the Suns, ranking seventh among small forwards in turnover ratio, 52nd in assist ratio and 46th in overall PER. His 40.5 percent shooting from the field easily made 2012-13 the least accurate season of his career, and all of this inefficiency came in spite of career-low minutes.
While the organization probably didn't need an advanced metric to warn them about Beasley's off-court struggles, the Suns decided it was time for a numbers guy all the same.
New general manager Ryan McDonough reminded us immediately why the Suns wanted him in particular, saying at his introductory presser, "We’re going to try to be at the cutting edge... We’re always trying to find the next big thing."
You know—the kind of big thing that warns you to stay far, far away from guys like Beasley.
Before becoming Phoenix's latest hope, McDonough was Danny Ainge's assistant GM with the Boston Celtics, establishing himself as "part of a new breed of talent evaluators who have been making inroads into the highest level of the NBA in recent years," according to SB Nation's Paul Flannery.
McDonough wasn't the only representative of that new breed to cash in this summer.
The Philadelphia 76ers snagged a rising star from the Houston Rockets, hiring Sam Hinkie as their new general manager earlier this year. Guys like Hinkie (and former boss Daryl Morey) have become synonymous with the role of advanced metrics in scouting and all manner of decisions.
Nilkanth Patel writes in The New Yorker:
What Hinkie says sets him and his cohorts apart is that they are always looking to collect additional data and clean up misinformation. Analytic thinkers in basketball argue that the old “eye test” has its limitations and rely on statistics to fill that void. Most coaches can tell what kinds of plays allow their players to succeed: some thrive in the pick and roll, some while posting up, others in isolation, and on and on. But statistics measure exactly how much better these players do in those particular situations.
Despite the obvious value sophisticated metrics provide for scouts in confirming or revising their visual intake, there remains a very small and stubborn school of thought that resists all of the confusing math. Doug Collins' famous claim to media that the analytics were in his head has become emblematic of the old-dog-new-trick routine.
In fairness to Collins, every team relies on analytics to some degree, either using in-house analysts or consultants. In fact, Collins' team hired MIT product Aaron Barzilai as their new Director of Basketball Analytics in 2012. And with Hinkie now running the show in Philly, they'll be relying on those numbers all the more.
The rest of the NBA certainly is.
But that doesn't mean the end of the scout as we know it. As Daryl Morey and Sam Hinkie explained in their 2011 contribution to Grantland:
Information with real power comes in a variety of forms: both in the stereotypical form...of databases and spreadsheets and analysts and predictive models, but also in the form of expertise and experience acquired only via a lifetime of playing and coaching the game.
For those studying the game, assimilating useful information isn't a zero-sum proposition. It's important to see players in action, and it's important to supplement that visual data with performance data. Thanks to the latest advances in geeky scouts' best friend, technology, those are becoming one and the same thing.
How the Tech Craze Is Still Changing Things
The only things more useful than numbers to today's teams are the ridiculously tricked-out cameras that tell talent evaluators and advance scouts alike everything they need to know. Probably way more than they need to know. Grantland's Zach Lowe with the lowdown on the latest:
Fifteen of the league’s 30 teams have purchased a data-tracking camera system from STATS LLC that records every single movement on the court—the ball, the players, the referees, etc.—in three dimensions. The cameras can measure just about anything, and the teams that are using them best have moved far ahead in developing their own algorithms to measure whatever they wish—which team forces pick-and-rolls left most often, where corner 3s typically rebound when they miss, and how often a player accelerates from “jog” to “sprint” during a game.
In other words, we now know everything.
Or we're getting there, anyway. In his nifty explanation of how all of this stuff works, ESPN's Zach McCann admits that while "coaches and stat guys" get tons of data delivered to them in seconds of plays actually happening, "They don’t always know what to do with the information."
Ball Don't Lie's Dan Devine offers one example of information overload, demonstrating that it's as much about processing complexity as volume.
Basketball is a five-man game; with the exceptions of pure one-on-one isolation plays, virtually every individual defensive action depends heavily on the action of at least one (and likely more) teammate earlier in the possession. Team communication, defensive systems, assignments within the context of a scheme ... that's all stuff the SportVU optical tracking data and subsequent analysis of it can't quite get at yet.
Devine highlights one program called Eagle that aims to convert all of this data into user-friendly readouts, including "quick and interactive visualizations of the information (like shot charts and heat maps) and more."
But the software race is only just beginning. Daryl Morey told Slate's Jason Schwartz that he didn't "think anyone has the killer app there—the thing that comes out of that data that gives someone a very significant edge."
Scouts aren't the only ones benefiting from this technology—just like they're not the only ones looking at advanced metrics. But the technological evolution is changing the data environment in which they now work.
Thanks to camera tracking, we'll develop better ways to quantify and understand off-the-ball tendencies, like how a club rotates defensively or where its screens usually come from. Whereas most traditional statistics measure something in relation to the ball (e.g. who's scoring with it, who's rebounding it, etc.), we're getting better and better at monitoring everything else—things happening away from the ball or things speaking to the ball's movement.
That kind of information is power. And organizations are treating it that way.
Analytics departments and front offices around the league remain besieged by siege mentalities. What seems cutting edge now might be old news were all 30 front offices to actually put their collective heads together. But that would require them to lay down their guards in a high-stakes competitive sport.
Not happening anytime soon. According to Slate's Jason Schwartz, "NBA teams guard their data—and whatever conclusions they draw from it—with about the same paranoia as a government official sitting on bomb codes." Zach Lowe likened the phenomena to kids hoarding candy on Halloween.
Compounded with the sheer amount of raw data streaming into teams' hands, that kind of cloistering inhibits these tools' potential. Research mastered by one organization may remain a black hole to another. This is a wise move, competitively speaking, but it's a tragedy for our collective awareness of everything a scout could possibly want to know.
The final frontier might not be so much about technology as it is pursuing an environment in which organizations can develop and apply that technology together, sharing best practices and information in ways that could even be mutually beneficial.
On the other hand, competition often drives innovation. The NBA's information-based Cold War may be one of the reasons we've come so far. Threats of secretly developed stealth-team planes and double-agent scouts notwithstanding, teams pour money into these resources to win—not for the betterment of mankind and its boundless nerdiness.
The Final Frontier
Fortunately, there are some really smart people out there interested in bettering mankind—or at least its most hoops-obsessed constituents.
When he wasn't studying biomechanical engineering as a senior at Stanford University, Muthu Alagappan was busy winning the Best Evolution of Sport talk at Daryl Morey's 2012 MIT Sloan Sports Analytics Conference. Ben Cohen's exceptional profile of Alagappan in GQ dubbed his innovative approach to data analysis "Muthuball."
Hoping to find out what's transpired since last year's conference, I caught up with Muthu, who's now a student at the Stanford School of Medicine and a Sports Data Scientist for Ayasdi. If talking about centers and small forwards is going the way of the dinosaurs, scouts probably need to know about this.
In his original presentation at the Sloan Conference, Alagappan contended that the traditional grouping of five "positions" in basketball had become antiquated. He used software to perform topological data analysis that divided players into 13 classifications according to their statistical production.
Instead of labeling players as "shooting guards" or "power forwards," this system uses the numbers themselves to delineate a "scoring rebounder" from a "defensive ball-handler" and so on. In his presentation, Alagappan used the example of "point forwards" as evidence of our growing tendency to already think along these lines, classifying players by functional designations rather than the overly simplified roles implied by the traditional five positions.
When NBA scouts are evaluating college prospects or talent on other teams, these classifications offer a different lens for assessing whether a player would fit in and fill an unmet need, whether their skill set is too redundant with those already on a roster and a number of other insights that can either confirm or modify previous assumptions.
The possibilities for applying this kind of analysis (and representation) are virtually endless, depending on what kind of numbers you feed the software and how you want to use them.
Unsurprisingly, the applications stemming from Alagappan's research have already grown alongside the emergence of improved data collection (like that nifty camera-tracking system).
On the one hand, that's useful for personnel scouts because really defining a player requires more than data about that player's production. As Alagappan put it, "Two guys can have the exact same box score, but they can do it in very different ways."
It could also help us rethink how we define team styles, though, helping advance scouts more seamlessly interpret and use data about the opposing side. From Alagappan:
Using spatial data and representing it in a very visual manner allows someone to understand the flow of the game more meaningfully. Understanding patterns of ball movement and player movement in a spatial manner gives you a sense for their style very quickly. And it's actually not advanced analytics, it's just advanced visualization. You're not really crunching the numbers, you're just painting them in a more visual way. But I think that's really where a lot of scouting is going to go before games.
Of course, that's not the end of the story.
Players can be defined and visually represented in a lot of different ways.
Maybe a team is investigating its free-agent or draft options and looking to assess injury risks. According to Alagappan, "If we can use medical data to predict a player's career longevity, their career trajectory or their likelihood of injury, you can save teams millions and give them a huge leg up."
And using that kind of data can also help manage players already on a team, says Alagappan:
As we get better at performance tracking, we can start to say...This guy's heart rate tends to reach 140 within 10 minutes of playing, and we notice at that point his performance drops, his speed drops, his change of direction drops. When we start implementing a lot of those wearable devices that track player health and performance, we can start coming up with some really interesting correlations between player health and player performance, which will tell us a lot about injuries but also about overall fitness of players—when to take them out, when to sub them in and so on.
Once we explore all the different ways to synthesize and represent the data, the next big hurdle may be soliciting buy-in from those who ultimately need it most.
"We're getting pretty good at analytics as a community. The question's going to be how do we make those analytics more serviceable to actual results, i.e. how you cross the barrier from analytics to actual players making different decisions," Alagappan explains. Players' habits are already hard enough to change without basing coaching directives on potentially counterintuitive data.
The scariest implication for scouts of the future isn't getting pushback from the organization's resident Luddite; it's the increasingly real possibility that scouts of the future will actually be robots. Not so fast, says Muthu—we'll never get away from that eye test entirely. Robots still need help from people who know the game.
A lot of times I'll run a network and get a result that, to me, doesn't make basketball sense. I do believe it makes mathematical sense, but it's just been a question to me of what I'm doing wrong in terms of picking columns or picking metrics that's causing the software to misinterpret the question. So I think that's where the real disconnect is. The software is always going to do the right job; it's just giving it the right task (which is sometimes not done properly), and that's where I use my intuition.
A fitting sentiment during a year in which we lost a scout with intuition like none other.
And that's the beauty of scouting's future. Even as innovators continue following in Marty Blake's footsteps, improving our ability to make sense of basketball, they'll continue relying on the same fundamentals.
In the last 60 years, Blake's genius hasn't been replaced. It's been democratized, disseminated to all corners of the game in the forms of rapidly developing technology and the very human resources who have to use it.
Even while they're spending the night at Penn Station.
What is the duplicate article?
Why is this article offensive?
Where is this article plagiarized from?
Why is this article poorly edited?