NFL

I try to do this weekly during the season. Third year now. The previous edition can be found here.

#What is ANY/A and why does this matter?

ANY/A is a QB statistic which correlates extremely well with wins. Aside from EPA/play (which as far as I’m aware is much harder to calculate as an individual, I think it’s the best available (to us plebeians, that is). Its full name is Adjusted Net Yards per Attempt, and it’s basically a modification of yards per attempt in the following manner:

(20 x TDs + passing yards - 45 x INTs - sack yards)/(pass attempts + sacks)

A higher number is better.

Is it perfect? No, obviously not. It doesn’t account for a lot of things, like fumbles and rush yards and pick sixes and several other things like torrential rain games which skew stats in weird directions, but it is a pretty good stat and is not too terribly difficult to calculate by hand.

ANY/A is often computed over at least several games, but most often over a season or career. This is because it’s very prone to volatility. As you’ll see below, QBs can put up ANY/As in a single game that can be far better or worse than their career ANY/A.

So, why should you care about this? Frankly, if you don’t care than feel free to ignore this post. I just like statistics and thought it was be interesting to put up. But my argument for it is that even at a small sample size, the larger ANY/A in a game seems to win a lot (generally because it means the QB played better than the other QB). So maybe it’ll give you one more aspect you can pay attention to in a football game.

With all that being said, here’s the data I’ve accumulated.

#Data Table for Week 12

#Grades

Now, what are those little letters listed after the ANY/A? Well, those are my grades.

Before you start screaming at me about your favorite player’s grade, let me just say that it isn’t just random where the cutoffs are. Specifically:

An A grade corresponds to a single-game ANY/A greater than or equal to 10.

A B grade corresponds to a single-game ANY/A greater than or equal to 7.5, but less than 10.

A C grade corresponds to a single-game ANY/A greater than or equal to 4.5, but less than 7.5.

A D grade corresponds to a single-game ANY/A greater than or equal to 2, but less than 4.5

An F grade corresponds to a single-game ANY/A less than 2.

Now, these grades aren’t like those in school, where the average ends up usually being a B, and very few people get D’s and F’s. This is a curve where the average is intended to be a straight C. Similarly, the vast majority of single-game performances also are intended to be a C, because C’s should be enough to be competitive in most games. As such, B’s and D’s are somewhat unusual but not extraordinarily so, and F’s and A’s are extraordinary games, for either good or bad reasons.

Note that single-game ANY/As do not necessarily match up with season-long and career-long ANY/As. Single-game ANY/As are much, much more volatile and will yield a much larger spread than a typical season-long ANY/A spread. For that reason, you cannot treat them the same as you would a season-long ANY/A, where an exceptional, MVP-caliber season would be an 8+ ANY/A. It’s kind of like the PFF system, where a lot of consistently good performances will get you a higher grade than one great performance and a bunch of mediocre performances.

Well then, what do the grades mean? While they are certainly somewhat subjective, this is what I intended them to represent:

A: This grade represents an exceptional game through the air. This performance was nearly flawless and is incredibly difficult to replicate game in and game out. It’s nearly unsustainable, even for the best QBs. There should be no more than a few of these per week. A string of these would probably result in the greatest season of all time.

B: This grade represents an excellent game through the air. While some mistakes were present, the good vastly outweighs the bad. These are certainly more sustainable than the A-graded games, and the best QBs can sometimes have these games for long stretches at a time. A lot of these games will probably put you in the MVP race, and there should be a handful of these per week.

C: This grade represents a mediocre or satisfactory game. There were mistakes and success, but neither vastly outweighed the other. This kind of performance will put most teams in contention to win most games, and the majority of QBs in a week will have this kind of performance.

D: This grade represents a bad game. Many mistakes were made, enough so that they significantly outweighed the successes. A team with a QB playing like this will be hard-pressed to win games, and if your QB is playing like this often, it’s probably time to look for a replacement. There should t be too many of these per week, but there should definitely be some.

F: This grade represents a terrible game. The QB had essentially none or very few positives throughout the game, far outweighed by the negatives. With this kind of performance, it’s almost impossible to win a game. A string of these warrants a benching almost immediately. Just like for the A’s, there should only be a couple of these per week at the most.

My opinion is that my cutoffs do represent these grades well. The vast majority of grades fall between B and D, with the most in C, and A’s and F’s are few and far between. I didn’t just throw darts at a board either, I spent quite a few weeks looking at the numbers before coming up with these cutoffs, sometime last year (unfortunately I do not remember exactly when). Obviously you can always make slight changes here and there, but I’ve been using this system for a while and it’s easier for me to keep using it than to do some statistics with standard deviations and percentages and the like and figure out what the perfect cutoffs would be. [And actually, most of the time, these track pretty well with the idea of standard deviation - usually the number of C’s is pretty close to the amount within a single standard deviation of the mean, and usually there are only a couple A’s and F’s, which are ideally meant to be outside 2 standard deviations from the mean. While this hasn’t been numerically tested or anything, I feel pretty good about it in general].

If you don’t like them even after this explanation, feel free to ignore them. I just made them for fun anyway.

#Tl;dr

ANY/A is a QB stat that tracks well with wins; larger is better. This includes a list of the single-game ANY/A for any QB who played significant snaps (aka both meaningful in number and meaningful in value) this week. There are attached grades which are somewhat arbitrary (I set the cutoffs once in the past but I don’t make changes to individual grades). I think they make sense and fit my goal, but if you don’t like them feel free to ignore them.