NFL franchises spend tens of millions of dollars to scout prospects, so of course I can improve the process using 8 hours of spare time and my remedial understanding of statistics. I’m of course joking, but I set out to see if objective numbers from physical characteristics and combine results could add something in terms of quantifying potential. The exercise has limitations. No testing allows you to quantify the player’s ability to quickly process visual information. Some positions are so heavily influenced by processing speed that the correlation between physical abilities and eventual performance is thin. This includes QB, CB, RB and LB. The other positions still require processing, but the core duty of the position relies heavily on the player’s ability to accomplish some defined task at the beginning of the play. Much of a DE/DT’s success depends on their ability to rush the passer and beat the blocker in front of them. The OL know their blocking assignments pre-snap, and a WR/TE largely knows their route pre-snap. Because of the pre-defined nature of their core task, physical traits played a larger correlation to eventual performance.
The source data is also questionable. Many prospects do not even complete all of the drills. Those that do have significant variance in their combine preparation. Much like the SATs, results on the combine drills can significantly improve with proper preparation. Some prospects do not have the resources to prepare, and some simply don’t care to prepare because their tape is good enough.
Physical traits are also 1/3 of the overall equation. Technique/mechanics are not measured, nor is their acumen to learn. At best, the scores developed (YAS) could show the likelihood that a given prospect’s physical traits checked off 1/3 of the boxes for the overall equation. The draft is like throwing darts at a dart board though, and if some new information can increase your hit rate from 35% to 38%, that methodology has value. If a YAS can help identify potential UFA that can contribute, that also creates value.
The overall score for each position was crafted differently. Certain characteristics and times matter more for certain positions (3-cone splits for DTs), and some do not matter at all for some positions. Many of these were surprising (timing of the last 30 yards of the forty mattered significantly for WR, and arm length only mattered for DEs that rely on speed-to-power rush techniques). Using past combine data, equations were crafted that appear to link old combine data with eventual performance, which fit a statistical normal distribution. There are always outliers (Orlando Brown), and the equations will constantly evolve. Also, the times for this year’s combine were adjusted during our calculations. For example, 40 times were adjusted .05 slower due to a new track that appeared to produce historic results.
Much thanks to The Beast, a draft guide by Dane Brugler, which makes a year’s subscription to The Athletic 100% worth it. There would have been no way to collect all of the drill times and measurements without his guide. With no further delay, here we go:
Two, very different, equations had to be used for DE YAS score calculation, each with a range (essentially) from 1-10. Most of the successful past performs had a high score in at least one of the two equations (Speed and Power), and some had a high score in both (Watt, Ware, Bosa Crosby). Surprisingly, arm length was only pertinent for what I would call a power rusher. There were past busts that had high scores (Vernon Gholston, Chase Winovich, Chris Wormley), but aberrations were limited.
Aiden Hutchison, for all of the talk about his physical limitations, scored extremely well for speed and power rushing. Even his short arms were not able to knock his power rush score below a 7. Travon Walker had a 7.38 in power but very little on speed. Kayvon Thibodeaux was the biggest surprise, with no scores over 4. However, he did not appear engaged during his combine performance, so his scores may be a factor of apathy. Most of the remaining highly rated prospects had sub-5 scores in both power and speed. Drake Jackson had a 5.32 in speed and Dominique Robinson posted a 5.82, putting both above their likely draft slot. The Dallas Cowboys have been linked to Sam Williams, who had a 5.6 on power. In terms of late-round or UFA flyers, Mike Tverdov, Luiji Vilain, Joshua Onijiogu and Brayden Thomas breached the 5 point mark. We also have high scores from LB Brandon Smith, who would need to make the adjustment to edge. I am sure that perennial pro bowlers will come out of the Johnson/Karlaftis/Ojabo/Mafe/Enegbare group, who had relatively low YAS. When they do, their data will help us adjust the Edge YAS in the future when identifying what we missed.
Again, two different DT YAS had to be developed. One group was primarily for 3-techniques, and the other was for noseguards. The sample size was too small on noseguards to consider that calculation valid. As for the rushers, their 3-cone drills relative to their weight was a significant factor. Jordan Davis and Perrion Winfrey did not complete enough of the combine drills to develop a score. Of the rest, Travis Jones was by far and away the high scorer. Even when his weight influence was capped, he produced scores on par with Fletcher Cox. In terms of late-round flyers, Thomas Booker, Curtis Brooks, Jordan Jackson and Kurt Hinish had scores significantly above their expected draft range.
Again, two types of WR YAS equations were necessary, for cut and for body. “Body” receivers were not necessarily slow, and included players like D.K. Metcalf. The ”last 30 yards” of the 40 turned out to be a new measuring tool, that was extremely relevant to WR YAS. I actually suspect that the MOST important tool for speed would be how fast they run the distance from 10 yards to 30 yards, but since no 30 yard splits are taken, this hypothesis was impossible to test. Historical outliers for the derviced equation included Josh Reynolds and Denzel Mims, as well as (on the negative side) Davante Adams and Stefon Diggs. Wilson, Williams, Drake London and Chris Olave did not complete enough of the combine drills to obtain a result. Treylon Burks did not appear to have sufficient lateral quickness to obtain separation via cuts, and insufficient (height/length/vertical) numbers above the median to qualify as a ”body receiver”. He also exhibited pedestrian speed/agility results so the model did not assign him a high WR YAS. Christian Watson had outstanding WR YAS for both cut and body. Alec Pierce and Kevin Austin also posted above-average cut and body scores. Calvin Austin posted a 7.04 on a cut score, but his height is an issue. He may provide data moving forward regarding a ”minimum height requirement”, which could override WR YAS. Of DEEP prospects, Samori Toure posted an eye-popping cut score, and there were some interesting results from Jalen Brwoder, Dahu Green, Tay Martin and Tyshaun James.
The last 30 yards of their 40 times were, again, important in the historical data modeling. The 3-cone was also important, along with length/height/vertical scores. I think TE is the position where statistical modeling could be of the most use, because they are rarely fully-developed in college. There simply isn’t enough time to teach pass blocking, run blocking and route running before the NFL level, so this position requires the most projection. Of this year’s crop, there are Jelani Woods and Charlie Kolar, and then there is everybody else. Chigoziem Okonkwo presents some interest, but he is essentially a heavy WR and will likely be covered by DBs, and it is unclear if he has a significant enough size advantage for that to work. Of the late round prospects, Armani Rogers, Lucas Krull and Rodney Williams present high enough TE YAS to warrant a flier.
This is where statistical modeling goes awry. There is just too much technique for physical traits to correlate to future production. Players like Slater (short arms) and Orlando Brown (terrible agility) overcome physical limitations fairly often. Gifted players like Andre Dillard and Ja’Wuan James often squander physical traits. I have moderate confidence in my OT YAS modeling, and Evan Neal did not complete enough of the combine drills to receive a score. Of those that completed drills, the highly thought-of prospects (Ekwonu and Cross) produced OT YAS that roughly correlate to borderline-starter. Trevor Penning and Abraham Lucas had the highest scores, with the third highest going to guard prospect Zion Johnson. He played spots at tackle (poorly) but has a high acumen, and I wonder if he could be the Dallas Cowboys “guard now, tackle later” prospect. However, that goes beyond our objective modeling. Of the remaining prospects, Matt Waletzko had a high score, and scores from Braxton Jones, Nicholas Petit-Frere, Luke Tenuta and Sean Rhyan also indicated that their traits outpace their college performance (and likely draft spot). (4/30/22 Edit) – Additional UFA that appear to have starter athletic traits include UFA Ryan Van Demark, who was signed by the colts, and former Longhorn Denzel Okafor)
If the OT modeling was shady, the OG modeling was a blind shot in the dark. I suspect this is because all OT provide some ”apples to apples” comparison traits because they all have to pass protect on an island. Whereas, OG performance is so tied to their scheme fit that ”apples to apples” comparisons of physical traits is near impossible. Zion Johnson scored high again, but the OG YAS for Kenyon Green was a 2.3, which was my first indication that the model is likely severely flawed. If we accept the results from the model, however, there is value to be found late in the draft with Logan Bruss, Zach Thomas and Eric Wilson.
Runningbacks appear to fit into two molds. Burst RBs have fast 10-yard splits and good lateral quickness (short shuttles). Power backs have good agility when multiplied by their weight (so, agility relative to size), and good long speed (40 time, or last 30 time). RBs that outpace their projected draft slots include Tyler Badie, Ty Chandler, Master Teague and Isaiah Pacheco.