ESPN hates the NFL quarterback rating. They always have. After years of complaining and mocking the statistic every chance they got, they have finally decided to do something about it. Their solution is a bold one – they have created a whole new statistic. Their Total Quarterback Rating, which they have strangely abbreviated as the QBR, is designed to be a far more comprehensive and accurate assessment of what a NFL quarterback is capable of and how they are actually performing.
Before we take a closer look at what the QBR is and how it might be useful for football bettors I feel a brief need to defend the quarterback rating. It’s far from a perfect stat, and making it out of 158.3 is just awkward, but there are some things about the stat that are very powerful if used effectively. Most significantly, there is a very strong correlation between quarterback rating differential – the difference between offensive and defensive passer ratings for a team – and overall success. While ESPN has been very aggressive in their negativity about the QB rating smart bettors should not be in a hurry to reject it entirely until they have explored what it can do for them.
Now, back to the QBR. To develop this new statistic ESPN looked back at all the plays over three seasons in the NFL – nearly 60,000. Instead of just developing a simple statistic that inevitably has limitations they have developed an approach that looks at every aspect of quarterback play – from passing to rushing to getting sacked to avoiding the rush and so on. On each play the quarterback will be assigned only the credit or blame that they are due. That means that if they throw a perfect pass that is dropped by a receiver then it won’t reflect on the to the same extent as a pass that badly misses the receiver. All of the calculations will combine together into a single score between one and 100, with 50 being an average NFL QB.
It’s an ambitious goal. It’s also a very labour intensive one, and one that is open to interpretation. ESPN will have to have a person looking at every single play in every single game to evaluate them and their impact. They will also have to decide just how much credit or blame they are due on the play. There are a few potential problems with that. For starters, one person obviously can’t evaluate every game, so different people will be looking at different games. That means that it is quite possible that a quarterback could get different QBR numbers in a game based on who is evaluating the game. The differences might not be large, but if they exist at all then it creates a potential weakness. It reminds me of the Beyer ratings in horse racing. They are a good way of evaluating speed, but they are far from perfect because of the human interpretation that goes into assigning them.
Second, it is hard for someone to accurately assign credit or blame if they don’t know what the play was supposed to be in the first place. If they don’t know what the NFL QB and the other players are supposed to be doing then it is a challenge to be accurate in assessing how well they did that job.
Third, whenever there are so many factors that are combined into one single number it can be difficult to assess what needs to be improved if the statistic isn’t as powerful as it should be. With a simple statistic like quarterback rating it is easy for anyone to understand how to calculate the rating, and they can refine and adjust it as they desire. The QBR is incredibly complicated and inaccessible, so it is impossible for a handicapper without a statistical degree to fully understand how it is derived and how it can be tweaked or improved for their purposes. You are forced to use it exactly as it is, and that’s a real concern. If the number is consistently not performing as you would hope then you can’t easily adjust something to improve the results because you can’t easily tell what is causing the problems or what should be tweaked. As a general rule complexity is the enemy of good handicapping, and the QBR is nothing if not complex.
While I have some concerns and skepticism about the QBR as a football handicapping tool there are a few things I like about it. First of all, it should do a better job of assessing the real value of a quarterback to an offense. It has always been a challenge to evaluate dissimilar quarterbacks. For example, how does a guy like Michael Vick who runs so much compare to a guy like Peyton Manning who is in no hurry to leave the pocket unless he absolutely has to? Because the QBR looks at every aspect of QB play and looks at how it contributes to the final outcome of the game it should make it easier to compare these quarterbacks in a meaningful way – both over their careers and over a shorter time frame.
Another factor that is appealing is the clutch index. This looks not just at what the quarterback does but when he does it to assign significance. For example, a 30 yard completion is far more impressive in the final seconds when down by six then if they were down by 40 and playing against third stringers. A lot of NFL stats do an ineffective job of factoring out or properly valuing the actual impact of the play, so if the QBR can deliver on that promise then it could be effective.
QBR is not a miraculous answer for football handicappers, and it will take a lot of work to figure out if and how it can really be useful. Despite the issues, though, there is certainly enough going for it that it should be examined and experimented with a lot over the first few years it is used. It won’t instantly provide answers, but it has the potential to give some meaningful clues to how things will turn out and maybe help sports bettors make winning NFL picks.