ESPN recently unveiled their “QBR” which is, purportedly, an improvement upon the NFL’s confounding Quarterback Rating. It probably is, but it suffers from many of the same deficiencies as my own “QBER”, all of which are nicely outlined by Brian Burke in his post on advancednflstats.
Weaknesses aside, one aspect of QBR that I really like is its adoption of Brian Burke’s “Air Yards” concept. “Air Yards” tries to divide the credit for a completed pass between the quarterback and the receiver by giving credit for the “air yards” the pass travels to the quarterback and the “ground yards” gained thereafter (or yards after catch) to the receiver. This, I believe, is the ultimate way forward for anyone who attempts to come up with a comprehensive “Win Score” win credit delegation for participants in the sport of football.
Last football season I tinkered with such a concept, but ultimately gave up. My concept would have been based on each football participants contribution to “forward yards”. If you look at it, although scoring has increased dramatically in the NFL, each and every season the number of points scored in the NFL per yard gained remains roundabout 0.065 points per yard (or a little more than six points per 100 yards, which makes sense). Meaning, I thought, that every yard advanced or denied could somehow be translated into points scored or denied which in turn could be translated into wins produced in NFL history. I still think it probably can, but it was the “somehow” part that tripped me up and forced me to abandon ship.
Here is a brief description of some of the problems I could not easily overcome.
First, to make the “forward yards” concept work, one has to count the forward yards produced by each punt. That’s tricky. If a team maintains possession of the football and is able to punt, in one sense it failed (it had to turn over possession), but in another sense it succeeded (by maintaining possession, it earned the right to push the opposition back further with a punt). But whom do you credit with the forward yards? The punter?? All he did was come in and kick the ball. But the punter must get some of the credit for the distance of his punt. Yet, if you credit him with creating every forward yard that the ball advanced as a result of his punt, suddenly each punter looks like one of your most valuable players behind your quarterback. That’s ridiculous.
Second, and in the same vain, is the placekicking aspect of the game. When a kicker comes in and boots a field goal, it advances the ball, but in a different way. So how do I translate that into “forward yards”? Same thing for kickoffs. Which portion of the return yardage do I credit to the kicker?
Finally, the most important cluster I couldn’t unravel: shared responsibility. In football, nearly every play features some form of shared responsibility. A QB cannot throw a complete pass without the line blocking defenders from tackling him, and the receiver separating himself from the defense and catching the ball. An RB cannot advance the ball without some blocking. All of these things have to be accounted for.
Its my belief that it can be done, and I think the philosophy behind Burke’s “Air Yards” is ultimately the platform that must be used to do it. In some way, the metric must give each player credit for the forward yards he creates or helps create or helps deny. But that kind of a system is not only labor intensive, it opens up many difficult areas of distinction.
Let’s say, for instance, you decide that running backs shall get all credit for the “forward yards” they accumulate after contact or after they encounter an open tackler, and the nearest linemen get all the other yards created. Ok, but what if the RB is an idiot who cannot remember where the assigned hole is that he is supposed to run through, and constantly runs into apparently unblocked defenders even though obvious holes are created for him? Do you blame the linemen for his failure? How can you?
Beyond all that, a “data specific” wins produced metric for football would be time limited. If one is dependent upon “play-by-play” one is necessarily limited in how far back one can go with his numbers. That stinks. One of the great things about any “win” statistic is that it gives you the ability to compare players from different eras. A “data specific” football metric could not do that.
So for now a “wins produced” metric for all 22 football players remains a pipe dream. At the moment the best we can do is Professor David Berri’s wins produced metric for QBs (“QB Score“) and for RBs (“RB Score“) alone, and Brian Burke’s WPA and EPA. (which credits players for the impact their plays had on the ultimate outcome of the game).
However, if and when a more comprehensive metric is created, I think it somehow must be based on something similar to the philosophy behind Air Yards.