Football, Analytics, and the Big Data “Rabbit Hole”
Ever since I can remember, I’ve been a fan of the Buffalo Bills. I know, I know, they’re usually not in the playoffs, but what can I say: what I lack in intelligence, I make up for in loyalty. I bring up my Bills fanhood because they recently hired a Director of Analytics for their team, going down a road that all baseball teams and a few football teams have already traveled. This got me thinking…is it always great to measure and analyze data if it’s not the RIGHT data?
As you may recall, the Oakland Athletics of baseball brought national attention to sports analytics by employing the “Moneyball” strategies of their GM Brad Pitt (played in real life by Billy Beane). Simply put, “Moneyball” referred to the use of advanced statistics like WAR, slugging percentage, and OPS as opposed to more widely accepted stats like runs batted in, etc. This allowed the A’s to build a roster that could compete and win against teams that more than doubled their payroll.
Now football, from an analytics perspective, is much different than baseball. Rather than just a pitcher vs. a batter, football plays consist of 22 large, fast people running into and around each other at full speed. Accordingly, there are an extremely large number of variables on any given play. Further, with only 16 games per season, the sample size for football data tends to be smaller than in baseball, where there are 162 games.
Is there a point to this blog post? Glad you asked, I was just getting to that. How, in a sport as complex as football (and with such a small sample size), will you know which data to measure and analyze? It’s a problem regularly faced in the “real” world as well…access to data is no longer the problem; knowing what to analyze is the main issue.
Football analytics has already had an effect on in-game coaching decisions (like going for it on 4th and 1 in your opponent’s territory) with stats like Expected Value and Win Probability, so that’s a good thing. But will it ever be useful for grading individual players, as in baseball? I’m not so sure, simply because the amount and arbitrary nature of the data on each play. The current world of player evaluation is ruled by “football guys” who make personnel judgments off of “what they see” and “gut instinct” (sound like marketing twenty years ago?), and I think they’ll always play a role, but analytics could become a useful tool at least. The problem is that success or failure in football is so reliant on the play of the entire team, making it tough to measure on an individual basis. For instance, if a player fails to gain yardage because one of his teammates didn’t execute his block, how would that be graded? And is it really worthwhile to measure things like a player’s success percentage on intermediate routes against zone coverage during night games on the road in December? Realistically, probably not.
Which brings me to the real world issue here, the Big Data “Rabbit Hole”. With so much data available to marketers, how do you decide which data are worthwhile without losing sight of your goals in the process? I think, as marketers, we need to move past the point of “Oooohhhh look at this shiny new data I can analyze” to the point of “OK, how is analyzing this data going to affect our bottom line”. Until we get there, I’m not so sure that Bigger Data will always be Better Data.
So what do you think? Is all data good data? Or is it “garbage in, garbage out”? And will the Bills beat the Chiefs this week? Let us know in the comments section!