Like body mass index, IQ and the gross national product, the NFL's passer rating roughly measures something important but yields deeply flawed results. It's always been a jury-rigged stat just waiting for the sabermetric revolution to kick out its struts and build something better in its place.
Well, ladies and gentlemen, the wrecking crew has arrived, and the extreme makeover has commenced. Welcome to Total Quarterback Rating (QBR), a new way of looking at QBs based on ESPN's analysis of nearly 60,000 NFL plays over the past three seasons. QBR looks at every facet of quarterback play, from passing and rushing to fumbling and taking sacks, and allocates credit or blame to QBs according to how each and every play they make contributes to their team's success.
Do we all really need another uber-stat? Well, yes, because there are two basic problems with traditional passer ratings: what they measure and what they don't measure. The official formula for passer rating is actually less complicated than its reputation. It takes completions, passing yards, touchdown passes and interceptions, all on a per-attempt basis, compares each to a league-average figure, and mashes them into one number. But passer rating doesn't attempt to weight its categories by their importance to winning football games. It just averages them together, which tends to bias scores heavily in favor of QBs who complete a lot of short passes, driving up completion percentage without necessarily generating more yards or points. It's even possible, absurdly enough, to improve your rating by throwing passes for negative yards.
Another issue: the league averages that passer rating uses to grade QBs come from the Paleozoic 1970s, when a special NFL committee put the stat together, and when football's rules and strategies were both far less friendly to passers than today. How much has the game changed since then? Well, in 1973, the year passer rating became an official stat, Roman Gabriel, then with the Eagles, led the league with 3,219 passing yards, a total that would have ranked 18th in the NFL last year. A QB's passer rating is partly, often largely, a product of the time in which he played, rendering era-to-era comparisons nearly useless.
And what good is a stat if it can't help settle who's-greatest brawls?
Let's look at what's left out of the official passer rating formula.
Michael Vick led the NFL with 6.8 yards per rushing attempt last season, but the system the league uses to rate QBs gives him no credit for the 676 yards or nine TDs his legs generated. Conversely,
Jay Cutler lost 352 yards on sacks, but
Osi Umenyiora could still be planting Cutler near Jimmy Hoffa and passer rating wouldn't notice. It counts the four categories it cares about, and only those four. Just as important, passer rating doesn't consider how or when a QB racks up passing yards, TDs or INTs. Throw for 300 yards and a couple of scores as you're trying to avoid getting shut out in a hopeless loss, and you'll inflate your rating; heave a Hail Mary jump-ball INT as the clock runs out in the first half, and you'll drag your rating down.
We shouldn't bash the inventors of passer rating too harshly, though. Don Smith of the Pro Football Hall of Fame, Don Weiss of the NFL and Seymour Siwoff of the Elias Sports Bureau heeded a call of duty from then-commissioner Pete Rozelle and did their best with the stats they had. And their formula is probably better than any of the eight other systems the league tried between 1932 and 1972. But passer rating is a perfect example of everything that can go awry with kitchen-sink stats, where an inventor throws together a bunch of data that looks important, breaks out his calculator and sees what he can come up with. QBR takes a different tack; it measures the connection between plays and points on the field, between points and team wins, and then gives credit where credit is due.
QBR starts with this insight: Any possession in a football game has an expected value -- the average number of points the team with possession can expect to score, based on all the historical outcomes for teams facing the same down, distance, field position and time remaining. And that means we can evaluate any play by how much it increases or decreases a team's expected point total.
For instance, if your favorite team is playing at home and has a third-and-7 from its opponent's 45-yard line with 14:55 remaining in the fourth quarter, it will score, on average, 1.8 points. Suppose your team then gains 42 yards on its next play, giving it a first-and-goal at the 3-yard line with 14:30 left in the game. In that situation, teams average 5.6 points. So the value of that play is the difference between 5.6 and 1.8, or 3.8 expected points.
It doesn't matter whether the play was a bomb, a screen pass, a draw up the middle, a recovered double fumble or a pass-interference penalty. It's worth 3.8 points, and if you add up the expected values added by all of your team's plays in a game or a season, you will get something very close to the number of points it actually scored.
QBR allocates the points added by every play in an NFL season to each of the players involved, every play. On completed passes, for example, it splits credit among QBs, receivers and blockers, depending on factors such as whether the quarterback was under duress, where he threw the ball, how far it traveled and how many yards the receiver gained after the catch. QBR splits the blame for sacks on quarterbacks and offensive linemen and attributes QB fumbles to QBs. Further, QBR weights every play by its clutch value -- its contribution to a team's chances of winning, given the score of a game, not just to scoring points.
And if you care at all about stats, that's the key: QBR finally brings all the advantages of
win probability to football. To determine who the most valuable player in the NFL is, compare the average value of, say, wide receivers and defensive backs, or figure out whether it's
worth it to onside kick, you need a system that tells you how much various plays affect a team's probability of winning a game. QBR does just that for quarterbacks. A 5-yard completion on third-and-3 is much more valuable to QBR than a 5-yard completion on third-and-15 because, in real life, it gives the quarterback's team a much better chance of coming out ahead.
QBR is scaled from 0 to 100, with 50 representing league-average performance. For a single game, a rating in the 90s is terrific; last year, Vick's six-TD Monday night symphony against the Redskins topped the charts at 99.8. For a season, any QBR above 65 is Pro Bowl-caliber, and
Tom Brady led all starters with a 76.0 QBR in 2010.
But those are just the top-line numbers. Delve into the finer details of QBR, and we can quantify the true greatness of
Peyton Manning, who has added 107.5 clutch-weighted points a year to the Colts' offense since 2008, significantly more than any other QB in the NFL. We can see who has killed his team worst over the past three seasons with sacks (Cutler, minus-55.5 clutch-weighted points in 2010), interceptions (Cutler, minus-44.1 in 2009) and fumbles (that's right, Cutler again, minus-15.6 in 2010). Vick helped his team the most with his legs last season (22.1 clutch-weighted points).
Matt Ryan adds more than a touchdown a year to the Falcons' offense just by inducing opponent penalties. We can also start to appreciate players and skills that used to go unrecognized.
In fact, we can … well, you get the point. Now that the lockout is finally over, we've got all season to play in this shiny new house. And the difference between QBR and all the ramshackle metrics that have come before is that this rating isn't just convenient, it's also comprehensive and meaningful. We're moving on up.
