This year’s Super Bowl matchup shows you don’t need a particular type of quarterback to win in the NFL. The Ravens’ Joe Flacco has 38 rushing yards this season. The 49ers’ Colin Kaepernick ran for 56 yards on a single touchdown gallop against the Packers a few weeks ago. But in the long term, when you’re building a franchise, which kind of signal-caller is the better bet?
Conventional wisdom says a runner is more likely to get hurt than a stay-in-the-pocket statue. Just ask Joe Flacco, who told the assembled press on Wednesday that “quarterbacks like [Kaepernick] are eventually going to have to become mostly pocket passers to survive in this league.”
This belief is shared by NFL personnel gurus. During the 2011 season, then-Colts vice chairman Bill Polian was deciding whether to draft pocket passer Andrew Luck or the mobile Robert Griffin III as Peyton Manning’s successor. Polian was ultimately fired before he got to make that call, but he let Sports Illustrated’s Peter King in on his thinking regardless. "I'd probably pick Luck,” Polian said. “When you boil it all down, you worry about running quarterbacks getting hurt."
But is this correct—are mobile quarterbacks like Kaepernick, Michael Vick, and RGIII, more prone to getting hurt than conventional passers such as Flacco, Peyton Manning, and Tom Brady? Anecdotes can’t offer a clear answer. Vick, RGIII, Manning, and Brady have all suffered serious injuries. Kaepernick’s career also demonstrates how contradictory the evidence can be when cherry-picked. He wasn’t allowed to run the ball in high school for fear of injury, yet he won his starting role this season because Alex Smith—who ran the ball much less frequently than Kaepernick—suffered a concussion, likely as the result of a quarterback sneak.
Theorizing about the risks of mobility can’t take us very far, either. Don’t Brady and Manning seem more secure, protected by a phalanx of 300-pounders at all times? Or maybe they’re in even more danger, since they survey the field while defenders charge at them from all angles. Mobile quarterbacks can at least choose to some extent which hits to take. They can slide. They can run out of bounds. And with their attention not always focused 30 yards downfield, they may be better able to prepare for impact.
We tried to shed some light on the injury question by collecting quarterback injury data and applying some basic statistical tests. Our first challenge was to decide how to measure injury. Since Football Outsiders’ “adjusted games lost” index goes back only a few years, we pieced together our own data. Players hobble to the bench during games and miss a few plays or entire quarters, but starts lost due to injury (as a percentage of total likely starts) is a commonsense measure of “how injured” a quarterback is during a given season.
Using online databases and news reports, we put together a list of regular-season games started and starts lost to injury—not suspension or benching—for each team’s primary starting quarterback between 2002 and 2012. This yielded 324 total observations over 11 seasons. We omitted a handful of seasons for which quarterback carousels made it difficult to judge who might have started had all QBs been healthy. We also omitted midseason replacements like Kaepernick, instead choosing one primary starter per team per year. (Based on his play as Alex Smith’s replacement, Kaepernick will go into next season as the 49ers’ starter and be counted that way in the next iteration of this study.) Next, for each quarterback season, we collected data on a set of variables we thought might explain injury rates: rushing and passing numbers, sacks, age, and weight.
The test we were most interested in was the most straightforward. If we separate the mobile quarterbacks from the conventional ones, which group misses more starts due to injury? We used a couple of different metrics to separate the Vicks from the Bradys: rush attempts per start and rush attempts per total number of plays called for the quarterback (what we call “rush share”). We also realized that our results might depend on whether we looked at games lost over a quarterback’s entire career instead of treating each QB season as a separate observation, so we decided to measure both. Finally, we ensured that each of the four total ways of separating “mobile” passers from the rest yielded a reasonable set of names. For instance, when mobility is defined by four or more rushes per start over a regular-season career, nine of 82 players in the dataset qualify: Michael Vick, Robert Griffin III, Vince Young, Daunte Culpepper, David Garrard, Quincy Carter, Colt McCoy, Cam Newton, and Tim Tebow. (Kaepernick would also qualify under the four-rushes-per-start criterion.)
As you’ll see in the chart below, regardless of how we sliced the data, there was no statistically significant difference in injury rates between mobile and conventional quarterbacks. Quarterbacks of both types tend to lose 11 to 14 percent of their starts to injury. Even without counting the thus-far injury-free Kaepernick, three of the four tests produced a lower injury rate for mobile quarterbacks. The gap, though, is small enough that a statistician would call it zero.
This was an intriguing finding. We were worried, though, that we had made too coarse a distinction between quarterback types, so we turned to regression analysis to understand if any combination of variables might explain why we see high injury rates in some quarterback seasons and low rates in others.<a name="_GoBack"> Again, we found no relationship between rush share and starts missed due to injury. This held even when we controlled for age and weight—neither of which were found to have an effect on health—and also for injuries in the season prior, which did turn out to be a weak but statistically significant predictor of injury in any given year.
It turns out that the only gameplay variable that explains injuries with any statistical significance is sacks. On average, a 1 percent increase in sack share—the percentage of plays called for the QB that end in a sack—is associated with a 2.6 percent rise in starts missed due to injury (0.7 percent standard error). This link holds when we use the career-wise dataset and when we use sacks per start instead of sack share.
Skeptics might protest the main finding for two reasons. First, running quarterbacks could have shorter careers because of their bruising style of play. If this were true, though, we’d expect them to sustain more injuries while in the league, and they don’t. Second, maybe types of injury are different between mobile and conventional quarterbacks. We tallied up all the start-quashing injuries suffered by each quarterback in our sample, not double counting any repeated damage across years to the same body part, and classified each into four categories: head/neck, hands/arms/shoulders, torso/back, and feet/legs/groin.
A comparison using this admittedly simple procedure reveals almost no difference between injury-type distributions for the two groups. In each group, the lower body took the most punishment, and just more than 10 percent of injuries affected the head or neck. If mobility is a major determinant of quarterback health, we would expect this comparison to reveal a difference, but the injury patterns are remarkably similar.
In sum, it seems that standing in the pocket is just as dangerous as scrambling around. Yes, RGIII left the Redskins’ playoff game with his knee twisted so badly that you hoped Fox was experimenting with in-game CGI. But when we take the long view, serious injury doesn’t discriminate based on one’s ability to race.
At least one puzzle remains, though. Since sacks are the only significant predictor of injury other than prior medical history, it is tempting to tell a story in which mobile quarterbacks evade sacks more successfully, thus compensating for the injury risk inherent in rushing upfield. Yet the dataset reveals that mobile quarterbacks are sacked slightly more often than are conventional QBs. We can show that mobility doesn’t have a negative effect on health, but we can’t explain exactly how running quarterbacks are able to avoid additional risk.
We aren’t necessarily jumping on the read-option bandwagon, either. There’s still plenty to debate regarding the effectiveness of mobile quarterbacks. They may be less useful in two-minute drills, and their main advantage—speed—may wane quickly with age. But we hope to have shown that, based on the evidence available, conventional fears about quarterback rushing and injury risk may be overblown. If a general manager wants to protect a top draft pick, he should shore up the offensive line and cross his fingers.