The Vikings have a well-known need at linebacker, though they may not want to invest a high pick there with the rising importance of the nickel defense. With that in mind, the Vikings will have to find a linebacker that may not be the slam-dunk you tend to find in the first round. To that end, the Vikings could improve their odds with the smart use of data.
But what data to use? There’s a lot of information out there, and most of it is useless. Sometimes information that seems useless is useful if applied properly and a lot of our old understandings of which statistics matter may conversely not stand up to scrutiny.
Conventional wisdom indicates that speed and agility scores at the combine are critical for linebackers, but thorough testing of the athletic data can reveal that every drill at the combine matters for off-ball defenders. Beyond that, all of the scores mean nothing without accounting for the weight and height of the linebackers you’re measuring.
Because of that, it can be quite complicated to come up with holistic scores using a simple process that really accurately weighs the importance of various measureables. That might be a reason that NFL teams have been doing a poor job of incorporating athletic testing into their evaluation of linebackers.
Couple that with a poor understanding of which college statistics can predict future success in the NFL and you can end up with a distrust of using data overall that you would not have had if someone presented it appropriately.
I’ve combined the athletic measureables work I’ve been doing for the past several years with new measures of productivity in the college game to see which linebackers were most likely to bust and which were most likely to pay back their teams’ investment. Before I list the winners and losers in this year’s draft class, take a look at the linebacker scores and results by round in the following slideshow:
As you can see, prospects are separated into three bins: scores of 8.0-10.0, 6.0-7.9 and 0.0-5.9
Prospects in the first group met their round expectation far more often than players in the third group (for the first round, players need to be headliners; for the second round they need to be consistent starters and for the third round they need to stick around the league and teams do not actively try to replace them).
The determinations used above are quick judgments, but the results bear out if using Pro-Football-Reference’s approximate value or Pro Football Focus’ player grades. Jon Beason and Dont’a Hightower may deserve some quibbles, but chances are if you ask most people, one of them has played like a second-round linebacker (still very good) and the other like a first and you may get different responses from different analysts. Taken as a whole, here are the relative hit rates for the different groups:
(*excluded are 2015 rookies, players without reliable workout or college statistical data, as well as David Pollack and Odell Thurman because they busted massively for off-field reasons—both would have been marked as mid-round booms otherwise)
90 players isn’t a definitive sample size, but it’s certainly compelling. The work does bear out into the fourth round, as well, though the hits and busts are harder to define and the number of players with a high score drops significantly.
Another way to look at it is to look at the results by the “expected value” of a draft pick, using PFR’s Approximate Value metric; one they developed that roughly measures a player’s contribution level based on solo tackles, assisted tackles, interceptions, pass deflections, sacks, forced fumbles, scores, team defensive performance, teammate performance and postseason honors.
It’s an imperfect metric, but for the most part does a very good job giving quick numbers to run over large samples of data. As expected, the expected Approximate Value of a player decreases as you go into later and later picks. By figuring out what the average player returns to a team for each pick, you can see if a specific player has beaten their expected value.
Below are two tables that show the average performance above expected value for a player’s pick, sorted by linebacker score. The first table shows every linebacker with the score above a particular number and the second one shows every linebacker with a score below a particular number. In either case, it’s pretty clear: higher linebacker scores are good, and teams should try to grab players with scores of eight and above.
You can also tell from the general success rate of each score category:
With all that in mind, and most of the linebacker scores from pro days in, we can see which players are smart investments for the round value they’ve been given by CBS. Potential off-ball linebackers below, sorted by their player score:
[Updated to include all the linebackers projected to be drafted by CBS – if others have asked me about a specific linebacker, they are included here. I previously indicated that B.J. Goodson’s score was 3.3 but it is in fact 7.7 because of an input error]
A few caveats, however: Jaylon Smith and Myles Jack have estimated workouts scores used, and the players from Ohio State and Alabama may have inflated scores because of a correction designed to take into account teammate draft position. With so many Ohio State and Alabama players potentially drafted (and early), it could be overcorrected.
For now, however, this shows clear warnings for Kentrell Brothers (athleticism reasons), Su’a Cravens (athleticism reasons), Dominique Alexander (athleticism reasons) and Kyler Fackrell (age reasons). Reggie Ragland had a production red flag, but teammate corrections took care of it. Still, it would be prudent to note that for a team attempting to determine whether or not to put Ragland in the first round.
Remember, the scores are supposed to alter expectations in a given round, so it’s not saying that Joe Schobert is better than Leonard Floyd, only that Schobert is more likely to outperform his pick value than Floyd is.