Clustering the 2022 NFL Draft QB Prospects
Who do the prospects compare to in the NFL?
The 2022 NFL Draft is a little over a week away from the time of this writing. Overall, the general mood on this draft seems to be lukewarm at best. This screenshot from Peter King’s Football Morning In America sums up the feeling around the league on this particular crop of prospects:
This draft also falls short on the excitement factor, which is due in part to the perception of the QB class. According to Grinding The Mocks, no QB is expected to be drafted within the first 10 picks (according to Expected Draft Position), and only 4 have EDP’s in the first round. Last year the QB draft class had 5 QB’s with EDP’s just in the top 10. This doesn’t mean this crop of QB’s are doomed from the start, but rather their odds of succeeding appear to be a lot steeper than previous classes.
Regardless of how you approach the draft (film/numbers based, etc.), everyone loves to compare prospects to players currently playing or who have played in the league. Comparing things to one another is simply just part of our nature as humans. The numbers based approach to comparing prospects to other players is to use a technique very familiar to this newsletter: K-Means Clustering!
The Clustering Approach
Clustering is something I have done a couple times on this newsletter, including with the 2021 Draft QB Prospects. Essentially we are going to take a bunch of variables, standardize them so they’re all on the same scale, then feed those into the algorithm which will give us our grouped clusters. The NFL players we will use include all QB’s drafted in the College Football Playoff Era (2015 Draft - 2021 Draft) that have taken at least 70 snaps (~2 games) in their career. This was done in order to limit the amount of clusters and so we could make comparisons to players that have actually played in a NFL game. As for the statistics, a wide range of variables were used in order to amass as much information as possible:
Career Efficiency Numbers: The stats used here are the guts from ESPN’s QBR metric and QBR itself. This includes both passing and rushing points added, sack points lost, how often the QB rates in the 75th percentile of QBR’s, as well as how often they have games under the 25th percentile.
Last Year Of College Numbers: This gives us a profile of what they were right before they entered the league. Efficiency (EPA/Play), Explosiveness (explosive play rate and EPA/Play of successful plays) and other advanced metrics are lumped into this category.
Draft Profile and scores: Every year Next Gen Stats and NFL.com produce scores and grades for the draft prospects. The Next Gen Stats score uses the prospects athleticism and college production numbers to give each prospect a grade, while the NFL.com grade is from analyst Lance Zierlein. The prospects height and weight are also included here (Shoutout @_TanHo for compiling the NGS and NFL.com numbers)
Recruiting Rating: The only recruiting number taken into account was their HS recruit rating. This simply shows how the prospect was viewed coming out of High school.
Clustering Results
Here are the results of the clustering. The unmarked axis represent the two principal components that explain the majority of the variance in the data, which were determined by performing a principal component analysis (PCA). Six clusters ended up being a perfect amount of clusters for these group of players, and a quick scan through shows everyone is largely in a place that makes sense (Yes, Mason Rudolph did in fact have a very good college career!).
Another way we can look at our clusters is by arranging them into tiers:
A vast majority of the 2022 Draft class falls on their own tier, along with star Dallas Cowboys QB Dak Prescott and backup QB’s Brett Hundley, Cody Kessler, Kevin Hogan and Brett Rypien. Dak Prescott was a star at Mississippi State, but scouting saw differently and ultimately the Cowboys were able to scoop him up in the 4th round.
2022 Draft Prospects Notre Dame’s Jack Coan, Western Michigan’s Kaleb Eleby and Kansas State’s Skylar Thompson fall onto the tier of players that have largely not succeeded in the NFL. Of course, the one outlier here is the king among outliers: superstar Bills QB Josh Allen.
The remaining two 2022 QB prospects included are Western Kentucky’s Bailey Zappe, and Nevada’s Carson Strong. Zappe is grouped with guys that have gone number 1 overall in the draft (Goff, Winston, Lawrence), star in the making Justin Herbert, and other guys that were highly touted in their draft process. So why is Zappe not gaining as much attraction? Well, despite breaking the FBS record for TD passes, Zappe is seen as a mid round talent primarily due to the competition he faced, and the air raid style of offense he played in that typically boosts a QB’s volume stats.
Carson Strong was a strong QB1 contender in the beginnings of this draft cycle, but ultimately his stock has fizzled. Questions about the health of his knee surfaced, and ultimately his college production wasn’t the greatest. While the QB’s in his cluster may not be lighting the league on fire, they all have pretty much earned opportunities to play, which is all you can really ask for. The odds of success may not be high, but they are definitely not zero.
Cluster Analysis
By averaging key variables within the clusters, we can make broad observations about the QB’s in those clusters. The first three clusters stand out from the rest based on their college production. There is also a wide gap from the first three clusters and the rest when you look at their NFL grade. Cluster 3 tended to have a turnover problem in their final college year, which unfortunately followed some of those QB’s to the league (Im sure you have seen plenty of Jameis memes). Its also interesting to see how low Cluster 1’s average HS recruit rating was, given how much success they have seen in both in college and the NFL.
Considering the majority of our 2022 Draft prospects reside in Cluster 4, we will focus on that cluster. The biggest thing to point out is while posting average QBR’s and pass points added, they ranked 2nd out of the clusters in rushing points added. From Malik Willis and Sam Howell, all the way to Dustin Crum, this class is not short of QB’s with a dual threat ability. Aside from the rushing ability, there wasn’t too much other positives in regards to these cluster groupings. Sacks were a big problem for this cluster, which you can see from their average sack points lost.
Per NFL.com Grade, the cluster has the worst average out of the six clusters. That is a problem not only for the cluster, but for the 2022 draft prospects in general.
This shows that, while there is a good quality of depth among the class, the higher end talent just isn’t there. You could make the argument that without Trevor Lawrence the 2021 class mirrors the 2022 class, but when you couple that with college production (Next Gen Stats Score), you can see the 2021 class is still a much stronger class.
QB Similarity Scores
Our clustering gave us a cluster largely composed of the 2022 draft class. It revealed that this draft class is basically on its on island compared to the other classes in the College Football Playoff Era. We could stop here and conclude the best comparisons are to each other, but our goal was to compare the prospects to NFL players. To get our draft comparisons for the top QB prospects, will take the variables we used to cluster and create a similarity score between a prospect and the other QB prospects outside of their draft class. First up, the current favorite to be QB1: Malik Willis
Malik Willis
If you recall, we looked at just rushing production and concluded that Malik’s rushing comparison wasn’t quite Lamar Jackson. Well, when you throw in the passing production and everything else, Jackson jumps to the top spot in similarity score. As you can see the similarity is rather weak, and Willis expectations in the league should not be Lamar Jackson level. A range of Brett Hundley on the low end and Lamar Jackson on the high end is the sort of low floor high ceiling range that was followed Willis this entire draft cycle.
Sam Howell
Heading into the 2021 season, Sam Howell was coming off an amazing 2020 season and some had him as the #1 overall pick in early 2022 mock drafts. Unfortunately, Howell could not overcome the losses of some of his weapons and could not replicate the 2020 season. While his 2021 season wasn’t bad by any means, it still exposed some warts in his game. Aside from Lamar Jackson, this list doesn’t exactly inspire anyone. Sam Howell has the dual threat ability and arm strength you want to see in QB prospects. If he lands in the right situation with the right weapons, you may get lucky and see more of 2020 Sam Howell vs. 2021.
Desmond Ridder
Desmond Ridder will always be the QB for the first ever Group of 5 Playoff team, and has brought Cincinnati to new heights. The film community seems to love Ridder and his mechanics, but from a numbers perspective it just does not look as good. Again, the similarities aren’t very strong when you take out his 2022 counterparts, but this list isn’t very good either. Perhaps his stellar mechanics plus great situation could lead to a sequel to the Josh Allen miracle growth.
Kenny Pickett
Kenny Pickett had a meteoric rise up the draft big boards as he took Pittsburgh to new heights with an ACC Championship. His rise in play wasn’t as big of a jump as Joe Burrow or Zach Wilson’s blow up years, and his career before this year was not NFL quality by any standard. This reflects in his similarity comparisons, which consist of mainly backups and Daniel Jones, who may be out the door in New York next year. He will need to continue his fast climb up the QB play ladder if he wants to carve out a place in the NFL.
Matt Corral
Before we overreact, remember these similarity scores are without the rest of the 2022 class, which would have filled up the spots before these QB’s. These similarity scores are all under 50, denoting a very weak connection. Still, this is by far the best list of QB’s out of the rest of the prospects. Matt Corral had a lot of success in college, and finished with the highest Next Gen Stats grade in this draft class. Some knock him for the high RPO usage offense that he ran in college, and it remains to be seen how that will translate to the NFL game. One thing is for certain, if he starts his NFL career like any of these QB’s, you will consider that a win any day.
Conclusion
The 2022 QB draft class has been called a down year by most of the draft community. Using our favorite clustering technique, we found the class is basically on its own tier, and is more deep than it has higher quality talent. No matter how much information you acquire, a “surefire” QB can still fail and a “likely bust” could end up setting contract records. Clustering is a great way to take a large amount of data and group it in a way that is appealing to our minds and allows us to form the comparisons that we love to make. All that is left is to utter the phrase heard ‘round the world: “We shall see what happens!”
If you want to dive in to the data like I do, check out @CFB_Data and @cfbfastR on Twitter, where you can learn how to get started in the world of College Football data analysis!
If you want to see more charts and one off analysis, follow my twitter page, @CFBNumbers