An analytics revolution comes for every sport sooner or later. MLB had Moneyball in the early 2000s and has moved well beyond it in the years since. The NBA has used efficiency to all but kill the mid-range jump shot. Soccer has seen an influx of countless new ways to measure passes and scoring chances down to the finest detail.
The NFL’s change became most evident in 2018. Computer models that looked at thousands of games found an inefficiency: Coaches were being too conservative on fourth down, when teams can either punt the ball away or go for an all-or-nothing conversion. That year, they got a little bit braver, attempting fourth-down conversions on 15 percent of their chances, up from 12 percent in the preceding few years. The quants, it seems, won the battle for football’s decision-making soul. In accord with various metrics, NFL teams now pass the ball more now than before; going into the current season, every NFL front office had at least one staffer, and often many more, primarily doing analytics work.
But somewhere along the way, football ended up with an analytics backlash. Across social media and on TV, fans and broadcasters are constantly pillorying the nerds. Last season, after the Baltimore Ravens coach John Harbaugh came up empty on a late two-point conversion to seal a loss, a crew of CBS commentators took turns hitting him like a piñata. “They’ll show you a spreadsheet and say, ‘This is why I made that decision,’” said Nate Burleson, one talking head and former player. Another, the Super Bowl–winning coach Bill Cowher, was blunter: “Paralysis by analysis. We overanalyze things. It’s not that hard.” You can find similar analytics hatred in the college game. After Texas Tech University faltered on a fourth down earlier this month, the Fox play-by-play announcer Gus Johnson said, “Analytics! Throw ’em in the garbage!”
Such is the crossroads where the sport exists in 2022. On the one hand, analytics have helped countless champions, and have made football, America’s foremost entertainment product, even more entertaining. On the other hand, the fancy stats are tearing football’s commentariat apart, and even inviting scorn from coaches who have spent their careers doing whatever it takes to win. The very concept of analytics has become a football bogeyman that no one saw coming. Maybe we should’ve.
In theory, sports are the ideal place for intense number crunching. The stock market and the weather are naturally numeric, but “we’re the only place where you have a scoreboard,” Alex Auerbach, a sports psychologist for the NBA’s Toronto Raptors, told me. “Sports already quantifies the most extreme way of benchmarking where people are,” he said.
The simple box score has been around forever, but even for casual football fans, advanced analytics are now unavoidable: Amazon Prime Video, the new rights holder for Thursday Night Football, runs a stats-y simulcast to its main broadcast every week. Player grades from the statistics-and-evaluation empire Pro Football Focus appear regularly on Sunday Night Football. Move into the depths of the football internet, and you’ll run into an alphabet soup of stats: expected points added (EPA) per play, completion percentage over expectation (CPOE), and DVOA (which nobody even knows by its full name). It is a Sunday ritual to see real-time, robotic evaluations of fourth-down and two-point decisions.
Some football fans adore these innovations. Others very much do not. On Twitter, a fourth-down robot’s assessment of a decision often leads to responses such as this one from last week: “I don’t want to see this type of ridiculous stat anymore.” Some football media, especially on TV, take a similar approach. “It’s still reflexively negative, like, ‘The nerds don’t really know what they’re talking about,’” Bill Connelly, an ESPN writer who covers sports through an analytics lens, told me. “The end.’” Analytics has become a catchall pejorative applied to any bold, unconventional decision a coach might make—especially one that fails. What happens, absolutely, is that “when people do a quote-unquote aggressive move, it is often ascribed as an analytics play even if the numbers do not say so,” Seth Walder, an ESPN analytics writer, told me. (Ironically enough, projection models shrugged their shoulders at the Ravens’ much-derided two-point try, seeing it as a toss-up.)
Plenty of coaches also recoil at how analytics have encroached on football. Consider the two most accomplished coaches of this generation: The University of Alabama’s Nick Saban and the New England Patriots’ Bill Belichick, who have seven national titles and six Super Bowl wins respectively. Saban has said he is “not an analytics guy” and described the job of a quant analyst as “some guy who hasn’t played football ever and he sits at a computer and he puts a bunch of stuff into a computer.” Belichick, meanwhile, once said about analytics, “I don’t care what they say.” Both coaches employ analytics staffers, however. Saban is famous for employing a small army of coaches whose job title is literally “analyst.” So, what gives?
Maybe this is all simple. Becoming an elite athlete, or a coach of elite athletes, requires a lifetime of work that goes well beyond figuring out the wisest analysis of data. The NFL’s precise motion tracking of players, often illustrated in moving dots, does not know the play call or a million other subtleties, and in turn, neither does all of the data derived from it. “If I want to know how to cook a beef bourguignon, I’m not going to ask Einstein,” Hugo Mercier, a cognitive scientist at the Institut Jean Nicod, in Paris, told me. “People have their areas of expertise. And even if people might be able to tell you that the MIT crowd is smarter on the whole than [an MLB] scout, they would still think that the scout knows more about baseball.”
For us fans, perhaps the whole contradiction comes down to the idea that numbers can be what Mercier calls “a black box.” Consider a computer that spits out the difference in pre-play win probability if a coach decides to kick a field goal instead of going for it on fourth down. Humans are geared to trust information sources that we can argue with, Mercier told me. There is no arguing, not really, with a fourth-down model.
I am a fan of the Pittsburgh Steelers, a normally solid team that currently has one of the worst records in the league. Before the season started, my more optimistic brethren had difficulty accepting that bad times were coming, even as various statistical analyses suggested an impending crash. “If you see someone on TV and they talk at great length about how the Steelers are not doing great this year, and for this and this reason, things are going to go bad, they might convince you,” Mercier told me. “But if you just see a statistical analysis that doesn’t explain its reasons, I don’t think it’s going to convince many people.”
In a sense, sports analytics are stuck on a hamster wheel. Many who have played and coached the game harbor natural skepticism about them, which comes out when they’re asked questions about analytics or talk about stats in their post-career media roles. Then the backlash filters into the public discourse and reinforces itself over and over again before regular audiences of millions. We value what athletes and coaches say about sports, the same way we trust what doctors say about medicine or chefs about cooking.
But perhaps the simplest reason for all of this resistance to analytics, in locker rooms and TV studios and everywhere else that football is played and watched, is simply that America has analytics fatigue. Escaping the algorithmic world that inundates us with an endless stream of information is impossible. I rely on a fitness watch that tells me exactly how long I slept and how hard my heart pumped for every minute of the day, then gives me advice on how intensely to exercise the next day. TikTok users can’t escape an opaque algorithm that queues up an endless scroll of videos. Political observers everywhere rely on a computer model that simulates an election and lets them track probabilities for months, or for a few hectic hours via a moving needle. Numbers are both the background noise to our daily lives and the battleground for so many of our societal fights.
But sports, after all, are supposed to be a form of escapism to take us out of these troubles. What we really want, up to a point, is to argue. In that sense, analytics should be a godsend. They’re an extra weapon in any fan’s crusade to talk about their own teams or their rivals. But the cardinal sin that sports analytics commit against our brains is to make arguments that are hard to counter on their face. I might tell my friend that their team’s quarterback has an inaccurate arm, and they might respond that, in fact, the QB’s aim mimics a precision missile. But if I then counter that the QB’s motion-camera-generated completion percentage over expectation is well below the NFL average, then what is left for my sparring partner to say, other than that the stat itself is junk? Where is the fun in that?
There is, of course, a way for an advanced stat to find approval with someone who believes they are skeptical about such things: It supports your argument. The analytics backlash “is kind of the same thing every year, but at least the teams change,” Connelly said. “The fan bases change that are yelling at me, because it really just comes down to: ‘If the numbers say what I want them to say, they’re good. And if they don’t, they’re ridiculous.’”