clock menu more-arrow no yes mobile

Filed under:

Remarkably agile; analytics and the 2023 defensive back 3-cone drill

Yes, this is about math

Clemson v Notre Dame Photo by Michael Reaves/Getty Images

If you’ve been following my recent exploits you’ve likely already seen the latest Bills Mathia on standard deviation here, or the follow-up article on wide receivers and speed here. And if you haven’t seen either of those things yet, go check them out. It’ll bring some context to this dive into the 3-cone-drill performance for defensive backs at the 2023 NFL Scouting Combine.

Why this group? The Buffalo Bills could likely use another defensive back or two and this isn’t a bad drill to find a desirable trait. Or in other words, we’re about to dive into the last decade of data and see who has remarkable agility and who doesn’t, as we set our sights on the 2023 NFL Draft.


The wide receiver article linked above is a good read because I’m about to short-change this one. I took the last ten years of combine performances for the 3-cone drill, widely associated with change of direction and “agility” for our purposes. Specifically, only defensive backs and their times from the combine were used (no other positions or pro days, etc).

I’m not about to teach standard deviation for a third time in a week, but I calculated the mean and standard deviation for this group and we’ll use that to define who’s remarkable at this drill.

The Fun Stuff

Very brief recap and chart

Here’s the bell curve crappy drawing for this one followed by a brief recap of concept.

The average time for a defensive back in the 3-cone drill was 6.97 seconds over the last decade. The standard deviation for this sample was 0.204 seconds. The vast majority of players fall between 6.77 seconds and 7.18 seconds — aka between -1 and +1 from the mean. From a statistical point of view regarding this characteristic only, that means there are a ton of guys in this range — which makes them “replacement level.”

That’s a good segue into what this is all about. Between -1 and +1, the stat point of view would say: “Look for other reasons to like or dislike a player. This performance doesn’t stand out.” Another way to look at it is that numbers between -1 and +2 take average from a single point (6.97 seconds) and turn it into a range (between 6.77 and 7.18 seconds).

Between 1 and 2 standard deviations from the mean in either direction, there’s a possibility that you have something significant going on but it’s not definite. You might bump players between +1 and +2 down the draft board based on this trait, but if they have other great qualities this shouldn’t be a prohibitive performance. From -1 to -2 it’s the same deal, but in reverse. You wouldn’t draft a guy thinking he has some significant trait, but it might be enough to cover up a wart or two.

Anything greater than two standard deviations from the mean is definitively significant. That’s where we begin our real journey.

Fun Facts

The question we want to end on is if there are any players remarkable enough at this drill where you might be tempted to take a gamble on them from this trait alone. We’ll get there. Before we do that, some fun facts.

  • Pro Football Reference had 317 defensive backs with recorded times for this drill in the last decade (including 2023).
  • There were 9 players (2.5%) in that time frame at +2 or greater standard deviations from the mean. This is the group that’s on the wrong end of things.
  • One player was more than THREE standard deviations above, which is a truly remarkable performance (but again, not in a good way).
  • Despite that, four of the players were drafted including the guy at +3 (none higher than the fifth round). I point this out because I don’t want to suggest players are undraftable with these numbers, only that you better really like some other things about their game.
  • For the better performers in this drill, we also have nine players at -2 or greater away.
  • We also have Jordan Thomas with the NFL combine record in 2018 with a 6.28 time. That’s three standard deviations away from the mean, which is insane.
  • Five players in this group were drafted (none lower than the fifth round). Thomas was not one of them after having a blah season his senior year and some other concerns.

What about this year?

Finally, the thing we really want to know. Does any player this year stand out when it comes to change-of-direction measurements from the 3-cone drill? Just like the wide receiver group with the 40 times, the answer is...


Julius Brents out of Kansas State was pretty close to the -2 cutoff with a time of 6.63 seconds. Daniel Scott was the next-best defensive back this year and was barely below the -1 cutoff at 6.75 seconds.

A huge part of these results is incredible variability in who actually does this drill. Only seven defensive backs took part in this based on information I could find. Compare that to the 2021 combine where 77 defensive backs recorded a 3-cone time.

In summary

I chose the 3-cone drill for this position group because I’ve seen during film review how important change of direction can be, and of the drills this is my favorite to get some idea of that. I’d hear arguments for the shuttle drill too, for the record. For this exercise, I like the result.

With the 40 times for wide receivers, I think the potential is there to draft a player solely on that measurement and take that gamble with plenty of flaws aside. For this grouping, I don’t see that as a possibility. Inconsistent information/participation and a couple other flaws relegate this drill to valuable information, but nothing more than a piece of the puzzle.

And that’s why I like this result so much. I spent over 1,000 words on this and covered only a single measure that really didn’t create any great clarity on the position group. Evaluations are incredibly complex and need to factor in a vast amount of information. I like to remind myself of this at times. Even when you find a statistical anomaly in a preferred measure (hello Jordan Thomas) you don’t want to lose sight of the rest of the data.