What Happened To Football's Next Great Stat?

ESPN's QBR passer rating was supposed to be the rating that quantified the most complicated position in sports. Instead, it more or less disappeared.
Share |

It’s football season, and so men are shouting. The merits of their respective arguments are measured in proportion to the decibel level at which those arguments are offered. Business as usual at ESPN.

Specifically, five men are sitting around an oblong table, yowling hymns to powerful notions they are unable or unwilling to define: intangibles, willpower, heart, effort, desire. It could be any NFL pregame or halftime or postgame show, but this is ESPN's Monday Night Countdown, and it is helping viewers get ready for some football. Those five men shouting are not enough, and so there are three other men at the site of Monday's game; they sit around another oblong table, voices also raised. Of these eight men, six are former players or coaches, and the other two are ESPN Personalities Chris Berman and Stuart Scott. Across the bottom of the screen, one of the most complex, involved and fascinating statistics about football ever created—the result of years of work and multitudes of dollars that only a profit-machine like ESPN could afford to spend on a statistic—is scrolling across the screen. No one is shouting about Total QBR, though. No one is even really whispering about it.

Total QBR (or more accurately, QBR, since the 'total' is a marketing prefix) was fighting an uphill battle from the start. Jon Gruden, former-coach-turned-ESPN-shouter, welcomed it into the world by cautioning, “I’m worried about it because it sounds like a lot of stuff and it’s another quarterback statistical rating formula.” And this was Gruden, a quarterback-obsessive who, in our high-volume/low-wattage national football converation, qualifies as an intellectual. The rest of the football discourse—all limp n' loud machismo and soft-focus Tebow slashfic and sentimental pomp—is backwards and anti-thought enough to make Tony Siragusa look like Slavoj Zizek. This was the environment in which QBR was expected to grow.

Last year, after years of expensive and time-consuming research, ESPN unveiled QBR with great fanfare as a replacement for the conventional (and widely unloved) passer rating. It was supposed to change the way quarterbacks were measured and compared, and since it was coming from the Worldwide Leader in Sports, it was sure not be left in academic obscurity. And that would be, or would have been, a good thing: passer rating is strange, from its maximum possible score of 158.3 on down, and both over-precise and vague; QBR, on the other hand, was supposed to be easily accessible to the average fan and complex enough to encapsulate both the evaluative concepts that stat-nerds love and the dizzying range of facets that comprise quarterbacking. It was supposed to be a tough thread of reason running through the noise-field emanating from those former players and coaches around those oblong tables. The shouters might even have recognized it, in time, and used it to support or rebut those endless arguments about the Most Clutch Quarterback, or help make clearer just what it is that a QB has when he has that indescribable and undeniable "it." QBR was supposed to be all these things, and so of course it is none of them.

"It’s got a lot of really cool stuff in it," Aaron Schatz of Football Outsiders told me about QBR. "I wish they used it more." Instead, QBR has become a literal sidebar to your regularly scheduled NFL programming—another number vanishing under the crush of all that noise.

***

Until recently, ESPN wanted no part of advanced statistics. Websites like Advanced NFL Stats, Football Outsiders, and their equivalents in every major sport evolved and thrived in the vacuum created by ESPN's proud lack of interest. These online intellectual communities developed in response to the dearth of any such analysis in the mainstream, and created their own statistics to better understand what we saw. But the online stathead scene has always been just a community, never a mass movement. Advanced stats were unlikely to reach the type of person who watches Around The Horn. QBR was, at least in part, an attempt to bridge that gap.

At its core, QBR is an amalgamation of several other stats. Dean Oliver (that's him at left), who's widely regarded as a stats genius and is ESPN’s director of production analytics, consulted the work that Brian Burke and others had done at Advanced NFL Stats and made it the foundation for QBR. The resulting hybrid stat incorporates the concepts of Wins Probability Added and Expected Points Added, both used by Burke. But QBR does several unique things Football Outsiders and Advanced NFL Stats don’t do in their ratings: it combines passing and rushing values, doesn’t adjust for the strength of the opponent, and incorporates ESPN’s game-charting data to assign responsibility to the quarterback depending on each play's outcome. The game-charting is a labor-intensive task that only ESPN could do, Schatz says, and it allows QBR to account for dropped passes, value sacks differently based on the situation, add new wrinkles to air yards, or count scrambles and planned rushes differently, among other things.

Those are, obviously and inevitably, a lot of "other things." But that's quarterbacking, and taking the position's many facets all into account helps make QBR such a useful metric. "The goal of QBR is to come up with the best possible measure of how well a quarterback played," Schatz says, and he emphasized it was pretty good at this. (It's worth mentioning that Schatz, aside from his longstanding role at Football Outsiders, is a regular on ESPN programming. He noted several times to me that he understands how his praise might be perceived, but stressed he truly does value QBR as a good metric.)

As with everything having to do with ESPN, QBR is not without its controversies. Most notably, QBR is proprietary—as are Football Outsiders stats like DYAR and DVOA—meaning that the statistical models behind the stat are still secret. This is an old conflict: the challenge is popularizing a product without losing the ability to commodify it. It's understandable, but it has brought QBR in for some criticism. Schatz defended it—so long as the components of the metric are revealed, which ESPN has done with QBR. "It’s a black box, but I’ll tell you what’s in the box, I just don’t let you see it," Schatz says. "I tell you it’s a ham sandwich. I don’t tell you how many slices of ham it is, or how thick it is, but you know it’s a ham sandwich. It’s the same way with QBR."

Without a doubt, the most controversial aspect of QBR—and the one that would surely please ESPN's bellowing knights of the oblong table—is the Clutch Index, which weights every play according to the situation's import. In football as in baseball, mentioning the word "clutch" is the advanced stats equivalent of yelling "fire" in a crowded theater; people get panicky, incredulous and angry at once, start shouting, and ultimately try and run away as fast as possible. Stuff gets trampled. It's not a good look.

It's also not necessarily a defensible concept. If Quarterback A plays in a blowout, and Quarterback B puts up identical stats as Quarterback A but does so in a close victory that features a fourth quarter comeback, Quarterback B will have the better QBR. (Schatz has written that offensive performance in the first quarter correlates better with winning than performance in any other quarter.) As Schatz clarified, in QBR, "quarterbacks are not rewarded for being in clutch situations, but rather for performing well in them."

But nobody knows how ESPN defines a clutch situation or how it’s weighted, which only adds to the community’s nervousness over the concept. Still, most people outside the comment sections of Football Outsiders do believe in the concept of clutch, and ESPN built QBR for everyone.

Which is a big part of how and why ESPN has not succeeded just yet in crossing-over QBR—the stat is for everyone, which alienates the purists in the stathead community, and yet it took 600 words for me to fully explain the damn thing and so is clearly not for the “average” fan as ESPN conceives that fan to exist. Those who need something simpler than the traditional passer rating because of its infamous 158.3 scale have, in QBR, a metric normalized on a 0-100 scale. But it's not simpler, really, because it couldn't be simpler and even begin to do its job. And it could, arguably, not be simple enough to resonate with brainiacs like Woody Paige while still deserving to be called an idea. The market for QBR is, inevitably, strained.

***

After ESPN's early season QBR-propaganda blitz in 2011, it gradually stopped being talked about on TV. Jaws and Gruden mentioned it less and less on the Monday Night Football broadcast; by my unscientific count, it hasn’t been mentioned once this season. QBR stopped being featured on Monday Night Countdown, and its appearances on SportsCenter tapered off. It is prominent on ESPN.com’s sub-blogs and displayed on Football Outsiders side-by-side with their own stats, but seldom discussed out loud. On Monday Night Football broadcasts, ESPN uses the old passer rating. Nobody is shouting about QBR anymore.

Which makes sense, since this is not a stat built to be shouted about, and because shouting is a lousy way to have a conversation. QBR requires the kind of discourse Burke, Schatz, and their sites have created over years of conversation; it's a metric built on several thousand lines of code, years of investment, and several complex concepts, all of them reduced to one number. It's a great idea, in a discourse broadly resistant to ideas.

Brian Burke isn’t surprised to see ESPN giving QBR the sink-or-swim treatment. "You cannot possibly watch and accurately process all 16 games in one weekend, no less all the games of a season, including the playoffs and Super Bowl," he says. "The purpose of a statistic is to be a first estimate towards the truth, because your brain can’t process all that information. So it’s going to be flawed, but you have to accept that." What ESPN wanted—and what ESPN's way of talking about the NFL demands—from QBR was an argument-ender, a metric that doubled as truth-with-capital-T. But stats don't do that. They can't.

Instead, the complications inherent in advanced statistics only increase the number of things to argue about. "Every time you’re factoring something in, you’re bringing in a possible source of error," Burke says. "Inherent in all statistics is the Central Limit Theorem, which is that a lot of these immeasurables are uncorrelated with what you’re trying to get at. So sometimes it will be the case that a quarterback did make an awesome, exceptionally accurate throw, and sometimes it will be the receiver [who made a big play happen], but on average it’s going to wash out in the long run. That’s why you need the long run."

QBR is maybe best understood as a metaphorical representation of ESPN’s own Central Limit Theorem. In the short run, it seems like a failure because no one is talking about it. But it hasn’t yet been given the long run in which to show its worth. It’s stuck scrolling below the Monday Night Football/Countdown crowd, who are doubtless as skeptical as Gruden about the stat ("I'm concerned about it," Gruden said after it launched. "I’m excited about it. But I’m skeptical.") and almost certainly less excited. If football's designated talkers can’t coherently discuss their feelings about QBR—or if they're so dedicated to the ambiguities of clutch and intangible that they find the very idea of quantification offensive—how can they be expected to discuss QBR itself?

QBR deserved better than this. It’s too useful, too intriguing, and too rich to be dismissed, and not too complicated to be discussed. You don’t have to agree with the clutch index in order to see its potential. You don’t have to embrace its lack of adjusting for defenses to see the value in measuring quarterback performances. But as it stands, QBR is stuck. ESPN programming can’t talk about it coherently, and as Schatz succinctly stated, "This is not meant to be a metric for Brian Burke." So if the ESPN crowd is concerned-excited-skeptical-worried about QBR, but QBR is not sufficiently hardcore for the stat nerds, what is QBR's audience?

There is a chance that QBR is itself the statistical noise that will wash out in the long run per the Central Limit Theorem Burke described; in that case, ESPN programming will continue to be ESPN programming, and the idea of QBR will wither, gather dust, disintegrate back into the internet. Or QBR could be the start of a little renaissance, an idea good enough to last and maybe earn spots for Aaron Schatz, Brian Burke, Dean Oliver and others at those oblong tables. It's still too soon to tell QBR's fate. But this is an easy idea to cheer for, and not just because some quiet reason would bring the volume down in Bristol.

NOTE: For more on the "Clutch Index," and an explanation of QBR from architect Dean Oliver, check out this update.


Share |

Comments

"The purpose of a statistic is to be a first estimate towards the truth, because your brain can’t process all that information. So it’s going to be flawed, but you have to accept that."

This is a great summation of the point of a statistic and exactly why I can't understand why ESPN would spend years and (presumably) many thousands of dollars to create Total QBR.

Passer rating is not a perfect measure by any means. But the point is "to be a first estimate towards the truth." If you look at a year-by-year breakdown of passer rating rankings, you'd see that the stat does a perfectly sufficient job of representing the truth about which quarterbacks were the best for any given season.

The only odd thing about passer rating is that it's based on a scale of 158.3, which might be a problem if it was just being introduced today, but it's been around for over 40 years - long enough that most of today's NFL fans have grown up with and grown familiar to the odd scale.

The reason ESPN's tQBR isn't catching on is because there's no need for it. You can look at a list of the top 10 passer ratings for any season and be reasonably sure that you're looking at a list of the 10 best quarterbacks. There are no major flaws in the statistic despite the fact that it doesn't take certain elements into account (rushing, game situations, opponent strength, etc).