It's been a bit since I've last posted anything in the workout series-trying to figure out whether or not players at certain positions performed better or worse as a result of combine workouts. With college pro days scheduled to take place over the next couple of weeks, it's important that we don't get infatuated with certain numbers that might turn out to be meaningless.
After looking at wide receivers and offensive linemen, I decided to take on defensive tackles. This process took much longer than the other processes, largely because the NFL just has more defensive tackles take significant snaps. It's not just that there is heavy rotation along lines (there is), but that the NFL is really bad at drafting defensive linemen.
Normally, I found that the relationship (defined by the correlation, which I described in small detail in the receiver article as a number between -1 and 1) between performance and draft pick to be extremely low for defensive tackles (at least at a particular position-more on that later).
Using the value of a draft pick derived from Chase Stuart at Football Perspective (formerly of pro-football-reference), it's pretty clear that a logarithmic model makes sense for determining whether or not a pick is a "hit" or not. This works with the draft value chart that Jimmy Johnson created (and nearly every website uses now) and is a good way of comparing picks.
From that perspective, it's pretty easy to compare draft picks to value. While the analysis of running draft picks against performance (however you choose to measure it for a position group) is wanting when it doesn't control for experience, aging and system, a clear relationship generally emerges.
While I generally found a correlation of -.35 or so from draft pick to performance at other positions (which would probably end up at -.5 when controlling for UDFAs, aging players and rookies), defensive tackles were found to have performed far below their draft status. In fact, one-gap tackles ended up with a correlation of nearly 0, with only a slight but insignificant bias for higher picks.
This should be particularly confusing, because draft picks correlated with snap counts at -.4. You would expect that players who get on the field perform well, but either defensive tackle is a really, really difficult position to have depth, or NFL evaluators are sort of poor at finding the right ones. Given that defensive tackle was my largest sample size, even after splitting them up by scheme, I doubt it's the first conclusion (one-gap defensive tackles beat out wide receivers by a hair).
Individual metrics are a little better for defensive tackles, but only one single individual rate metric exceeds .1 in magnitude of correlation-tackles, tackles for loss, run grades, overall grades, etc. That individual metric is total pressures per snap, and that comes in at a piddling -.24.
Generally my goal is to find a metric from the Combine that exceeds the correlation of draft picks (when treating them as having logarithmic value instead of absolute arithmetic value), and I usually fall a little behind after adjusting for UDFAs, age, etc. Here there lies no such problem.
I imagine I'm missing something, however. Different players are drafted for different roles, and I may not have completely controlled for that. I painstakingly went through every tackle and team they played for and separated them out by their primary role. I got rid of the 3-4 tackles and ends so that I could focus on the Vikings. Then I split the players who played in 4-3 systems into primarily 3-technique (outside the guard's shoulder, between a tackle and guard: Kevin Williams) and 1-technique (inside the guard's shoulder, between the guard and the center: Pat Williams) roles.
Within those constraints, NFL teams didn't do well. Presumably, not all schemes demand 3-techniques or 1-techniques to perform similar functions, but the results for NFL teams are not encouraging.
As always, I take a look at other research on the topic before I delve into my own. DraftMetrics takes priority again, because they've looked at everything. Instead of splitting players up by scheme, they took a look at defensive tackles overall, and found that there were no significant differences between Combine numbers and the number of overall starts a player had.
The most significant result, but not one that was particularly encouraging for them, was 10-yard splits. 58% of all Combine participants became 3-year starters, but 82% of those who ran a 10-yard split in 1.75 did the same. It sounds significant, but it's not much more significant than the general difference in athleticism that you find between drafted players and undrafted players.
After that, they found that the vertical jump and 20-yard shuttle were somewhat less significant, but still more significant than the average difference between starters and nonstarters.
Most interesting, however, were their red flags: none of the 23 participants they studied who ran the 40-yard dash in 5.31 second or slower became 3-year starters. None of the 24 participants who had less than 21 bench press reps became 3-year starters.
While I generally appreciate their work, I have to reject their claims here. First, because this is the first group of players who I found that had more snaps as a result of their draft position than their performance, and therefore should not be judged by their number of starts.
By that I mean that picks correlated with the number of snaps at -.5 (which would theoretically mean it explains 25% of their snaps), while grades per snap, tackles per snap, etc. all performed at a lower correlation. Again, pressures per snap did exceed draft slot in their ability to put players on the field, but because run performance did not, this implies that NFL teams are willing to generate pressure at the cost of giving up the run.
But there's an easy way to avoid this, apparently. Paying more attention to the 20-yard split, the 3-cone time, or even just the bench press would instantly improve the NFL's ability at finding players who can generate pressure without performing as poorly against the run.
I'm sure I'm wrong about something here, perhaps about something as fundamental as player grades, so I'll just move on to what I know for sure.
The first is that these don't make sense until you control for scheme and position. Under tackles/3-techniques have significantly different metrics than nose tackles/1-techniques. I eliminated all seasons played by players in a 3-4 system, and ended up with 83 seasons from draft picks picked between 2005-2010 that played the 3-technique and 78 such players who played 1-technique for the majority of their snaps.
For incoming three-technique players, the most important predictive workouts were the 20-yard split, the vertical jump, the three-cone drill and the bench press. Specifically, the 20-yard split was the single-best workout for predicting NFL success, which is admittedly odd.
This ended up being significant in the descriptive ("backwards-looking") model as well, so it seems that the 20-yard split is significant. At first, I chalked this up to Geno Atkins ruining all the results, for having an absurd split while also having the most pressures per snap (it wasn't close, either).
Removing Atkins from the sample then led to bench reps getting disproportionate weight in the formula. This is because Brodrick Bunkley has been a consistent leader in pressures and also has the highest bench rep total in the sample by far (44).
After controlling for outliers, it was fairly clear which variables mattered, and the 20-yard split consistently stood out as an important predictive workout for three-technique tackles.
Feel free to theorize as to why that's the case. Personally, I imagine part of its importance 10-yard burst is included in the 20-yard split, which is fairly important. Interestingly, if you take a look at what professional trainers have to say about the 40-yard dash, the first ten yards sound exactly like the skill set you need to be an excellent lineman:
The start of a 40 yard dash is first based on the athlete's explosive power to help get them from a static position out into the drive phase of the sprint. Many coaches today have their athletes start in a 3 point stance athlete stands with front foot 2-6 inches from line depending on the athletes size and back foot 2-4 inches from front foot with toes facing forward. The athletes front knee should be bent nearly at 90 degrees and back leg around 120 degrees with hips slightly above knees, back flat and chin tucked.
Once the athlete has left the static position the athlete is now in the acceleration or drive phase. Michael Gough (2006), defines the acceleration phase from the initial movement of ground contact until the athlete reaches top end speed. A powerful triple extension of the hip, knee, and ankle joints is important for maximum power development off the start. Forward body lean is critical during the acceleration phase with the shoulders always over the hips.
The focus on knee flexion, footwork and bursting out of a three point status all align with what defensive linemen have to do. Body control is paramount, and all of the important defensive linemen muscle groups are stressed.
The 20-yard split includes both the drive phase (as it isn't complete at 10 yards) and the transition phase as the player moves from focusing on explosion to maximal velocity:
Michael Young (2007) of the USA Military Academy and Human Performace Consulting explains there are three primary goals of maximal velocity sprinting: preservation of stability, minimizing braking forces and maximization of vertical propulsive forces. Preservation of stability is the body's ability to stay in perfect posture for the sprint because when stability is disrupted the loss of elasticity occurs. This stability relates to the athletes core for the most part, think of a squat an athlete holds their breath on the way down to support their back and keep their spine protected.
Stability, balance and maintaining a trained posture are critical for linemen as well and the transition from burst to balance is critically important to both 40-yard dash times and defensive line play.
Or at least that's my justification for the importance of the 20-yard split over the commonly held convention for the 10-yard split.
Overall, determining "how athletic" a defensive tackle is from their combine numbers comes down to this:
This does a very good job at predicting NFL success, more than the other formulas I've presented. The top scorer is also the best 3-technique tackle in the league, Geno Atkins (9.98 points). Other high scorers include Ndamukong Suh (7.5 points) and Brodrick Bunkley (7.98 points). The biggest miss was Clinton McDonald (8.67 points). The lowest scorers are almost unanimously forgettable. The bottom three are Anttaj Hawthorne (0 points), Frank Okam (.62 points) and Atiyyah Ellison (1.3 points). Still, there was one miss among the bottom scores, which was Brandon Mebane (1.4 points).
There is a way to reduce the formula to just predict pass-rushing or run-defending performance, although it is a bit weaker. For pass-rushers, the formula is:
There's a good run defense formula as well, expressed here:
The most surprising part of this formula was the use of the "Flying 20," which refers to the second half of the 40-yard dash. Simply subtract the total 40-yard dash time from the time at the 20-yard split, and you'll get the "Flying 20" measure that I initially thought was useless until I ran some numbers
It may be because the Flying 20 isn't just a measure of maximum speed, but smoothness in the transition and an ability to switch positions while maintaining balance. The ability to maintain that balance in unusual positions will often show up in the run game. That, and generally being able to move a big body around is good. The Flying 20 is at the very least indicative of that.
The fact that the vertical jump was important in both formulas speaks to how versatile the generic measure of leg explosion really is. NFL evaluators will talk about this to some degree, and many have famously used what's called the Explosion Number, which simply adds the number of bench press reps to the vertical jump (in inches) and the broad jump (in feet). Any number over 70 is supposed to be worth something.
I found that the Explosion Number itself did a surprisingly decent job at predicting pass-rushing ability, but the vertical jump alone is more predictive. Adding two other numbers without correcting for how important those other numbers are reduces the effectiveness of the Explosion Number to predict pass-rushing performance. In fact, by themselves, the shuttle, three-cone, 10-yard split and 20-yard splits were also more effective by themselves than the Explosion Number.
From there, the next task is to evaluate 1-technique linemen, or nose tackles for the 4-3 system. These players are often asked to plug up holes, take on blockers, redirect traffic and occasionally collapse the pocket. Without a premium placed on their pass-rushing ability, there aren't as many hard statistics to measure performance. Here, I've adjusted most things to Pro Football Focus' grades for play against the run and pass rushing in order to come to solid conclusions about my data.
Nose tackles are much harder to evaluate. While I don't have to worry about freaks like Geno Atkins or Brodrick Bunkley messing up my data, there's not a lot of clean indicators that point to performance. As I've often learned, simple is just sometimes better.
In general, however, there seemed to be a bit of a tradeoff when it came to comparing pass-rushing prowess to physical run-defending. Generally, it seems, you'll have to choose which you would rather specialize in when picking your nose tackle (if you're going to be influenced by Combine numbers).
For the most part, the better pass-rushing nose tackles didn't always last for very long. Folks like Adam Carriker (10 pass-rushing points) switched to a 3-4 defensive end, and others like Corey Peters (8.49 points) never made a big impact. It seems that for a 1-technique tackle, you would largely want to judge their run defense.
But if you wanted to know the pass-rushing formula for incoming nose tackles, here it is:
It's not a particularly useful way for evaluating nose tackles. Yes, the better scorers in this metric do generate more pressure (and even generate more tackles for loss), but also are more likely to miss tackles and get pushed out of plays.
More interesting, it looks like GMs are overpicking for these qualities, with a correlation of the pass-rushing score and the log of the pick (again, the best way to evaluate how GMs value a pick) at -0.5, which is highly significant. Finding a pass-rushing interior player and lining up between the guard and center is something that not only happens too often, but is often counterproductive.
An instructive example is Letroy Guion, who played 3-technique at Florida State and for a brief time with the Vikings, but then switched to nose tackle. These might not be players who were overpicked necessarily, just incorrectly used.
For incoming nose tackles, run defense can be characterized thusly:
Finding this equation was perhaps the most strenuous part of the process. It's clear that physical dominance is more important at 0-tech than maybe any other position, but the ways in which they express this dominance can vary. Some of the problem was that I was OK allowing nose tackles playing in a hybrid or 1-gap 3-4 system, but I had to get rid of them as they were obfuscating the result.
It should come as no surprise that the 40-yard dash doesn't matter, but the fact that the "Flying 20" is very important is fairly shocking once again. I've sort of explained my thoughts on that already, though.
Also, who would have thought there should be a mild penalty applied to weight? At first, I thought the negative correlation with weight had to do with some outliers who weighed a lot but did poorly, but after controlling for maximum and minimum weights, that turned out to be the case. In general it is good to keep weight down, and NFL teams should be wary of 1-technique tackles who weigh more than 320 pounds or below 305.
Of course, weight isn't all bad. When finding a total score for the grade, I found that there was a slight corrective to the small penalty assigned to weight—running short distances quickly with weight is good, but so is anchoring yourself.
Incidentally, the correlation of these qualities and generic draft pick slot was nearly 0.
The final score for incoming prospects on a 1-10 scale can be characterized with a more complex equation, although its significance is fairly easy to explain.
While I've already discussed how weight is both good and bad in the formula, the most shocking values might be the 40-yard dash and the height. I'm not sure how to explain the height, except to say it probably is more difficult to find players with good leverage and a natural ability to find the right pad level at the more extreme heights.
The 40-yard dash is not extremely important, but relevant enough to include. It only carries one-tenth the significance of the 10-yard split, and might be more useful as a proxy to determine work ethic. The highest graded 1-technique prospects between 2005 and 2010 were Jason Shirley (9.93—he was the guy the Bengals converted to guard in the 2009 season of Hard Knocks), Kyle Williams (9.92), Roy MIller (7.23) and Brandon Mebane (6.63).
It was much more fun working with the numbers of current 1-technique and 3-technique tackles, because my sample sizes tripled and the numbers made a lot more sense. Even though it didn't control for age or experience, the results are a lot more interesting and are probably a better jumping off point both for evaluating current NFL talent and incoming rookies.
For 3-techniques, I still think it's more important to focus on their pass rush. For that, the equation was relatively simple:
The single most important workout was the 20-yard split, where the raw number mattered much more than any correction for weight or height. The fact that this holds true across different position groupings tells me that this is a severely underlooked portion of defensive tackle performance.
After that, weight is the most important, although that is largely to correct for the penalties applied to weight when calculating burst for the 10-yard split and agility through the shuttle. The 10-yard split is next in importance and it seems like any prospect at a decent weight who pulls in a fast time in the 10-yard split and 20-yard split should be worth looking at.
The three worst players in this metric were Atiyyah Ellison (0.022), Sedrick Ellis (0.36) and Craig Terrill (0.54). Two of the three best are unsurprising: Geno Atkins (10.03) and Fletcher Cox (10.00) should surprise people, as those two can move with power. Tony Hargrove (8.68) is a bit surprising, but not discouraging for the system in the slightest.
For run defenders, the formula looks overcomplicated, but it has a theme that is good to go by:
Every speed score added new information to the formula, but the 40-yard dash was again weak compared to the other three. More importantly, the most relevant speed scores mattered most when put in context with weight. The best defensive ends by this formula were Alan Branch (10.03), Mike Martin (9.89) and Kevin Williams (9.57).
You'd think you could combine the run grade scores and the pass grade scores, but redundancy and noise strip the total formula of meaning. A more streamlined and accurate formula looks like this:
And yes, that is in fact a simpler formula. The takeaway from this one is that the 20-yard split is still the most important determinant. Best scorers were Fletcher Cox (9.99), Geno Atkins (9.72) , Mike Martin (9.08) and Kevin Williams (8.59). The worst were Craig Terrill (0.00), Spencer Johnson (1.00), Kendrick Clancy (1.41) and John Hughes (2.32).
Here's 1-tech run defense:
And 1-tech pass-rushing:
And the total grade:
The grades bore out how you expected. Desmond Bryant, who is the best pass-rushing nose tackle in the league right now, scored at the top of the pass-rushing grade overall. Behind him is fellow Raider John Henderson (maybe the Raiders have already done this sort of analysis?). Just below that are 1-tech flameouts that should have played in 3-4 systems (Alex Carrington, Adam Carriker) and familiar names like Alan Branch and Gerald McCoy.
For run defenders, the top physical performers were John Henderson, Desmond Bryant, Alan Branch, and those 3-4 flameouts (Adam Carriker and Alex Carrington).
The overall best performers were John Henderson (10.56), Alan Branch (9.23) , Desmond Bryant (8.99), Adam Carriker (8.62) and Frank Okam (7.04).
So what did this all tell us? Too much to process all at once, that's for sure. But a recap:
- You are looking for different physical attributes when comparing 5-technique, 3-technique, 1-technique and 0-technique players.
- Speed scores should be taken in context of a player's weight and height. Don't let people tell you the 40-yard dash isn't important for defensive linemen. It is, but only in context. Speed is stride length times stride frequency. How do you increase both? Power in the legs.
- Split those 40 times in half. It's surprisingly important. The "Flying 20"—time after the 20-yard split—is an indicator of balance, footwork, work ethic (long-term over a season and short-term in a single snap) and body control.
- For 4-3 tackles, the most important work out is right before that, at the 20-yard split. DraftMetrics has figured out that there's value in the dash, but only at the 10-yard split, which I found to be mildly predictive but significant.
- At the same time, don't get caught up in 40 hype. The 40 is more important than a lot of analysts say, but a lot less important than coverage of it implies.
- Get tall guys at 1-tech. Seriously. I'd say 6'3" and above. It's not a deal breaker, but it turns out to be somewhat significant. Guys who played below that height that did well in the last five years: Brandon Mebane, Kyle Williams, Jurrell Casey. That's it, and Brandon Mebane did much better when he switched to 3-technique with Rocky Bernard and Alan Branch playing the nose. Several more were average, but if you want a home run, the guy better look real good on film. Consider it a "red flag."
- Don't get caught up in bench numbers for 1-technique guys, unless it is lower than 21 bench reps. It's all in the legs, anyway.
- Organizations do a bad job using combine data. Shocker.
- There's still more to learn. I'll be more comfortable with two more years of data.
Some scores are in, so I can give you approximate results, but I couldn't find any 20-yard split times (which is a real shame). I guess at those, but the correlation of the 20-yard dash to 40 times and 10 times is not as strong as you would think (which I suppose is further validity of the measure as an independent test).
|Name||Pressure Score||Run Score||Total Score|
|Name||Pressure Score||Run Score||Total Score|
What does this mean? Well, we don't have scores for a ton of defensive linemen. Some haven't had pro days yet (Star Lotulelei) and others have refused to participate in the Pro Day (Shariff Floyd). I was down on Hankins a little bit after the combine, but no longer. He's my second-ranked 1-technique prospect once again.
The only warning signs for nose tackles were for Brandon Williams and Cory Grissom. I no longer want to pick Williams at the bottom of the second or top of the third like I previously did, largely due to these results. A negative score, or one near 0 is a red flag to me. This could also be an indication that he should stay at 3-technique, instead of switching to nose, as is being discussed. Only one nose tackle had a lower score than Williams and had a good career in the NFL: Brandon Mebane. Everybody else under two points has been completely underwhelming.
On the other hand, I would take T.J. Barnes in the seventh round instead of waiting until the free agency period to get him. Eye-popping numbers that compare to Alan Branch, and worth a look.
I quickly ran through the numbers after inputting average and favorable scores for the missing workouts of Shariff Floyd and Sheldon Richardson. Nothing stood out worth noting for 3-technique play. Bennie Logan is missing a 40-yard dash, so we'll have to wait on him. Based on an unofficial 5.08 40-yard dash and somewhat reasonable guesses for the 20-yard and 10-yard split, he graded out to a 2.3 as a pass-rusher, 7.55 as a run defender and 4.4 in his overall grade. Nothing worth deviating from your tape study.
At 3-technique, nothing stood out as a red flag. The player I'd be interested in the most is Nicholas Williams, who has very, very little experience in football. One year of high school football (four years basketball), and two years in college (and one year as a redshirt freshman). He scores highly, and his 8.01 makes him a UDFA target to me, worth probably the $10,000 signing bonus instead of the usual $5000.
Hopefully, these Pro Days are more revealing. What do you think?