Buffalo Bills and the Josh Allen Dilemma: A Fan’s Plea For A High Variance Offense
I’ve been meaning to write this article for a while, and it’s difficult to tell where to start. I think the best place is my feelings about Josh Allen when he was first drafted.
I was upset. I thought it was just about the worst thing the franchise could do, not only taking a QB whose archetype had failed time and time again at the NFL level but also trading up *twice* in the first round of the draft to take him with the *seventh* overall pick.
As I let the dust settle, I cooled off some. I tried to do something that is difficult to do – separate the player from the process. It’s going to be difficult to ever convince me the process was correct (much like with the Sammy Watkins draft-day trade), but it’s clear there were things I and others missed.
More importantly, just because I was disappointed with the process didn’t mean Josh Allen couldn’t succeed. I think a look back at the Football Outsiders QBASE rankings/projections sums this up best, albeit in an oversimplified fashion.
QBASE tells us you’d never want to invest the type of draft capital the Bills did into a QB prospect like Josh Allen:
“Since 1997, there have been 27 different quarterbacks chosen in the top 100 with QBASE ratings below zero. The best of these quarterbacks was either Josh McCown or Brian Griese. It’s a terrible group of quarterback busts. Negative-QBASE passers chosen in the first round include Mark Sanchez, Josh Freeman, Kyle Boller, Rex Grossman, J.P. Losman, and Patrick Ramsey.”
Despite the negative QBASE and comps, the projected range of outcomes for Allen by that very same system weren’t as doom and gloom as many analysts, myself included, made it seem:
Yes, the bust rate is alarmingly high, but more than a third of a time Allen is an adequate starter or better and upper-tier/elite over 15% of the time. Those certainly aren’t numbers that match up with Josh Allen’s cost, but they also aren’t the numbers of a QB that has no chance of being successful. Additionally, there are a couple of things that QBASE and analysts like me missed about Allen that suggest his range of outcomes should be looked at more favorably, albeit still with a skeptical eye. I’ll touch on those in a bit.
With this perception of Allen’s range of outcomes in mind and trying to disassociate the team process from the individual player, let’s evaluate Allen’s rookie year:
The Josh Allen Positives
One thing that became obvious in regards to Allen from the beginning is that his college rushing statistics hid his upside as a rusher at the NFL level.
Allen ran for just 727 yards between his 2016 and 2017 seasons for Wyoming, averaging a disappointing 3.1 yards per carry. However, college rushing numbers do include yardage lost on sacks so it is likely Allen was a better rusher than that number indicates.
Perhaps we shouldn’t be so surprised given his size and athleticism. It’s not as if Allen didn’t test well at the combine. He tested really well:
While his 73rd percentile SPARQ-x score isn’t quite Cam Newton (97th percentile) per PlayerProfiler.com, Allen is a big-bodied QB who can move well. That showed itself in his rookie season.
In just 12 games (only 10 he started and finished), Allen rushed for 631 yards off of 89 attempts (7.1 YPC) and scored 8 touchdowns on the ground.
While there are some people who cringe when Allen rushes, either because they are concerned he will get hurt (reasonable) or because they’re ignorant (something we saw often with Tyrod Taylor), the fact of the matter is his scrambles are extremely valuable. I’d go as far as to say they dramatically shift the risk/reward spectrum on Allen as a prospect. Perhaps he’s still not someone who you’d want to dump in all the equity that the Bills did, but likely someone with a more favorable range of outcomes than the QBASE chart above indicates.
Just how valuable were Allen’s rushes last season? According to NFL ScrapR (I used Python to look at the data), the average EPA (Expected Points Added) of Allen’s rush attempts last season was 0.465.
For reference, Patrick Mahomes posted an absurd 8.89 ANY/A last season when he threw the ball 580 times for 5,097 yards and 50 touchdowns against just 12 interceptions. Yet, Mahomes’ average EPA on those pass attempts was 0.44, less than Allen’s scrambles.
Now, before we get too crazy here, we need to point out a few things:
- QB scrambles, in general, have a high EPA, so Allen isn’t some sort of unicorn in that regard
- QB scrambles tend to be a noisy statistic year over year, so it’s not necessarily indicative of what Allen will be able to repeat in the future
- You can only scramble so many times (Mahome’s EPA may have been a little less on his attempts, but he had 6.5x the volume of Allen rush attempts)
The point is not to get carried away here. The point is that Allen’s size and athleticism allow him to scramble successfully, and those scrambles are of high value and should remain an integral part of his game moving forward if he and the Bills can help it. If they can, Allen’s rushing alone changes the floor-ceiling combination on his overall prospects.
One thing Bills fans will be quick to point out last season is how much Josh Allen improved from his first 6 games (5 starts) to his final 6 games (all starts), missing a month and a half in between due to an injury in the Houston game Week 6.
It certainly felt like the game slowed down a bit for Allen, and that he played much better. Do the numbers bear it out?
Yes, they do.
Over Allen’s first 6 games, he took 21 sacks while averaging just 6.0 YPA.
Over Allen’s final 6 games, he took just 7 sacks while averaging 6.9 YPA.
We’ve boiled a pretty complex game down to just two statistics without a lot of context, but they do the job for our purposes. Allen was much more comfortable after returning from injury, and the statistics bear that out.
The Josh Allen Negatives
I think at times Allen’s improvement over the second half of last season combined with some gaudy rushing numbers allow fans to shield just how horrific he was as a passer.
Allen completed just 52.8% of his passes. It’s true that completion percentage is flawed as a direct reflection of a quarterback’s accuracy, but that number is bad no matter how you frame it, no matter who was protecting him, no matter who was catching the ball.
Yes, Allen took shots down the field, but he ultimately had and ANY/A of 4.37. Of quarterbacks since 2009 who have attempted at least 200 pass attempts, Allen current has the eighth-worst ANY/A. ANY/A is not perfect, especially over smaller samples, but I like it because it includes sacks, touchdowns, and interceptions into its calculation.
Now, as I mentioned above, completion percentage isn’t a perfect reflection of a quarterback’s accuracy because it does not take into account things like depth of target, drops, and throwaways. Let’s try and dig deeper to find out if Allen’s supposed “fatal flaw” was as bad as it looked on the surface.
Pro Football Focus has an adjusted completion percentage that accounts for many things outside of a quarterback’s control. It’s not favorable towards Allen:
If you’re looking for some sort of hope, here, it’s that all four non-Mayfield rookie QBs struggled. Additionally, adjusted completion percentage does not account for the depth of each target.
Allen led all of football in intended air yards per target at 11.0. He was pushing the ball down the field, and the more you do that, the lower your completion percentage is going to be, regardless of accuracy.
However, even if we account for that, Allen was inaccurate. Here’s his completion percentage by depth of target when in a clean pocket:
The image above, courtesy of airyards.com, shows that at almost every depth of target, Allen had a completion percentage lower than the league average.
Furthermore, in addition to their adjusted completion percentage, PFF tracks accuracy percentage and avoiding uncatchable passes. Allen ranked worst in the NFL at avoiding uncatchable passes and third-worst in the NFL in accuracy percentage (beating only fellow rookies Josh Rosen and Lamar Jackson).
What Now For Josh Allen?
Pass Happy League
So, we’re left with an insane athlete at QB (poor man’s Cam Newton rushing the football with the strongest arm in the league) who struggles mightily with accuracy, which if we’re being honest, isn’t likely to markedly improve.
Let’s juxtapose that against what we’re seeing across the entire NFL. Much like the NBA gradually underwent a revolution leading to more 3-pointers, the NFL is undergoing a gradual revolution leading to more passing.
The math in the NBA was exceedingly simple, boiling down essentially to a modified version of “3 is greater than 2”. The math in the NFL, despite what those resistant to change will tell you, is almost as simple. Just as the expected value of a 3-pointer is higher than a 2-pointer, the expected value of a pass is higher than a rush.
Rather than sloppily recreate the work, check out these two pieces to see why:
Passing is more efficient than rushing, so why do NFL teams run so often? I’m going to run through some common answers to this question that have proven to be unsupported by data 1/
— new-age analytical (@benbbaldwin) February 15, 2018
JJ Zachariason on NumberFire: https://www.numberfire.com/nfl/news/19933/saquon-barkley-may-be-a-generational-talent-but-he-ll-still-be-overdrafted
We’re going to see more and more teams go to pass heavier, spread approaches – creating mismatches that can be exploited through the passing game, which is more efficient than running the football.
Where does that leave the Bills? This is where things get tough. If the Bills don’t transform to a pass-heavy attack, they are going to be operating from a mathematical disadvantage. However, if the Bills do transform to a pass-heavy attack, how do they counteract a success rate that’s bound to be worse than the league average given your quarterback’s primary weakness (accuracy)?
In my mind, the smartest thing the Bills can do is ratchet up the variance. If we go back to the basketball analogy, when underdog teams face heavy favorites, one smart strategy is to further increase the amount of 3-point attempts they take. Doing so naturally creates more variance in the game since it’s the shot with the highest point value but lowest success rate.
In the case of the Bills, I’m not suggesting they ratchet up the variance by passing even more frequently than sharp teams. I’m suggesting they focus on picking up chunk plays. Accept the inevitably below average success rate that is going to come with Josh Allen as your quarterback, and put him in a position to make as many big plays as possible.
The Bills have shown some encouraging signs of being open to this type of attack. For starters, last year Allen, as previously mentioned, led the NFL in intended air yards per target.
We also pointed out his improvement pre-injury to post-injury, and I think Offensive Coordinator Brian Daboll doesn’t get enough credit for this. Let’s look at the change in the team approach, pre-injury and post-injury:
|Time Period||Att||Deep ATT||Deep%|
As a quick aside, there’s a third party that’s responsible for Allen’s pre/post-injury improvement alongside Allen himself and Daboll: Robert Foster. Robert Foster was the man.
Of players with at least 33 targets last season, Foster ranked 20th in yards per route run (2.06) and second in yards per target (12.02). Find a way to get this man on the field, small sample size be damned!
Deep patterns not only create opportunities for Allen to pick up chunk plays and utilize that big arm, but they also create the type of space Allen needs to scramble, something it appears he and/or Daboll made a conscious effort of having him to do post-injury:
|Time Period||Rush Att per Game||YPC||Sacks per Game|
*I counted games 1 and 6 (Allen played partially in both games) as 1 game when calculating the per-game statistics above to provide a more accurate comparison.
Admittedly, when we divvy things up into small sample sizes, we can often see what we want to see. And what I cautiously see is a roadmap for the future – deep routes, space to run, fewer sacks (somewhat relevant, Adam Harstad believes sacks are a QB stat).
If we look at the offseason, there are some signs the Bills are continuing to go in this direction.
John Brown is a speedy, downfield threat. It’s going to be difficult for teams to make decisions when Brown and Foster are on the field together. Who do you give help over the top to? If both, who is spying Allen?
Then there is Cole Beasley. At first glance, I hated this signing. Where does Cole Beasley fit into the deep passing, high variance attack? While it was probably an overpay, the initial preseason games show that Beasley adds another dimension. Yes, we want this Bills passing attack to be down the field and high variance, but you need some mechanism to keep the success rate tolerable in the passing game. That dimension is Cole Beasley, who fits in the spread approach as a mismatch in the slot:
Cowboys Jerry Jones sure showed Cole Beasley what's what.
Bills Panthers 1Q
– Josh Allen 9/11 102 yards
– Beasley 5 catches – 5 targets 44 yards
– Tommy Sweeney 2 catches 47 yards
Buffalo Carolina pic.twitter.com/x0pm6RW5X1
— John Tharp (@JohnPTharp) August 17, 2019
The Bills also made meaningful upgrades to the offensive line, both in free agency and the draft.
Now, the preceding few paragraphs are where I pretend to be more of a football guy than I am. I am about 1% football guy (that might be generous), but I think intuitively you can see where I am headed with this. Perhaps a Bills analyst who is more of a football guy and good at breaking down film (@YardsPerPass) has more to add here. Bug him about it.
What I Hope to See
With all of that said, I hope the Bills show a lot of 4-WR sets. Get Brown and Foster’s speed on the field together. Get Beasley matched up with inferior coverage. Create rushing lanes for Allen to scramble.
I have vaguely addressed the Bills “throwing deep” and “taking shots” on offense. Let me clarify something here: we don’t need Allen to prove his arm strength by showing us how far he can throw the ball. Yes, the hope is Allen can convert some bombs here and there, but more importantly, he needs to be successful in the 15-30 yard range, something his 100th percentile throw velocity makes a bit easier. In fact, that’s where Allen showed the most signs of competency last season. He was average to above average in that range by Josh Hershmeyer’s PACR metric last season (from a clean pocket):
As you can see, Allen was horrific when he got beyond 30 yards. The samples tend to get smaller there so I wouldn’t take it as an indication that he’s definitively bad at throws of that distance. Regardless, the benefit of picking up extra yardage outweighing a lower completion percentage reverses itself past 30 yards:
Sticking this on the feed for the future. Longer passes have a higher expected value despite the lower completion percentage associated with them. pic.twitter.com/ah6LbpCWya
— Josh Hermsmeyer (@friscojosh) August 15, 2019
Furthermore, what exactly are chunk plays and are we sure they matter? If we define chunk plays as gains of 20-plus yards (admittedly somewhat arbitrary but easy to find), there was a 0.74 correlation coefficient last season between chunk plays and weighted offensive DVOA.
I’m still skeptical, for a myriad of reasons. Will Head Coach Sean McDermott be okay with a high variance approach? Is he willing to throw enough to keep up with the rest of the league (if the preseason is any indication, the answer is yes)?
And even if the Bills do commit to such an approach, it doesn’t mean it’s going to work. I believe it gives the Bills the best chance to realize a high ceiling offense and maximize Josh Allen’s talents, but the execution would still need to follow the strategy.
Step one is creating an environment for chunk plays to occur. Step two is the team capitalizing on those opportunities frequently enough that they are converting chunk plays at a higher rate than the rest of the league (making up for a presumed lower overall offensive success rate).
Step two didn’t materialize in the passing game last season when Allen posted atrocious yards per attempt numbers despite leading the league in intended air yards per attempt. The Bills finished 25th in the NFL in passing chunk plays, as defined a few paragraphs ago.
Still, it’s fun to envision an offense Allen can succeed in, even through skeptical eyes. When Allen’s name was called on draft night, I felt as if the Bills were doomed. Now, I see a path towards both the Bills and Allen being successful. Let’s roll the dice on a high variance offense and find out.