Jeff Dominitz grew up rooting for the Washington Redskins during their glory years under Joe Gibbs. So when an NFL team called him in 2006 about a job doing statistical research and analysis for the scouting department, he could not say no. He left his teaching position at Carnegie Mellon and, temporarily, his family and wound up living alone near the team’s headquarters.
The first sign that Dominitz’s work might not be fully embraced came when he reported to the team’s facility. According to Dominitz, who has a Ph.D. in economics, he was told that the head coach had been informed about Dominitz after he was already hired. The coach’s response, Dominitz was told, was, “We’re still about people here.” Dominitz then learned that he would be seated not with the scouting department, compiling detailed information about player prospects, but at a cubicle in a separate building, with the marketing department. Seven weeks later, Dominitz was gone.
Dominitz later spent a few years working for another team, predicting how athletes would do in the NFL based on their college and pre-draft performances, as well as analyzing overtime decision-making, among other tasks. Now the senior director at a Beverly Hills, Calif., economics consulting firm, he has given a lot of thought to why the rigorous study of advanced statistics is gaining a toehold among NFL teams years after it swept through baseball and, more recently, the NBA.
“It can’t just be money,” he said in a telephone interview. “It’s got to be some concern about the impact on the rest of the organization. You could see it in them putting me in the marketing building, trying to have as little impact on the rest of the organization as possible.
“To me, it’s crazy not to try. It can only be a firm belief that it just can’t work or the process of making it work will fail because you’ll have so much resistance from the coaching staff.” Perhaps. But when the Baltimore Ravens announced in August that they had hired a director of football analytics, it was a rare public signal of the growing interest among teams in weaving statistical analysis into game-day, draft and free-agency preparation, and even into the management of workouts and injury rehabilitation.
Several companies now study games to produce unique statistical analysis, including Football Outsiders, Advanced NFL Stats, Stats LLC and ESPN. Few teams like to talk about the degree to which they use analytics because they fear giving away a competitive advantage. One general manager whose team does delve into statistics, but who didn’t want to be identified, wondered why the Ravens announced the hire at all.
This general manager suspects that more teams do some form of statistical analysis than are publicly known. People who work in sports statistics, and coaches and general managers, agree that there has been a shift in the NFL’s guarded thinking.
“The culture has changed exponentially in the last 12 months,” said John Pollard, a general manager for Stats LLC. “Last year was a jog. Now we’re at a really good trot. This January, I think we’ll be at a sprint with the mass adoption of tons of information services.” The question remains how the shift changes a team and the game. Gil Brandt, a former personnel executive for the Dallas Cowboys, points to one way the league is already different. With advanced statistics, he notes, teams are able to see trends and adjust in real time. It used to be that teams would look back at how often they ran a play and how much it gained. Now, do they want to know Cam Newton’s completion percentage when a defence rushes three? Or four? Or six or more? That information is available week to week, allowing teams to tailor game plans with far greater specificity.
Much of the work is also centred on figuring out some of the game’s most vexing problems – when to kick a field goal versus going for it on fourth down; what to do under the new overtime rules; when to challenge a call; when to use a timeout – amid the chaos of the sideline.
The Jacksonville Jaguars, who this year created a football technology and analytics group, have pondered all those questions. The coaching staff had expressed interest in having information about how to function under the league’s new overtime rules, which call for each team to get a possession unless the first team to get the ball scores a touchdown. But because the rule was so new, there were simply not enough comparable examples of game-time situations to produce a reliable model, said Tony Khan, the son of the Jaguars’ new owner, Shad Khan, and the head of the analytics group.
“Baseball is an easier sport to prove out some of these concepts, because there are less variables,” Tony Khan said. “You can isolate the defence behind them; it’s essentially a one-on-one matchup. And there are so many more plays in baseball that you can look at. If you want to look at every fourth down and five from that team’s own 36-yard line, there aren’t going to be all that many of those. And the 11-on-11 nature makes it harder to isolate credit for success and harder to isolate blame for mistakes. There is a reason why this caught on a lot sooner and developed a lot further in baseball than football.” For most teams, though, the most intriguing application may come in player evaluation – projecting how college players will perform in the NFL and figuring out how valuable one player compared with another. The Jaguars, for instance, analyzed how often the receiver Justin Blackmon was targeted in obvious passing situations at Oklahoma State before they drafted him. Before they signed Laurent Robinson as a free agent, the Jaguars knew how many of Tony Romo’s touchdown passes he caught on crossing routes with the Cowboys last season.
The Jaguars are also using data to monitor injuries and recovery, hoping to tailor players’ practice regimens to keep them healthier longer, but also another potential boon when contracts are negotiated.
“Ideally, you want the objective and subjective to match up,” said one general manager, who spoke on the condition of anonymity. “The NFL is about resource allocation – you have a certain number of salary cap dollars and draft picks. If you found any area of the market that may be undervalued, you want to keep that information. At the end of the day, the tape is going to be our first choice. They have to look good on film.” That seems to point to the central concern of coaches and statisticians alike: that there remains no perfect formula for assessing many of the game’s positions, no football equivalent of OPS, the on-base plus slugging average now widely used as part of baseball player evaluation. While the quarterback can largely be isolated (ESPN created QBR as a way to try to improve quarterback evaluation by assessing his contribution to scoring points), it is much more difficult to assess each position on, for example, the offensive line, because the players’ work is so tied together. And then there are the intangibles.
“How do you quantify, statistically, Ray Lewis?” said Brian Billick, the former Ravens coach who calls himself a statistical nut. “You can’t. You still have to have a core of great players. I can wrap a lot of B players around those four core players. But you have to have those four.” The debate over the role of statistic analysis is likely to grow as its use becomes more commonplace in one of the last holdout sports. But Brandt remembers the Cowboys’ internal discussions about whether to draft Tony Dorsett or Ricky Bell in the 1977 draft. One of the Cowboys’ scouts, Red Hickey, wanted Bell. Others wanted Dorsett.
The Cowboys had a book that listed previous players, their attributes and level of success. The most important determiner was quickness, the Cowboys believed. After a 30-minute debate, coach Tom Landry asked Brandt to check the numbers.
“Dorsett was a cinch to be All-Pro, and Ricky Bell had a 60-per-cent chance of being an above-average starter,” Brandt said. “Do you know what Red Hickey said? ‘I bow to the machine.’ It’s hard to teach an old dog new tricks. But this is a coming thing.”