Analytics Antenna

Date01 July 2016
AuthorTim Chartier
DOIhttp://doi.org/10.1002/jcaf.22181
Published date01 July 2016
67
© 2016 Wiley Periodicals, Inc.
Published online in Wiley Online Library (wileyonlinelibrary.com).
DOI 10.1002/jcaf.22181
C
o
m
m
e
n
t
a
r
y
Analytics Antenna
Tim Chartier
During my childhood, my
family had an old TV in
the basement. It’s recep-
tion came via an antenna that
sat on the top of the rickety set.
To watch a program, I’d fiddle
with the rods hoping to see
images emerge through the fuzz.
Analytics is often the same.
Life has its serendipity. Such
events aren’t predictable but are
generally inherent. Within the
noise of randomness, analyt-
ics can often identify and even
quantify a trend. Yet seeing the
signal isn’t always easy.
Sports analytics is a sub-
field of data analytics and
encounters many of the issues
that arise in financial, retail,
or health care analysis. From a
last-second shot that bounces
around a rim to end a game in
victory or defeat to balls that
are tipped on the gridiron and
result in interceptions or mirac-
ulous receptions, sports has
the random elements that can
obscure the underlying trends
of skill and talent.
With my family’s television,
I easily saw when broadcast
images emerged into the pic-
ture. In data, seeing a trend
emerge isn’t always easy. In
fact, even when a pattern
appears, one must ensure its
generalized presence.
To give more context, let’s
turn to my recent research.
I’ve been working with ESPN
on a new analytic to identify
unpredictable teams. For each
season, we analyze over 5,000
games involving approximately
350 Division 1 NCAA men’s
basketball teams. I wrote a
computer program to analyze
the data with our new algo-
rithm. The result was a list the
teams classified as unpredict-
able and predictable.
One rarely knows how ana-
lytics will unfold until results
emerge. And, results are rarely
perfect. Randomness plays a
role. A high degree of accuracy
often indicates promising new
results.
For this work, I ana-
lyzed seasons spanning over
a decade. Sometimes, early
results led to only a moment
of laughter and then reflec-
tion regarding additional steps
that might improve the work.
Then, there are moments, like
this instance, were the results
make me pause and look again.
When I looked at the first list,
Isaw teams classified, as I
would hope. I looked at more
seasons. Every year, I saw
teams classified correctly, but
not all of them. I didn’t expect
perfection as one analytic
rarely captures every variation
of a phenomenon.
As I looked over the results,
I was admittedly stunned. I
paused, took a short walk, and
then returned again and looked
at the results. I did appear to
have a metric that classified
teams with attributes of inter-
est. One my collaborators, a
recent alum of Davidson Col-
lege with whom I’ve done a lot
of research both during her
time at the college and now as
an alum, asked about the work.
I showed her the lists.
She turned and said, “Wow!
We might have something.
Andwe may not. We need to
automate the process and see
if a computer can find what we
are seeing.” I smiled as that was
a speech I often gave. “Why?”
I asked. “Because we might
only be finding what we want
to find and overlooking ways
that this result isn’t as easy to
see as it seems.” She was right,
and it took time and other col-
laborators to solidify the base
of the work. Did the metric
work? Indeed, but interestingly
it was a variant on that origi-
nal method that will lead to
new work published in ESPN
Magazine.
Let’s pause and see
important parts of this process.

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT