Without question, Green Bay Packers tight end Brandon Bostick blew it when he muffed an onside kick in the waning moments of this year’s NFC Championship game, giving the Seattle Seahawks all the room they needed to take the game to overtime and win, earning a trip to Sunday’s Super Bowl.
But was it his fault the Packers lost? Not even close, says Louisville-based data analytics firm Edj Analytics. The NuLu startup has developed a simulation model that it’s marketing to NFL teams, crunching data to provide guidance on key decision points in games. The goal is to increase a team’s chance of winning — and not just scoring points.
It’s one part of a strategy for Edj that seeks to capitalize on our technology-driven obsession with data — and our limited understanding of how to put it to good use. The company’s growing portfolio, which includes partnerships to lower readmission rates in hospitals and attrition rates in higher education, suggests there’s a strengthening market for good information and the answers drawn from it.
But it all started with football. In 2006, Frigo and Chuck Bower, a former astrophysicist at Indiana University, began looking for ways to create a simulation model that would help notoriously risk-averse NFL coaches improve their play calling. Frigo, a former analyst at Royal Bank of Canada and world champion backgammon player, says the idea was drawn from game strategy: find ways to enhance your overall likelihood of victory, and stop getting so caught up in point spread.
“Big data right now is front-running good analytics,” Frigo says.
In other words, data doesn’t tell you much prima facie. This is a rather obvious problem in sports, where commentators often delight in sharing a player’s or team’s statistics without connecting them to the game you’re watching (probably because they have nothing to do with it).
Data without rules is basically a parlor trick. So Edj maintains a disciplined approach to its problem solving: What information is available? What are the constraints? And what is the goal?
In 2013, Frigo and Bower partnered with Sean O’Leary, founder and former CEO of Genscape, to create Edj Analytics. They’ve marketed their NFL model to a third of the league’s teams. But so far, they say, the coordinators and assistants they’ve met with haven’t been ready to tear up their playbooks — or even edit them.
Maybe the Packers should call a meeting. According to the company’s analysis, the absolutely terrible, cripplingly conservative play calling of Green Bay Coach Mike McCarthy did the Pack in way before Bostick seemingly head-butted the game away. We’ll let Frigo take it from here:
On fourth down, McCarthy decided to kick field goals twice at the Seattle one yard line, and again at the Seattle 22 yard line with a fourth and one. He also elected to punt in the 3rd quarter from his own 48 yard line on a fourth and one. As we have discussed so many times in the past, these are VERY poor decisions. In general, the science overwhelmingly supports more aggressive actions on fourth and short. The Edj NFL model accounts for score, clock, ball position and the relative strengths and weaknesses of the opponents. After simulating literally millions of unique outcomes, the relative cost of a poor decision can be quantified. The cumulative effect for McCarthy and the Packers was far more costly, on average, than Bostick’s single lapse of judgment. Brandon Bostick was willing to take accountability for his poor decision, will Mike McCarthy do the same?
Despite the obvious appeal of the company’s NFL model – gutsier play calling would certainly enhance the fan experience — Edj is finding its niche elsewhere. The company has grown its business sevenfold in two years, O’Leary says, and is on pace to go 10x by the end of this year.
In health care, it is using new performance-based requirements in the Affordable Care Act to help hospitals lower readmission rates — in part, by analyzing patients’ circumstances to more precisely measure the likelihood that they’ll have to come back soon.
If a patient lives alone, for instance, he or she is less likely to meet their medication schedule or go to the pharmacy on time. A provider could plan for that by outfitting the patient with a precise schedule to keep handy, or connecting him or her with a social services organization.
In education, the company is developing models to lower college attrition rates by better pairing students and schools before enrollment. And Edj has developed models to enhance trading in financial markets along with an unnamed partner, although the founders aren’t about to give up any trade secrets here.
“Some of the things you can really figure out [using data and analytics], they’re interesting and cool and meaningful, and they matter,” O’Leary says. “There’s a lot of stuff done internal to companies, to help them be more efficient or to save money, and that’s fantastic. Those are all important. We’re looking at trying to solve some bigger problems.”
But they’re not giving up on the risk-averse NFL, either. On Sunday, they’ll run the Edj model — designed specifically for the Seahawks and Patriots — and, if it yields something interesting, post about it online for all the Monday morning quarterbacks out there. Maybe Mike McCarthy will take a look.