Football's Moving Target
Issue   |   Wed, 09/19/2012 - 01:23

I don’t know that television stations can keep track of how infrequently viewers channel surf while watching their programming, but if they did, I think football games would win out in keeping remotes glued to coffee tables everywhere.

Admittedly, there are can’t-miss moments in every sports game, but between the 22 guys on the field looking to smash into each other, the constant potential for a game-changing score, and of course my fantasy teams, I find myself reluctant to miss a single down. It’s no coincidence that when the NFL holds its biggest game of the year, even the commercials are talked about just as much as the game.

So what’s there to be seen watching all the snaps? I buy in when players and coaches drone on about how every play matters. For a long time I didn’t actually believe it, but I’ve had my heart broken often enough to see the point. It’s a brilliant truism, if only because the idea that some plays don’t sit well with anyone. But assuming that every play does matter, and there really is a good reason that football games don’t just skip to the fourth quarter, what does every play amount to? What could that three-yard gain on 3rd and 20 in the first quarter be worth? How much did the holding penalty that negated a touchdown just before halftime cost? How much hope could a last-minute fumble deep in the opposition’s territory provide in a two-point game?
Seventy-four percent’s worth, give or take.

That’s how much that last scenario improved the New England Patriots’ chances of winning their game over the weekend, at least according to the winning percentage tracker Advanced NFL Stats has on their website. With just under two minutes to go, the Arizona Cardinals were sitting pretty: ball in their possession, a slim lead, only one Patriots’ timeout away from running out the clock for their second last-minute win of the season. Advanced Stats gave them a 92 percent chance of winning the game at that point. In the emotional roller coaster that followed for Patriots fans, the Cardinals’ Ryan Williams fumbled the ball, the Patriots recovered it, and with no timeouts remaining, clawed their way onto the Arizona 24 to set up for a 42-yard field goal and the chance to steal victory from the jaws of defeat. What was an eight percent chance for a victory ballooned to eighty-two in the span of a minute. Except those jaws clamped down as suddenly as they had opened when the field goal went wide.

A fitting end, since that game was nothing if not surprising. There’s a small fortune in the Nevada desert awaiting those of you with the foresight to predict such the outcomes of such games; for the rest of us, there are tools to incrementally improve on guesswork.

Effectively modeling football games has long been under the purview of the Madden franchise, but the NFL community has slowly seen a backdoor to the same kinds of insight creak open. The inspiration came from a previous favorite American pastime.

The cottage industry now known as sabermetrics (the name describes objective baseball metrics developed in large part by members of the Society for American Baseball Research) came from the machinations of a brilliant few, willing to sit down and examine the game they love in closer detail. Baseball fans were lucky in that sense; it’s a game all about numbers. For all intents and purposes, the action is limited to pitcher and batter, dancing around each other’s moves one pitch at a time. Elements of this discrete setup of in-game events are also present in basketball and football, but few games rival the calm math geeks seen on a baseball field. Incubated all over America for the past few decades, advanced stats found their way into baseball organizations eventually. Most important, easily digestible stats have now also made their way into living rooms, showing everything from a batter’s averages in different sectors of the strike zone (his “hot zone”), to his batting line given a particular count.

Basketball also recently saw the rise of more advanced bookkeeping. Plus/minus stats, which simply aggregated the change in score during the time a player spent on the floor, were among the first class of stats to be taken up.

Now the NBA’s StatsCube has reams of data available online for those with the wherewithal to make sense of them. In addition, Zach Lowe, who until recently ran the Point-Forward blog at Sports Illustrated, wrote extensively on a new frontier for basketball stats. A start-up named Synergy Sports decked out several NBA arenas last season with a slew of cameras that recorded all the action. Specifically, the same technology that had been used to track the precise position of baseballs and tennis balls, was now watching a basketball, along with the players dribbling it.

By tracking player movement as well, Synergy’s technology opened up a whole new dimension of analysis. Some of the generic reports Synergy now generates for its NBA clients include data on how high players get on their jump shots, how often they sprint at full speed in the fourth quarter, and just how high the perfect arc on a Ray Allen three-pointer is. So don’t be surprised if NBA broadcasts include information like this next year.

Football, in contrast, not only involves more moving parts than either baseball or basketball, but also has parts that are vastly more interconnected. So the learning and adoption curves have been much steeper in the NFL when it comes to advanced stats. To generate useful insights from advanced data, one needs to capture all the happenings on the field, identify essential bits, and make sense of them. The fundamental problem with football is in the first two steps. Quantifying data from the movements of 22 players on a single play, then culling through it is easier said than done.

Even seemingly simple events, like touchdowns, are conceptually difficult to model, because they can come from any place on the field, offensive and defensive. Making sense of moving the ball is another big hurdle. Is a 30-yard run from deep in your own territory more valuable than a long catch in the red zone?

What we’re left with for now are insights from the fringe, pending a more perfect football model. Trial and error is really the only way to move forward from here. In the meantime, teams will have to be content with their sideline iPads simply functioning as fancy playbooks.