Motivation for the Project
Two of the most common adages in sports are, “Defense wins championships,” and “The best offense
is a good defense,” revealing the long-held belief that a strong defense is a crucial element of a
successful franchise. In a 2013 New York Times article discussing his predictions for Super Bowl XLVII,
Nate Silver supported this with some numbers, noting that the 20 all-time best defenses went 14-6
in Super Bowl appearances, while the 20 all-time best offenses managed only a mediocre 10-10. A year
before, a Freakonomics blog suggested that the adages may be a bit of an exaggeration, but even if
defense isn’t the most crucial part of a team, it is still an important one, and this certainly holds
true in the NFL.
For example, on November 16, 2015, Rob Ryan was fired from his position as New Orleans Saints defensive
coordinator on the heels of a bad loss to the Washington Redskins. Ryan’s career on the defensive side of
the ball has been less than impressive, as this New York Times article discusses, and it’s natural to ask
what he, and other struggling defensive coordinators, could be doing differently. One obvious
responsibility of a defensive coordinator is to make playcalling decisions against an opponent’s offense,
which is difficult without knowing what play the opposition will call. As data scientists, we started to
wonder if these offensive playcalls could be predicted, based on factors such as field position, time in the
game, etc. So in this project, we have decided to undertake this challenge of predicting offensive playcalls
by testing a variety of classifiers on data obtained from Pro-Football-Reference.com. Our hope is that the
playcalling decisions of Rob Ryan and other defensive coordinators around the league can be better informed by
data, and that these classifiers may offer some insight into the playcalling strategies of offenses in the NFL.
