[dt_divider style=”thick” /]In 1999, Paul DePodesta was hired by the Oakland Athletics as an assistant to the then GM Billy Beane. DePodesta, and his trusty sabermetrics, led the A’s to unprecedented success given their budgetary restrictions. Along with the help of a successful novel and film adaptation, Beane and DePodesta helped usher in an era of analytics in sports that continues to grow by the day.
The Houston Astros won the World Series only a few years removed from being in the cellar of the league due in part to their unique analytics-aided roster construction. In the NBA, we have seen the advent of the analytic-driven ‘Morey-ball’, a style of basketball that puts an emphasis on three-pointers and layups, made famous by Houston Rockets GM and MIT grad Daryl Morey. Anecdotes like this don’t seem to exist in the NFL.
This does not mean the NFL has not invested in analytics. Most teams have a full-time analytics director on staff, while others have full blown analytics departments. Pro Football Focus, the current source for the best advanced player performance statistics, lists nearly the entire league as their clients. Analytics in football is not a taboo subject, it just hasn’t found its footing. It exists currently more to rule out certain decisions, like alerting a team when they are investing too heavily in a certain player/position, rather than being the driving force for decisions. Making this leap from boundary-setter to decision-driver is no small feat. It will require an extremely innovative thinker paired with a forward-thinking, and potentially desperate, front office and coaching staff. An interesting question arises from this consideration, why hasn’t analytics taken off in the NFL?
The cheap answer to this question would be to point to the old-school, stubborn, tough-guy mindset of the NFL. While partly true, to me, this does not answer the question. Football coaches and front offices may be slightly more stubborn or hard-headed than say an MLB front office. But, above anything else, they care about winning, and if something exists to help them win, they will use it as much as humanly possible. In fact, not too long ago, it was the NFL that was ahead of the curve in terms of willingness to change and implement new strategies. Football coaches all but invented ‘film-study’ and advanced preparation, and the game has undergone more league-wide changes in style of play than any of the other leagues combined (think Paul Brown’s forward pass, Bill Walsh’s west coast offense or today’s new spread-em out style of play). If crusty, old, run-first ‘football guys’ pushing back on analytics aren’t the cause for their lack of success in football, what is?
Analytics and Football – Oil and Water?
The game of football is uniquely designed to make implementing successful analytics very difficult. Let me explain:
Measuring player success in football is extremely difficult
There is a reason that baseball is the poster child for sports analytics. It is a binary game. On every play, you either succeed or fail. There is very, very little room for gray area in this regard. Basketball is not too far behind. There exists an extra layer of subjectivity when compared to baseball, but still, a relatively binary nature is still present. Football is quite the opposite.
Think of a running back that takes a broken play, where he should be tackled for a loss, and turns it into a 3-yard gain after making multiple defender miss. Then compare that to a play where the blocking is perfect, or the defense completely misses their assignment, and he runs for 50 yards without ever being touched. Which play is truly more indicative of the type of player the running back is? The answer is almost surely the former, but translating that into something quantifiable is a tall task.
There are few comparable grey areas of success to a heroic 3 yard run in other sports. They are pass/fail. Did you get a hit or did you get out? Did you make the shot or did you miss? In football, examples of subjective success are nearly endless. What about an offensive lineman who drives his man 4 yards off the ball, but as the running back hits the hole, he ever-so slightly slips off the block and the defender makes an arm tackle for a 3-4 yard gain. I would look at that play and say that the OL was the reason a positive gain was achieved in the first place and getting that much movement off the line of scrimmage was an incredibly impressive individual effort. Whereas, somebody else grading that play might think I’m crazy and the entire blame would be placed on the OL whose defender made the play.
Furthermore, if you are not part of the staff that drew up the play, you often will not know what was trying to be accomplished. As a former offensive lineman, I know all too well that many times you can be doing your job just fine but something in the play goes array, the ball bounces where you weren’t expecting it go and your man makes the play. To an un-educated eye, the offensive lineman clearly is at fault, but should he be? Well, the answer there is yes, because it’s always the offensive line’s fault, but the point is, if you don’t know the objective or scheme of the play, placing blame or credit can be extremely difficult.
Variables, Variables, Variables
When trying to obtain actionable insights from data, the make-up of your data is incredibly important. The easiest data to extract actual insights from will include a massive sample size with a limited number of variables. The ‘story’ of these numbers will not be hard to follow and even the most basic analytic tools can be useful when data is constructed like this.
At a high level, compared to the other major sports, football has the most variables (22 players on the field) combined with the smallest sample size (16 games). Going even further, simply having the most players on the field does not come close to telling the full story when it comes to the amount variables at work on a given football play. Each player on the field has a job on each play, and depending on the play, each player will have an actual effect on the play. In this way, there is no other sport like it. In baseball, the only people involved on a given play are the hitter, pitcher, catcher and whoever the ball is hit toward. In basketball (especially the NBA), while team defensive positioning is important, the game, at its core, still comes down to the man with ball and the man covering him. Similar to my sentiment regarding understanding the concept of a play, with the amount people influencing a play, it can become extremely hard to determine who should be credited or blamed with the success or failure of a given play.
Another variable that is unique to football is the lack of uniformity within the construction of a given play or series. In baseball, each play starts the exact same way, with absolutely no variation. In basketball, despite the fluid nature of the game, a natural up and down progression is followed. This is not the case in football. Where you are on the field, what the down and distance is, and whether it is a run play or pass play are all completely unique variables that exist only in football. Each play has a certain context around it that must be accounted for and because of that, gathering usable data is extremely difficult. A run that only gains 1 yard would, 95% of the time, be considered failure. But, if it was 3rd and inches, the opposite is true. Outside of extremely rare circumstances, missing a shot in basketball or getting an out in baseball, is always considered a negative outcome for the offense/hitting team. There are so many contextual variables for each play that occurs on a football field, variables that both do not exist in any other sports and variables that, when trying to gather usable data, must be quantified and accounted for.
Where Do We Go From Here?
All that was said above is meant to serve as an explanation for why we do not hear the analytics success stories that we hear in other sports in the NFL. This does not mean that analytics will not make the proverbial leap and will never become a fixture of decision making processes in NFL front offices and coaching booths. In fact, I believe the opposite is true. It is simply a matter of time until analytics finds its footing and the NFL has its version of Moneyball. With that said, the question must be asked: where are the most likely spots for us to see analytics shine through in the game of football?
1. League-wide changes in style of play
We have seen this occur time after time in the NFL even without the helping hand of big data. A football game in the mid-70’s seldom resembles a modern-day NFL game and this evolution will only continue to unfold. What will be the NFL version of three’s and layups? Could it be an exacerbation of the current wide-open NFL offense that moves closer and closer to a college offense? Or will it be a steady dose of run plays and play actions passes? The answer will lie in the numbers.
2. Quarterback evaluation
Many advanced metrics, while still not finely tuned, currently exist around the quarterback position. As mentioned previously, using numbers to evaluate football players is incredibly difficult and, in one sentence, is the reason why this article is being written. But, if there is to be a position where the code is truly cracked, it will be the quarterback. Some metric will emerge that quantifies how ‘good’ every throw is based on the pressure the quarterback is under, multiplied by the space available to put the ball, plus the distance the ball is thrown (or something along those lines). Naming the top 5-7 quarterbacks in the league is not a difficult task. Conversely, differentiating between the 9th and 15th best quarterbacks in the league is nearly impossible. Who is really better between Kirk Cousins and Case Keenum? Alex Smith or Blake Bortles? These answers often change on a yearly basis given scheme and surrounding talent; currently, these questions are a crap-shoot for NFL front offices, but someday, with the aid of advanced numbers, more clarity will emerge.
This will be one of the final frontiers of analytics in the game in football. Coaches already carry around their “chart” which, using numbers, lays out when they should/shouldn’t go for it on 4th down, when they can begin to take a knee along with other various situational breakdowns. This is Version 1.0 of what analytics driven play calling will look like. Someday, every play-caller in the NFL will have a line in their headset that patches them through to their analytics ‘guy’ who can provide the statistically best play for every situation that is encountered on the football field. There will be a synthesis of what currently exists and a new analytics driven play calling system. How much teams lean on analytics to drive their play calling will differ greatly amongst different teams, but in a game of trends, where everyone is looking for an edge, you are crazy to think that teams will not get creative with this type of stuff. The early success of these initiatives will be crucial in determining how widespread it becomes, but all it will take is one success story, and teams will be all but forced to test their luck.
Analytics has not found its true place in the game of football. The reason for this is more about the way in which the game is played, not the people involved in it. The NFL will have its version of Moneyball, the question is not if, but instead, when and where. Those answers are not clear, but someday we’ll know and hopefully, when that happens, it pushes the game forward in an interesting and innovative way that makes the game better for us, the fans. Also, hopefully we get a full feature movie so the best football movie in the last 10 years is no longer Draft Day, please and thank you.