QB Josh Allen: Were His College Stats Underrated?

Statistical analysis pushed many observers away from Josh Allen during the 2018 NFL Draft process, but below I outline why his stats were held down by 2 factors: The team he played on and the teams he played against.

These talking points are nothing new. They’ve been debated many times over, but I don’t believe they’ve been properly contextualized. I’ll dive deeper here as I look at how Allen compared to other draft prospects in regards to overall team quality versus the quality of competition.

Trying to quantify the quality of a team can be challenging and that’s probably why there hasn’t been much work done here. The analysis needs to be subjective and there’s too many variables to quantify. However, I was drawn to 247SPORTS “Team Talent Composite” score to help. These scores use recruiting rankings to estimate how much “Talent” is on a college football team’s roster.

There are obvious downsides to using recruiting rankings, but I like using them for 3 reasons:

1. They’re readily available
2. On the aggregate, they tend to translate to team performance
3. Players are judged independently from others on their college team

Looking at the teams for the 5 QBs selected in the 1st round of the 2018 NFL Draft, you’ll see the glaring recruiting deficit for Wyoming vs the 4 other “Power 5” schools:

Now I’ll shift the focus to the QBs and, before I go further, it needs to be mentioned that Allen also played worse competition than the others. To compensate for that I looked at the “net” recruiting difference between a QBs school and their competition for each game played. To do this, I take the “Team Talent Composites” (or “TTC” as I’ll refer to it here) for 2015, 2016 and 2017 (they only go back to 2015) and create a “Differential” metric by subtracting Opponent TTC from Team TTC. Here are the aggregated results:

So you can see that on this metric, Josh Allen certainly appears to play a higher level of competition when compared to the talent on his team. Now, does this matter? To test that, I took the individual game Differential scores for each QB drafted in the past 3 NFL drafts and looked at the median Passer Rating for 5 evenly-divided Differential tiers:

This Differential Score methodology does appear to have an impact on median Passer Ratings for a QB. Next, I decided to run a simple linear regression on Passer Rating versus TTC Differential Score and come up with expected Passer Ratings for each of the QB’s games. Finally, I take the actual Passer Rating, subtract the expected Passer Rating and find the Rating Over Expectation(ROE). Here are the results in both per game and the aggregate:

As you can see, Baker Mayfield shines in this analysis, Allen, Darnold and Jackson are all relatively similar and Josh Rosen underperforms. This result honestly seems a little too good to me (based on 1 year in the NFL), as I don’t think that this analysis should be used as a primary factor for draft projections. Instead, I think the real value is providing context for evaluating a QBs statistical performance versus peers.

We’ve all heard phrases regarding Allen like, “No Quarterback has succeeded in the NFL with college stats as poor as his”. Often with the emphasis of “Especially while playing against Group of Five level competition.” I believe that this shows how that thought process is flawed and hope that this pushes the discussion in the right direction.

Brian Organ (@BrianOrgan)