Make Data Work

strataconf.com
Presented by OReilly and Cloudera,
Strata + Hadoop World is where
cutting-edge data science and new
business fundamentals intersect
and merge.
n

Learn business applications of

data technologies
Develop new skills through
trainings and in-depth tutorials
Connect with an international
community of thousands who
work with data

Job # 15420

Data Analytics in Sports

How Playing with Data

Transforms the Game

Janine M. Barlow

Data Analytics in Sports

by Janine M. Barlow
Copyright 2015 OReilly Media, Inc. All rights reserved.
Printed in the United States of America.
Published by OReilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA
95472.
OReilly books may be purchased for educational, business, or sales promotional use.
Online editions are also available for most titles (http://safaribooksonline.com). For
more information, contact our corporate/institutional sales department:
800-998-9938 or [email protected].

Editor: Shannon Cutt

Production Editor: Nicholas Adams

Interior Designer: David Futato

Cover Designer: Randy Comer

First Edition

August 2015:

Revision History for the First Edition

2015-08-26 :

First Release

The OReilly logo is a registered trademark of OReilly Media, Inc. Data Analytics in
Sports, and related trade dress are trademarks of OReilly Media, Inc. Cover image
courtesy of Cambridge Consultants, www.cambridgeconsultants.com.
While the publisher and the authors have used good faith efforts to ensure that the
information and instructions contained in this work are accurate, the publisher and
the authors disclaim all responsibility for errors or omissions, including without
limitation responsibility for damages resulting from the use of or reliance on this
work. Use of the information and instructions contained in this work is at your own
risk. If any code samples or other technology this work contains or describes is sub
ject to open source licenses or the intellectual property rights of others, it is your
responsibility to ensure that your use thereof complies with such licenses and/or
rights.

978-1-491-92748-9
[LSI]

Table of Contents

Data Analytics in Sports. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

Predictive Analytics and the Secrets of the Dance Card
Nostalgia, Statistics, and the Challenges of Motion Sensors
The NBA: A League Driven by Data
Using Data Analytics to Course-Correct
Moving from Raw Data to Analytics
Data in Soccer: Crumbling Long-Standing Myths About
Performance
Wearable Devices and the Future of Sports Analytics
American Football and Strides in Technology

2
4
8
9
10

10
13
14

Data Analytics in Sports

Sports and statistics go hand in hand. As any child with a baseball

card intuitively grasps, any sport is a numbers game, converting
physical prowess into the scores we see on the board. Meanwhile, as
any statistician knows, the more numbers we have, the more ques
tions we can ask and answer: more specific questions, like how
quickly an outfielders reflexes must react to catch a ball; more busi
nesslike questions, like how well an NCAA team must perform to
make March Madness (and how much money each round will earn
their school); and more meaningful questions, like how much strain
a joint can handleand how we can improve performance to
decrease the chance of injury.
More so than in most other human endeavors, sports have objec
tive, direct, highly measurable, and timely outcomeswinners and
losers, margin of victoryas well as easily measured inputs or
parametersteams or players involved, site of game, pointed out
Jay Coleman, Associate Provost and Kip Professor of Operations
Management & Quantitative Methods at the University of North
Florida, one of the data experts with whom we will go in depth in
the next section of this report. All of the factors Coleman described
make sports ideal grounds upon which to hone analytic software
and processes.
Considering how naturally sports and data analytics fit together, it is
jarring to watch the general media narrative of the topic remain one
of combat: the geeks are invading, and the jocks want them out.
Sports are pure, and analytic tools are not, as an opinion piece in the
New York Times argued earlier this year.

Meanwhile, in an industry valued at $1.5 trillion globally and $485

billion in the United States alone, with revenues making more than
billion-dollar jumps every year, and technology surging ahead faster
than it ever has, leaders in the field have less time for philosophical
arguments about purity and more for seeking the purest data. In this
report, through conversations with experts at the intersection of
data and sports, we will explore advances in the sports industry,
from predictive analytics that work from the outside-in, to the kinds
of technology in which organizations like Major League Baseball
and the National Basketball Association themselves have invested.

Predictive Analytics and the Secrets of the

Dance Card
The most basic questions among sports fans will always be who will
win (a game, a set, a championship) and how a team can make it
happen, but it takes a great deal of knowledge to build up to predic
tions. Before the predictive formula comes the data set. Without the
historical patterns needed to build a model, there would be no pre
dicted patterns. So, when Jay Coleman discovered a treasure trove of
online data revolving around the college basketball teams that made
(and many that did not make) the NCAA Mens Division I tourna
ment from 1994 to 1999, he did what any good data scientist would
do: he used it.
Coleman, along with colleagues Mike DuMond of Economists
Incorporated, and Allen Lynch of Mercer University, is the creator
of the Dance Card, a formula that predicts which mens teams will
end up in the Big Dancea common term for the tournament
that spawns March Madness every year.
When March Madness rolls around, friends and coworkers across
America gamely fill out their brackets the way so many fair-weather
basketball fans do: with little knowledge of how the bracket arrived
at its current state. This cohort may delve into the world of statistics
to improve their bracket predictions, or to substantiate their hope in
their chosen teams, but they largely ignore or remain unaware of the
extent of the guiding hand of the NCAA Selection Committee.
March Madness draws its teams from 32 conferences, all with extra
ordinarily different ranges of size and skill, so naturally the NCAA
will not just award tournament placement to the top two teams from

Data Analytics in Sports

each conference. Instead, the winner of each conference receives an

automatic bid, and the NCAA Selection Committee fills out the rest
of the 64-team bracket with at-large bids. In a simpler world, these
at-large bids could conceivably rest on one factor alone: the RPI, or
Ratings Percentage Index. RPI calculates a score for teams based on
a host of statistics including their win percentage, their strength of
schedule (the difficulty of playing their opposing teams), and more,
with the goal of unifying these statistics across diverse conferences.
The Selection Committee, however, builds the final bracket off of
more than RPI alone, and they famously keep their exact criteria
secret. This creates a unique territory for predictive analytics to
exploit at the intersection of team performance and the psychology
of the selection committee.
[The Dance Card] simply identifies the patterns in the past Selec
tion Committee decisions, and assumes those patterns will repeat in
the near future, said Coleman in a recent interview. The Dance
Cards success at predicting indicates that they obviously do.
Indeed, the Dance Card achieved a 92% success rate in 2015 by cor
rectly predicting 33 of the tournaments 36 at-large bids. Over the
last four years, since the most recent update in 2012, its predictions
number 141 correct out of 146, for an overall 97% success rate.
Coleman, DuMond, and Lynch used SAS analytics to design the
Dance Card, which is a stepwise probit model. Probit models pre
dict one of two dependent variables, like the Dance Cards binary
variable: whether or not a team will make the bracket. (A probit
model would not be appropriate if, for example, the Dance Card
sought to determine the Selection Committees likely seeding
choices, which are much more complex outcomes than a binary
variable). Stepwise regression, meanwhile, suits the Dance Cards
need to sort through a great number of variables, like those that
compose the RPI, and others of unknown and differing importance.
The development of such a formula means that the creators can
update their predictions for the NCAA tournament after each and
every college basketball game, if they choose. It also means that they
can continually update the formula itselfand analyze what that
means for the evolution of the Selection Committee. The 2012
update, for example, disposes of the variable concerning the com
mittees bias toward teams within certain conferences. This is a
sharp turnaround from Coleman, DuMond, and Lynchs 2010 study
Predictive Analytics and the Secrets of the Dance Card

titled Evidence of Bias in NCAA Tournament Selection and Seed

ing, which suggested that there was substantial bias toward teams
playing in six major and six mid-major conferences, over teams
playing in nineteen minor conferences through two models, based
around the committees team selection and their seed placement.
These results, derived from ten years worth of Division I game play
from 1999 to 2008, made waves in the media, but within two years
such historical bias was nullified enough to necessitate a change in
the Dance Card formula. This suggests that the Selection Committee
members themselves were aware of the bias towards major confer
ences and the need to correct it. Perhaps even the NCAA knows
when data analytics are about to catch up with them.
By pulling aside the curtain on the opaque practices of the NCAA
Selection Committee, the Dance Card reveals the power of predic
tive analytics in areas that dont get as much attention as, say, how to
win your office betting pool. Achieving a place on the March Mad
ness bracket can mean big gains in money and exposure for the
playersand, on the flip side, losing a place can mean big losses,
especially for those unpaid players hoping to propel themselves into
the NBA. With stakes like these, expect more eyes on the Dance
Card in years to come.

Nostalgia, Statistics, and the Challenges of

Motion Sensors
With the Dance Card, we took a look at one algorithm and its
impact on one of the most lucrative sports events in history, the
NCAA Division I Mens Basketball Tournament. Lets zoom out now
for a full perspective on a professional sports clubs technology
needs.
As you know, baseball has a tremendous amount of nostalgiaand
almost limitless data, said Brian Shield, Vice President of Informa
tion Technology for the Boston Red Sox, but that data is changing
too.
Baseball, Americas pastime, and the nostalgia of its fans, have pro
duced a great wealth of data stretching back through the history of
the countrys professional leagues. As Benjamin Morris of FiveThir
tyEight, an ESPN blog founded by statistician Nate Silver after his
rise to prominence predicting the results of the 2008 American pres

| Data Analytics in Sports

idential election, recently commented, many young fans apprecia

tion begins with baseball cards and the numbers and details they
provide; because of statistics, their interest in the sport deepened to a
unique level. On the other side of the same coin, there exists a dis
trust in deviation from the traditional ways of analyzing and collect
ing these statistics. This is why Moneyball and the rise of sabermet
rics, a term for the objective analysis of baseball statistics, caused a
stir among baseball fans and commentators in a way no other sports
adoption of data analytics could. This is why, even in 2015, the New
York Times ran the aforementioned opinion piece titled Dont Let
Statistics Ruin Baseball, by baseball expert Steve Kettman, implying
that fans cannot truly experience or appreciate a game if they focus
too much on the numbers behind it.
Brian Shield acknowledges that the world of baseball has so far been
warier than other sports to embrace analytics. In many ways, it
hasnt, like football perhaps, tried to be the leader in adopting the
latest technology, he said. Thats changing. Thats what baseball is
now doing.
Unlike the haphazard adoption of sabermetrics that initially gave the
Oakland Athletics their advantage in the seasons documented by
Moneyball, the rollout of baseballs latest statistical innovation occur
red jointly for the teams of Major League Baseball. Statcast, a playertracking system developed by the MLB Advanced Media (MLBAM)
group with Amazon Web Services, began operating officially in each
of the 30 major league ballparks as of Opening Day 2015. Using a
network of cameras and radar, Statcast tracks the entire field to col
lect data on each player and each movement of the ball. The tech
nology won MLBAM the MIT Sloan Sports Analytics Conferences
prestigious Alpha Award for Best Analytics Innovation/Technology
earlier this year.
The MLB has previously teamed with Sportvision to develop pro
grams like PITCHf/x and HITf/x, which track and record data on
the trajectories of pitched balls and hit balls respectively, but HITf/x
data remained proprietary to each team. Statcasts promised availa
bility to public-sphere statisticians, commentators, and fans will cre
ate an unprecedented level of batted ball data.
The radar technology responsible for ball tracking, by TrackMan,
will work in conjunction with cameras by ChyronHego that will

Nostalgia, Statistics, and the Challenges of Motion Sensors

track player movement, going above and beyond PITCHf/x and

HITf/x to collect vital data on base running and fielding.
We all consider this the next significant jump in baseball analytics
that was never there before, said Shield. A lot of things that were
qualitative information will be quantitative. Did that player get a
good jump on the ball? Well, now you can see how many tenths of a
second it took for the person to react to that ball.
Claudio Silva, Professor of Computer Science and Engineering and
the NYU Polytechnic School of Engineering and a member of the
team building Statcast, also spoke with us regarding progress and
challenges of the technology. Silva works primarily in creating algo
rithms that translate radar data into game metrics, and although he
grew up watching soccer, he sees baseball as a unique opportunity to
make inroads into sports technology.
Baseball looks more like chess, where there are pieces that each
have certain properties, said Silva. [In baseball,] you know that
someone is going to throw the ball in a particular way, hit it so many
times. This creates openings for the gathering of more specific data,
and thereby the construction of better algorithms.
Still, Silva explained, creating high-fidelity models has been tough.
Statcast represents the next generation in terms of modeling what
the game actually looks like, he said, largely because of the raw data
sources: the advanced radar and cameras. As a data scientist, he has
found himself in the position of determining not only what
improvements he and his team need from software, but also what
kind they need from hardware.
The rate at which you can sample anything on the field determines
whatever you have on that particular derived metric, Silva said. If I
can track the players at 30 Hertz, there are certain implications to
what I can measure. Then if I only have the ball at 30 Hertz, that
would limit my knowledge of the ball, because the ball moves much
faster than the players.
The ultimate challenge with optical technology lies not with the lim
its of what it can do when its functioning properly, but with how it
functions in the real world. The simple matter of placing sensors
and cameras where they will be least affected by factors as elemental
as the weather can become a significant burden on any system.

Data Analytics in Sports

Plus, as Silva pointed out, there is no way to adjust to such sensors

on the fly. Comparing the system to something with smaller stakes,
he suggested thinking of the Microsoft Kinect, a set of motionsensing equipment primarily for Xbox gaming. Imagine, in your
home, trying to play with the Kinect, he explained. Youre probably
going to notice if you get too far, and the Kinect will actually tell
you: youre out of bounds! You have to get closer to the sensor. Dur
ing a baseball game, needless to say, there will be none of that. It can
only be an ongoing process of adjustments.
Brian Shield can corroborate the obstacles associated with placing
any new physical systems in a baseball park. The MLBs initiative to
ensure public-facing WiFi in all stadiums within the 2015 season has
spurred the Red Sox to install over 500 wireless access points around
Fenway Park, but, as a site of historical significance to the commu
nity, the site cannot display too much visible change.
Fenway Park is like a living museum in many ways. Any kind of
revision to the ball park is scrutinized, said Shield. We work quite
a bit with our Historical Society to make sure that we can both meet
our needs as far as delivering a quality WiFi experience and [main
taining] what we like so much about Fenway. The same goes for all
types of technological equipment.
Despite the importance of keeping up a traditional low-tech appear
ance on the (literal) surface, teams like the Red Sox are quickly
becoming their own technology creation centers. This year, Shield
mentioned, the Red Sox launched their first customer data ware
house. Public-facing WiFi naturally extends a great service to game
attendees, but it also has the power to feed important data back to
the teams.
We want to know how fans operate within the park so we can cre
ate better fan amenitiescreate a better fan experience overall, said
Shield.
For those watching from afar, any conversation about the impact of
data analytics on sports expectedly homes in on player statistics and
how teams can use predictive analytics to judge their players (and,
of course, any potential players). For a teams IT executive and his
department, however, like Shield, the in-house experience domi
nates the workload. The impact of team performance on hooking
and maintaining fans is only one aspect of the business of data in
sports.
Nostalgia, Statistics, and the Challenges of Motion Sensors

The NBA: A League Driven by Data

The National Basketball Association, like Major League Baseball, has
recently set its optical tracking system in place, and based on the
sheer nature of the game, the league stands to gain even more. As
discussed with Claudio Silva, traditional analytics suit baseball
easily: the game breaks down into discrete plays, hinging on the
one-on-one confrontation between the pitcher and the batter each
time. Until now, the seamless flow of basketball has rendered a simi
lar breakdown beyond the capacity of technology, but the era of big
data means working with huge amounts of micro data. Suddenly,
that seamless flow looks like an analysts vision of wave after wave of
data.
These data come in the form of coordinates sent from the sixcamera system now in each of the NBAs thirty arenas, using tech
nology called SportVU (pronounced SportView). This technologys
long journey to universal adoption by the NBA began in Israel,
where missile defense specialist Miky Tamir began adapting
military-style optical recognition to the soccer field in 2005. Once
American company STATS LLC acquired SportVU in 2008, they
saw an opportunity in the world of basketball.
Eight teams signed for and installed SportVU by the end of the
2011-2012 season, paying out-of-pocket, and seven more did so by
the 2012-2013 seasonnot yet enough to generate conclusive,
league-wide figures, but enough for two game-changing paper pre
sentations at the MIT Sloan Sports Analytics Conference 2013. Kirk
Goldsberry and Eric Weisss The Dwight Effect: A New Ensemble
of Interior Defense Analytics for the NBA and Philip Maymins
Acceleration in the NBA: Towards an Algorithmic Taxonomy of
Basketball Plays put SportVU on display for an audience of true
influencers. Goldsberry and Weisss sweeping look at how spatial
analysis can measure defensive success and Maymins deep dive into
the technologys ability to track a previously unquantifiable skill
won over fellow researchers and executives at Sloan. Its no wonder
that the NBA moved to partner with STATS LLC to use SportVU
across the entire league.
In the NBAs official press release following the decision, a statement
from Steve Hellmuth, Executive Vice President of Operations and
Technology included the boldly-put phrase, We are a league driven
by data.
8

Data Analytics in Sports

Philip Maymin, Assistant Professor of Finance and Risk Engineering

and the NYU Polytechnic School of Engineering, has a long history
of analytics consultancy for NBA teams, and has recently joined
forces with Vantage Sports, a new leader in sports analytics and a
finalist for Sloans 2015 Alpha Award for Best Analytics Innovation/
Technology, as their Chief Data Scientist. He spoke with us about
the value of new analytics technology for sports organizations and
his upcoming work with Vantage Sports.
Theres a lot of new data tech thats come out. You still have to have
a human element to make sure its reasonable, said Maymin. As
good as new wearable devices and optical tracking have become, he
explained, sometimes there are errors a human needs to spot. You
dont have a heart rate that jumps up to a billion, then goes back
down to zero. You dont have a guy whos on one side of the court
and he magically materializes on the other. You have to look for
these kinds of mistakes in the data and fix them. Theyre not com
mon, theyre rareall the data providers do a wonderful jobbut
things happen and you have to keep an eye on it, because any one of
those outliers could have a very large effect on what youre trying to
do.

Using Data Analytics to Course-Correct

High-profile media coverage of advancements in sports analytics
frequently feature division between supporters of data analytics and
those more leery of the shift towards new technologies, but within
any sports organization the true goal is harmony among all. As May
min says, You dont want to lose somebody who doesnt understand
data as well as you do, because they contribute in many other ways.
Proper use of analytics should serve to bring departments together.
As in any industry, using the data to show an associate how to cor
rect his or her course should come long before the statement that
the data show this person is performing poorlyso goodbye. To
illustrate this, Maymin uses an example of the scout who is always
wrong.
Every guy he thinks is great ends up being a bust. Every guy he
thinks will be a bust ends up a superstar, Maymin described. In
any regular organization, that guy would have been fired ten times
over. With an approach involving data analytics, using input from
scouts and the ensuing player outcomes, this scout may have courseUsing Data Analytics to Course-Correct

corrected much earlier, and many patterns of missteps that do arise

will be much simpler to identify and improve. In the same way,
coaches can better hone their players skills based on the results of
emerging technologies, so can departments within an organization
better train their workers.

Moving from Raw Data to Analytics

Vantage Sports, the company Maymin joined earlier this year, offers
teams this capability to hone their training and improve their play
ers with a combination of optical data and the human element that
Maymin considers still vital in a world of imperfect machine learn
ing.
When it comes to optical data, Maymin questioned, how do you
get from the x/y data, the raw data, to interesting basketball ques
tions, like was there a pick and roll? Since optical systems alone are
currently not capable of honing their data to this level, the program
is a mixture. Using an organizational system breaking down move
after move that can happen on offense or defense, the program tags
over 16,000 events per game.
For example, one thing we tag, that no one else on Earth has, is
whether a defender had his arm raised when defending a shot, said
Maymin. It turns out to be very important! Who knew? Of course,
every coach and fan knows it is important, but it is literally impossi
ble to derive from the raw x/y data, which measures a persons center
of mass, whether or not their hand was raised.
Vantage Sports CEO Brett McDonald gave a presentation to the
Sloan Sports Conference of 2014 on taking sports data from propri
etary walled gardens to open access, a shift for which all of those
at the intersection of sports and data should be on the lookout. The
company currently provides distinctive NBA commentary on its
website, using some of their data, but going forward, said Maymin,
they are aiming to put out a new fantasy analytics product.

Data in Soccer: Crumbling Long-Standing

Myths About Performance
Weve already established that basketball is a game of flow compared
to the discrete play of baseball, but what about the next level of flow?

10

Data Analytics in Sports

Basketball games in the NBAs 2014-15 season averaged 37.5 field

goals, so over the course of a four-hour game, there could be nine or
ten of these natural resets per hour. The shot clock ensures this. A
game without even these easily identifiable changes in state, how
ever, exists: soccer (hereafter referenced as football).
Garry Gelade, Director of Business Analytic Ltd., through which he
has advised teams like Chelsea Football Club, Paris Saint-Germaine
Football Club, and Real Madrid, as well as businesses outside the
realm of sports, speaks of this in terms of accurate judgment of per
formance. Analytics can do this in ways subjective observation can
not.
Players get feedback from their managers and teammates, but it
isnt easy to judge performance in a complex flowing sport like foot
ball or rugby, said Gelade. The quantifiable data that analytics bring
to the table ease this burden. Being able to track the important
components of performance and having sound benchmarks for
comparison can give a player and his manager a clear sense of what
they are doing well and where they need to improve.
It takes massive amounts of data, Gelade explained, to weed out the
key components of player performance. In team sports, measuring
the ability of individuals uncontaminated by the performance of
their teammates is a decidedly non-trivial task, he said. [Its] some
thing you can only do by analytical modeling.
As is the case with other leagues and teams throughout the sports
industry, top-of-the-line recording equipment to gather data for
descriptive analytics will come before prescriptive analytics. Highresolution tracking system TRACAB by ChyronHego, the afore
mentioned company that honed its player-tracking software for the
MLB, is now in use by the Barclays Premier League in the United
Kingdom, the Swedish premier league Allsvenskan, and the J.Lea
gue, the top division of the Japan Professional Football League,
among others. When it comes to processing stats, competing com
panies Opta and ProZone use a combination of recording technol
ogy and human analysts who tag events within the game (much
like Vantage Sports). Opta calculates that it tags between 1,600 and
2,000 events per football gameall delivered live.
These sources will produce vast amounts of data, which will proba
bly drive a move towards machine-learning and data-science
approaches, Gelade hypothesized.
Data in Soccer: Crumbling Long-Standing Myths About Performance

11

Some long-standing myths have already begun to crumble as datadriven approaches thrive and the information from them begins to
emerge. Gelade cited distance run as one of the pillars of football
that we can now examine more critically. Blindly assuming that
more running equates to a more effective performance was for a
long time an article of faith in football, but data modeling shows it
isnt true, he said. And this is where the shift from descriptive ana
lytics into prescriptive analytics will occur. Its all very well to know
how far or how fast a football player has run in a match, but the real
question is does it make any difference to the result, how much and
in what way?
Gelade sees a slower march towards wide acceptance of sports ana
lytics in Europe than in the United States. The singular champions
of analytics who popularized it here do not exist there. Billy Beane,
the executive of the Oakland Athletics who drove the events docu
mented in Moneyball, is the kind of figure other leagues need.
We dont have that kind of charismatic role model, said Gelade.
We cant point to any European managers or teams who overcame
their financial disadvantages in such a spectacular way.
Statisticians may brush aside the greater publics lack of enthusiasm
for analytics with the assertion that a club or a league does what it
needs to do regardless of the laypersons suspicion, but as Maymin
pointed out, sports organizations themselves contain multitudes. It
will take strong leaders to cause changes in company culture.
In America [there are] some high-profile and very successful sports
team owners who have espoused the analytics approach, Gelade
explained. People like John Henry, owner of the Boston Red Sox,
and Darryl Morey, GM of the Houston Rockets, are evangelists for
analytics in the USAand this has given analytics a credibility in
America that it still somewhat lacks in Europe.
Football does have its own John Henry nowin John Henry him
self, who is now the principal owner of Liverpool Football Club.
Gelade praised their strategic commitment to analytics, including
the creation of a sophisticated data repository, and the employment
of their own team of coders and data scientists. That way, Gelade
said, They have control over what they do, and can build some
invaluable intellectual capital.

12

Data Analytics in Sports

The growth of staff data teams is a trend across all sports, as Brian
Shield attested earlier, and they are only one way in which analytics
are changing the industry and the industry is changing how the
world sees analytics.

Wearable Devices and the Future of Sports

Analytics
What does the future hold for sports and data? Look for an
increased focus on injury prevention, especially through a greater
range of questions asked of players and, of course, data collected
through wearable sensors.
We tend to think of predictive analytics using statistics about player
performanceas a way of evaluating talent, said Brian Shield. I
think youll find predictive analytics being applied to health science
as well.
For now, the practical application of such data still goes a long way
toward player evaluation; as research accumulates, organizations
dont need to take risky bets on whether that player with a bad knee
will be viable long-termthe predictions are much closer to the
truth. Plus, as Shield pointed out, historical data like that bad knee
will begin to yield to closer scrutiny of whether currently healthy
bodies can withstand years of professional play. In terms of baseball,
Shield phrased it, Is their physique the kind of physique that can go
through two hundred innings a year?
Philip Maymin remains wary of anyones ability to mine data for
injury-related predictions just yet. The database of injuries is very
sparse, so its hard to make this kind of judgment, he said. Every
single injury for every single person seems to be very specific to
them and idiosyncratic. Its hard to get a very reliable data set.
Wearable bio-devices and the myriad types of data they can collect
promise to make this lack of information a thing of the past, though
they still have a long way to go. No major leagues currently allow
wearable devices during actual games, in part, Maymin noted,
because of the lengthy collective bargaining between the organiza
tion and the players union that would need to take place in a league
like the NBA. In 2014, however, the NBAs Development League (DLeague) allowed four trial teams to begin wearing devices manufac
tured by STAT Sport, Zephyr, or Catapult Sports to track movement
Wearable Devices and the Future of Sports Analytics

13

and heart rate in-game. For a sport in which practices are notably
far slower-paced than games, this is a major step.

American Football and Strides in Technology

These devices are one area in which American football has taken
strong strides ahead in technology. Over half of the teams in the
NFL use wearables during practice, and the NFL could conceivably
be the first league to use them regularly in-gamethey have, in fact,
already collected some in-game data using technology called Next
Generation Statistics, created in partnership with Zebra Sports Sol
utions, albeit no data that teams themselves can access. Using Next
Gen Stats, chips placed in players shoulder pads transmit Radio
Frequency Identification Signals (RFID) to receivers throughout the
stadium. With the NFL allegedly planning to dispense some of this
data to teams and create some distillations of it for fan consumption,
RFID will soon have a major foothold in sports, and could pave the
way for more complex wearable devices.
As Philip Maymin joked, Ideally, youd want every player wearing
those virtual reality suits that people use for video games, but of
course, there are some nods toward purity among the sports and
data set.
For now, optical tracking, wearable devices, and the minds and
machines that extract meaning from it all will be enough.

14

Data Analytics in Sports

About the Author

Janine Barlow is a writer and publishing professional currently
working with AMACOM Books, the publishing arm of the Ameri
can Management Association, where she promotes books on busi
ness, marketing, technology, psychology, and more. Outside AMA
COM, she brings a sharp editorial eye to content on business intelli
gence and advanced analytics. In the world of sports, look for her at
the NWHLs New York Riveters games in Fall 2015.