- David Fleming, ESPN Senior Writer
- 0 Shares
IT WAS CLOSE to midnight when new Jaguars coach Gus Bradley leaned forward in his chair and brought his palm down to his desktop like a gavel. "Let's get to the truth," he said.
The 2013 NFL draft was six weeks away, and the Jacksonville front office, with the No. 2 pick, had some tough decisions to make, starting with the future of third-year quarterback Blaine Gabbert.
Since the Jags selected him 10th overall in 2011, Gabbert had gone 5-19 as a starter with a passer rating of 70.2, second lowest in the league. Bradley needed to know whether Gabbert, at age 23, was already a lost cause, or whether his lackluster numbers were at least partly a reflection of his less-than-stellar supporting cast. The future of the franchise hung in the balance, starting with the approach to this year's draft: Take a new quarterback, find help on the defensive front or select an offensive lineman to protect Gabbert?
Bradley had watched plenty of tape. He knew the QB's basic stats by heart. He needed more. So before leaving his office that night, he made a decision that would have been cause for leaguewide ridicule just a few years ago: He asked the team's head of analytics, a man barely out of his 20s who had never played football, to uncover the truth about the team's quarterback.
At 5:15 a.m., Tony Khan, the team's senior vice president of technology and analytics (and the son of Jags owner Shahid Khan), walked into Bradley's office and placed three handwritten sheets of paper on the coach's desk. Later that morning -- once Bradley had deciphered the penmanship and presentation, which had the look of a seventh-grade book report -- he began to understand the significance of the data Khan had compiled. This wasn't a grizzled old scout asking him to trust his gut. This wasn't a spreadsheet of stats any fantasy football player could acquire. This was something different, something more. This was the future.
The page titled "Blaine QB Rating Stats" revealed that when adjusted for drops, throwaways and spikes, Gabbert's passer rating in 2012 was a respectable 82.8. The next page, "Blaine Time in the Pocket," detailed the Jags' woeful pass protection: The line gave Gabbert an average of just 2.56 seconds to throw the ball. When he had more than 2.6 seconds to throw, his QB rating jumped to 84.5. The final page, "Blaine Under Pressure," showed that when facing a six-man rush, Gabbert ranked first among QBs in completion percentage.
Bradley was convinced that Gabbert deserved another shot. Six weeks later the Jags used the second pick in the draft to upgrade his protection, selecting Texas A&M offensive tackle Luke Joeckel.
THE ROOKIE HEAD COACH had asked for the truth about his young quarterback; what he received was nothing less than a referendum on the current state of analytics in the NFL. Next-level metrics have long been an integral part of major league baseball and, more recently, the NBA. Yet the NFL has been slow to catch on because of a general resistance to technology combined with the subjective nature of the sport.
At the heart of baseball is a one-on-one battle -- pitcher vs. batter -- that allows for easy collection of clean, accurate, predictive data. In football, though, there are 22 moving parts on each play, along with an infinite number of variables, including score, field position and down and distance. STATS LLC, one of the leading analytics companies serving the NFL, tracks 90 separate bits of data in a single play. "There's a ton of information in football," Tony Khan says. "But that's the problem."
Even when it is possible to narrow the parameters of the game using, say, QB rating, the data doesn't show the full story or offer a window into future performance. A wide receiver is only as good as his quarterback. A running back needs an offensive line. A speed pass rusher needs a nose tackle. As former Ravens coach Brian Billick likes to say, "How do you quantify Ray Lewis?"
You can't. But over the past decade, a handful of the NFL's more progressive teams, including the 49ers and Falcons, have been quietly searching for other ways to apply advanced data collection. In 2001 Bill Walsh hired away Paraag Marathe from his job as a financial analyst at Bain & Co. to modernize draft boards for the Niners and study the team's salary cap. In 2011, based partially on advanced metrics, the Falcons gave up five picks -- two in the first round, one in the second and two in the fourth -- to draft explosive wideout Julio Jones. The move was widely panned, but GM Thomas Dimitroff had quietly crunched the numbers and found that less than 15 percent of fourth-round picks become starters and that Jones' ability to stretch the field and draw double coverage would be exponentially beneficial to the rest of the Falcons offense, especially veteran tight end Tony Gonzalez. In 2010 the Falcons ranked last in the NFL in passes of 25-plus yards with 14. Last year they more than doubled that total with 33.
More than two-thirds of the league's teams are now crunching numbers full time, including the decidedly old-school Bears, who established their own analytics department this summer. One of the notable holdouts? The Steelers. According to an industry source, they've shown "no interest or curiosity in that direction."
Until recently, neither had the Jaguars. In November 2011, Pakistani-born Shahid Khan, an auto-parts manufacturer whom everyone calls Shad, purchased the team for $760 million. The next summer he hired his son, previously the general manager of a biodiesel plant in Indiana, to build the Jaguars' analytics group from the ground up. Tony Kahn had been interested in advanced metrics since he was in high school and studied finance at the University of Illinois. His first hire was Daniel Adler, who was less than two years out of Harvard, where he majored in economics and served as president of the school's sports analysis collective. Yet even with Khan's pull as the owner's son and Adler's Ivy League credentials, their work initially went largely ignored. But after the 2012 Jags finished 2-14, the worst record in franchise history, Shad Khan fired coach Mike Mularkey and GM Gene Smith, who had been with the team since its inception in 1994. In their places, he hired Bradley, then the defensive coordinator for Pete Carroll's new-age Seahawks, and Dave Caldwell, who, not coincidentally, was the Falcons' director of player personnel under Dimitroff.
Caldwell, 38, and Khan, 30, set out to revive one of the league's most moribund franchises with the kind of nerd-jock collaboration that usually works only in a John Hughes film. All the number crunching in the world won't be enough to get the talent-starved Jags above .500 this season. But how they fare in Year 1 of the turnaround could go a long way toward determining the future of analytics in football.
"There were times when we all wondered if football would ever get it," says John Pollard, general manager of STATS LLC's Sports Solutions Group, which provides advanced analytics to more than 20 NFL teams. "Once football grabs ahold of analytics -- and that's where we are heading into 2013 -- it can redefine the sport."
BUT IF THE geeks are to inherit the turf, they first must figure out how to bridge the considerable gap that remains between their computer screens and the football field. When the Redskins front office hired an analytics expert in 2006, coach Joe Gibbs housed him in a separate building away from the football staff and then cut him loose after just seven weeks. Most analytics guys have never stepped foot on a football field or inside a locker room, never felt the reality-warping effect of momentum, chemistry or nerves on the black-and-white data they collect. It's no wonder the word most often used to describe them is "insufferable." Says one member of an NFC front office: "These guys are convinced the data should sell itself. It won't. They either have to learn coachspeak or get used to being ignored."
When first hired, Khan regularly stayed up all night producing detailed analysis of an upcoming opponent's tendencies, only to sit and watch on Sundays as the data came to life on the field -- and was ignored by Mularkey and his coaches. Kahn knew they had to change their approach. "If we're sitting in some back room making beautiful statistical models that no one is ever going to use, then we're in big trouble," Adler says. "My job is as much about sales as it is about stats."
To show their desire to bond with -- and learn from -- the "football guys," Adler and Khan accompanied Bradley, his coaches and Jags scouts to the Senior Bowl, the combine and various pro days. They talked football, drank beer and worked side by side for days at a time. At the Senior Bowl, an excited Adler began explaining to Bradley, who was hired in part because he is receptive to new and progressive ideas, the analytics of fourth-down decisions, clock management, two-point conversions and overtime options. The first-time NFL boss was grateful for the information. But as a joke, he pretended to memorize the spelling of Adler's name on his credential so he could call him out in postgame news conferences when the advice didn't work. "It wasn't a complete 180-degree change from the year before, but it was at least 150," says Khan, who spends so much time at EverBank Field that he had a daybed installed in his office and lists the stadium as his home address on his driver's license. "We spent more time interacting with Dave and Gus in those three weeks on the road than we did with our previous staff the entire season."
By the time they got to the combine in Indianapolis, Khan was sitting between Bradley and Caldwell while they interviewed players. As the players talked, the coach and GM would occasionally lean back to scan information on Khan's beat-up iPad before asking a question. When an SEC wide receiver claimed never to have dropped a pass, Caldwell says, Khan waved his finger across his screen, then held it up in the air to contradict him.
That player's chances of playing in Jacksonville were finished, but Khan was just getting started. As the Outland Trophy winner, Joeckel would have been a no-brainer as a No. 2 pick, but the Jags already had a good left tackle, Eugene Monroe. So Khan dug up a Pro Football Focus study indicating that right tackles, Joeckel's likely position in the NFL, were becoming just as important in today's spread-heavy offenses.
To help evaluate defensive backs, Khan created a metric he calls Passes Touched per Target: the number of throws a player intercepted and deflected divided by the number of passes thrown his way. The formula helped confirm that the Jags had found a potential starting cornerback, Connecticut's Dwayne Gratz, in the third round -- absurdly late for a starter at that position.
Deeper in the draft, Khan turned to Speed Score, a metric created by Grantland's Bill Barnwell, then of Football Outsiders. The ratio of a player's weight and his 40-yard-dash time is designed to project which college running backs will perform best in the NFL. Barry Sanders had a Speed Score of 110.6, while anything below a 95 is a one-way ticket to the Arena League. Most people saw Michigan quarterback Denard Robinson's best chance of NFL success at wideout. But Jags scouts had other ideas, and when Khan plugged Robinson's numbers (200 pounds, with a 4.41 40) into the Speed Score equation, his 105.8 placed him just a tenth of a point below the Miami Hurricanes' Mike James, the second-highest-rated back in the draft by Speed Score. With their fifth-round pick, the Jags selected Robinson, who had coined his own position name: OW, for offensive weapon.
Not even Khan suggests that the Jags' highly rated draft class (Mel Kiper gave it a B+) was the result of analytics alone. But no one denies that having objective data to complement the subjective nature of scouting was an asset. So Khan's contributions solidified his department's value inside the organization, as well as the working relationship between the two sides of the team's brain. "At the end of the day, you still have to scout," says Caldwell. "It's never gonna be like baseball. But there's no question analytics has its place."
THIS SPRING, ANTICIPATING an increased workload, Khan and Adler advertised for summer interns on a handful of sports analytics websites and were shocked to receive more than 100 applications. The team hired four students -- from Washington, Iowa State, Princeton and Harvard -- who have spent the summer camped out inside the conference room attached to Khan's office. Dressed in Jags swag, they sit elbow to elbow around an oval table covered with spreadsheets, newspapers, soda cups, laptops, video remotes and a spaghetti tangle of USB cables, extension cords and chargers.
Seemingly oblivious to time, visitors, nutritional requirements and the smell inside the room (a distinct mélange of Axe, football socks and overheated computer batteries), the interns work with the joyous fervor of undrafted rookie free agents. "All we've ever asked for was an audience," Adler says. "We've got that. Now the challenge isn't to come up with the perfect football stat -- that would be impossible and paralyzing -- but to challenge the conventional wisdom while maintaining people's trust."
When they aren't busy evaluating draft picks, analyzing game-day strategies and mapping tendencies of the Jags' 2013 opponents, Khan, Adler and team riff on the grand analytical quandaries of the game. At the start of training camp, they debated whether it's easier for running backs to gain yards in an offense with a running quarterback. If so, should that devalue their rushing yards? And by how much?
When members of the Jags' football staff began popping into the room to join the debates or pose queries of their own, Khan and Adler knew they had reached a breakthrough. Football analytics will never reach its full potential if the information moves only from the computers to the field. For the science to flourish, the data has to flow just as freely in the other direction, using the lens of real football expertise and experience to make the numbers infinitely more accurate and useful.
So if Khan's group creates a report measuring defensive ends by sack totals, Caldwell might instruct them to factor in blocking schemes, score and perhaps partial credit for disrupting the quarterback. This interaction helps safeguard against a major problem in the still-fledgling world of football analytics: false confidence. "If we're using 20 percent of what Tony gives us, I want to figure out how to make that 25 percent," says Bradley. "If we can understand them a little better and they can understand us a little better, then we can create something really special here."
Before camp started, offensive coordinator Jedd Fisch asked the analytics department about practice schedules: Did the amount of time the team dedicated each day to individual game situations correlate to how frequently they come up in actual games? For example, were they burning 20 percent of their practice time on fourth-and-goal when it occurs only 1 percent of the time in games?
In his hands a few weeks later, Fisch had a report detailing every NFL play of the past five seasons. The first thing he noticed was the frequency, and importance, of second-and-long, a game situation that is often overlooked. So Fisch adjusted his planning, hoping to improve tempo, efficiency and performance.
Problem was, early in camp the execution at practice wasn't matching what the data had predicted. Denard Robinson -- playing in a wildcat formation -- fumbled three days in a row. Gabbert, meanwhile, struggled with accuracy, ball security and rhythm. In 11-on-11 drills during the first week of camp, he and the backup quarterbacks went nine consecutive throws without a single completion.
On one series, Gabbert stood tall and relaxed in the pocket, worked through his read progressions and fired a laser to the tight end for a big gain. On the next series, he was slow and awkward setting up in the pocket and got cleated in the shin by his own guard. Gabbert sprained an ankle and had to sit out the next practice. The QB Khan had described in his report before the draft hadn't shown up for more than a play or two during the first week of camp and was competing with journeyman Chad Henne for playing time.
That's the trouble with trying to define truth in the NFL. It changes with every snap.