Last week, I discussed the importance of incorporating production metrics when scouting players. The biggest reason mentioned — aside from the predictive history for other positions in being able to determine good prospects — was that it allowed evaluators to break free from traditional means of winning opportunities and rewarding those who find ways to perform.
Another reason to incorporate data like production — and later, measured athleticism from events like the combine — is that it allows us to check biases, conscious or unconscious. If we’re biased for players with long hair, or schemes that free up offensive linemen for linebackers to block or linebackers dropping to the deep middle or whatever, we can implement some production metrics to re-center our evaluation.
While I have no evidence that offensive line production, at the moment, is definitively predictive of future production, there’s good evidence that college linebacker production can help us determine NFL success — evidence that we went over last year.
The production and athleticism scores in that piece hold up well, but adding age into the equation made it (marginally) more predictive than draft position for the first four rounds of the draft and significantly more predictive than the previous model.
Of the 43 linebackers who moved a tier as a result of adding age to the production and athleticism metrics, 26 moved in a direction that matched their success and 16 moved away.
Put another way, more players who moved from “high value” to “middle value” and “middle value” to “low value” were correctly identified as failures rather than incorrectly. The same is true when players moved up a category as successes.
Notably, the previous model saw some issue with low-scoring first-round picks incorrectly identified as big risks when they in fact were huge successes. This model, just by adding age, was able to move players like C.J. Mosley, Dont’a Hightower and Jon Beason up a value category while moving players like Bobby Carpenter and Dontay Moch down.
We don’t have athleticism data yet, which makes all these scores preliminary until the combine, but for now we can determine which linebackers are worth the most interest going into the combine.
Starting with the linebackers designed as “inside linebackers” by CBS, we can look at production — defined as solo tackle market share (percent of team’s solo tackles) and tackle-for-loss market share — after adjusting for highly-ranked teammates.
Essentially, if a player is a first-round caliber player (as defined by CBS), like Jonathan Allen at Alabama or Teez Tabor at Florida, that player will “give” credit for a third of his tackles to his teammates. If he’s a second-round caliber player, he “gives” credit for a fourth of his tackles to his teammates and so on.
That way, the fact that Reuben Foster played with Jonathan Allen, Tim Williams and Ryan Anderson won’t punish him like it would if we used true market share statistics. Market share teammate correction is what allows us to properly account for receivers Odell Beckham and Jarvis Landry coming out, who otherwise both would have had disqualifying market share.
Foster has 60 solo tackles, but he also gets credited for an additional 39.5 tackles because of those Alabama players mentioned earlier as well as Dalvin Tomlinson, Marlon Humphrey and Eddie Jackson. His 12.2 percent market share of tackles gets turned into 20.3 percent this way.
That puts him on more equal footing with players like Elijah Lee and Anthony Walker, who don’t have as many talented teammates taking tackles away.
Tackles as a measure of quality are rightly controversial. I’ve long argued that total tackles are an awful indicator of talent for linebackers. Given that the single-season leaders in solo tackles include such non-luminaries as Jordon Dizon, Tyler Matakevich, Chris Chamberlain, Orie Lemon, Taylor Reed and Scooby Wright III, there’s reasonable skepticism that using tackles as a statistic means much.
But the college environment is different, and avoiding total tackles in favor of tackle market share drives at the core of the reason why: most college players are not NFL quality players, so NFL-quality linebackers will end up accumulating the lion’s share of their teams tackle production. Not only that, they should be responsible for tackles-for-loss, another market share statistic that goes into determining the quality of linebacker play.
It’s related to a similar, but not identical distinction: good linebackers accumulate tackles; not all tackle accumulators are good linebackers. In college, that distinction tends to fall away, though a market share approach does a good job cleaning up a lot of the linebackers who fall through the cracks.
Our previous look into linebackers (linked here again) demonstrates the value of production metrics like tackle share for predicting NFL success. We’re only looking at production that centers around the run game (largely; as most of the linebacker tackles-for-loss and tackles in general occur against the run) and we’ll explore what production metrics that include coverage may look like.
Since then, I’ve added in a small correction for missed tackle rate, where a high proportion of missed tackles can impact the final score about half as much as a very poor tackle market share does.
As a reminder: these metrics do not compare linebackers directly to each other; they look at how likely it is a linebacker is worth more or less than their film grade. A linebacker given a fourth-round grade (Player A) with a higher production score than a linebacker with a first-round grade (Player B) isn’t necessarily better than the first-round linebacker — merely that Player A is more likely to outperform his fourth-round grade than Player B is likely to outperform his first-round grade.
First, let’s look at what CBS categorizes as inside linebackers. The ranks are taken from CBS’ ranks two weeks ago, so their official ranks have likely changed by now.
While the offensive line class was fairly old, the linebacker class is fairly young. As a result, we have more than a few green lights among the inside (and outside) linebackers, with only one inside linebacker who has a red flag.
The headliner of the class, Reuben Foster, is generally a good prospect. At seventh overall and with an average score, he is neither overvalued nor undervalued when accounting for his productiveness. His age is functionally average for a prospect in the top-150 and if his film really grades that well his production points to a high-quality player.
Potentially undervalued, however, is Raekwon McMillan from Ohio State. He’s not regarded as a particularly athletic player, but his production was fine enough. His production rating is more than “fine” here, but that’s a result of his being so young — born in November 1996, if Wikipedia is to be believed. Average production at an incredibly youthful age is precisely what Mike Evans, Sammy Watkins and Stefon Diggs put together.
Should McMillian test poorly, his final linebacker score will suffer but his age will keep him in the upper ranks. People tend to assume that bodies peak around ages 18-21, but athletes tend to peak closer to age 28.
On the flip side of that is Jarrad Davis from Florida. Despite an age that doesn’t hurt him too much as well as grabbing boosts in his tackle total from Quincy Wilson, Tabor and Alex Anzalone, Davis just has a poor score. He also should get a boost from Anzalone missing significant time (again) this year, but simply didn’t get to the ballcarrier that much.
If he was as young as McMillan, he’d have a somewhat above-average score. But he’s not and his development curve seems to be fairly disappointing.
After Davis are two linebackers who have noteworthy scores well above average, though not quite the score McMillan earned. Scores above 115 are still worth consideration for boosting value, and both Anthony Walker Jr. and Kendell Beckwith are interesting cases.
Beckwith is intriguing to me because he has a high tackle share despite missing two games and he’s fairly young for a linebacker graded in the third or fourth round. Walker is intriguing to me because I hadn’t heard much about him before the draft process but he evidently had an astounding 2015 after a rough freshman year showing up with 40 pounds too much muscle without flexibility or movement. If his 2015 numbers are used, he’d have a rating closer to 120 than 115, and that’s fascinating by itself… though his missed tackles from the last two years are a big, big concern. His youth is his best asset.
I don’t run FCS or lower data for a few reasons, but if I did, Connor Harris would come out very poorly. His tackle share, despite teammates who do not have draft consideration, was below average. Beyond that, he’s quite old for a prospect. He didn’t look great to me at the Senior Bowl (aside from a nice pick-six in one practice) and I’d wager he goes later than round four.
Now, outside linebackers:
There’s a lot to like about this outside linebacker class. Two of the included linebackers, Haason Reddick and Tyus Bowser, were edge players in college and therefore have to build their resumes off of an excellent tackle-for-loss showing rather than in solo tackles. You’d think this would poison the sample, but it works just as well, historically.
Von Miller, Anthony Barr and Jamie Collins were edge players who played off-ball linebacker in the NFL and their TFL market share was so overwhelming that it overcame the disadvantage they had generating solo tackles. They averaged a tackle-for-loss market share of 27.8 and generally speaking, you’d want to see a TFL market share above 25.
Reddick has a TFL market share of only 21.8 percent, but he actually has a high solo tackle market share for his position, and led his team in solo tackles despite playing on the line of scrimmage. That allows him to have an average production score, meaning the late first-round plaudits he’s earned from the draft community might actually be appropriate.
Bowser on the other hand, had very poor TFL numbers… though that comes with a big caveat. He was playing with perhaps the single best defensive player in college football next to him, Ed Oliver — who Pro Football Focus regarded as the top defensive player to be draft-eligible next year.
One might give leeway to Bowser on these numbers if they assumed that Oliver would be a first-round pick this year. I think it’s likely; in which case Bowser would be credited with 7.3 of Oliver’s astounding 22 tackles for loss. In that circumstance, Bowser actually ends up with a production rating very close to 100.
Generally speaking, I would bias myself in favor of film grades when there’s uncertainty, and if an NFL team judges him to be capable of off-ball work, I wouldn’t let his production change his grade too much—except in the case of a close tiebreaker.
Still, if one isn’t certain about Oliver’s talent level (say, if he were a second or third-round player), they should downgrade Bowser.
Zach Cunningham doesn’t seem to need any adjustment and his mid-first round grade as a linebacker seems well-earned. He evidently “misses too many tackles” according to one NFL scout, but even accounting for that, the methodology is perfectly OK with him. He has some, not many, missed tackles (per PFF) and when grading against how often he attempts a tackle, he misses only slightly more often than average.
Kendell Beckwith’s partner, Duke Riley was one of the beneficiaries of Beckwith’s absence and racked up some tackles in his place. He didn’t grab so many extra tackles that one would be wary of the statistic here, and his tackle production could counter the consistent theme in his scouting reports about getting swallowed up by offensive linemen.
He has the highest solo tackle market share of anyone in the draft class and an average missed tackle rate. To go along with that, he’s somewhat young.
One of my favorites at the Senior Bowl, Alex Anzalone, fared poorly. Even after pro-rating his eight healthy games to Florida’s full 13-game slate, he ends up with a production rating below 90. He has enormous injury flags and was excellent on film when healthy. He might remind Vikings fans of Michael Mauti, though their scouting reports are very different despite both being very talented.
That said, Anzalone’s evidence talent didn’t show up enough in the form of production even in the games he played despite teammate corrections. I wouldn’t be surprised if he’s been overrated by the draft community (again, I liked what I saw) and he falls further in the draft.
Finally, Elijah Lee benefits more from age than any other “outside linebacker” in the group; his tackle-for-loss production is iffy but his solo tackle market share is high enough to overcome it and his age pulls him past a production rating of 100. His general tackling numbers are below average for a draft prospect but players that young get more leeway.
Potentially interesting is the application of coverage statistics to the process. Unfortunately, coverage statistics are different from site to site and (unlike pressure statistics for offensive linemen) are divergent enough between services that using two services to fill the gaps produces a conflicting picture.
With that large pinch of salt in mind, the two groups really mellow out when adding coverage to the equation. Not knowing which coverage statistics project well to the next level, I used two specific ones: ball hawk rate and adjusted yards allowed per snap in coverage.
Ball hawk rate measures how often a player gets their hands on the ball (pass breakup or interception) when targeted, while adjusted yards per coverage snap measures how many yards, interceptions, first downs and touchdowns they cause or allow in their coverage. One measures ability when targeted and the other measures production in and out of targets; that way one can reward a player for deterring targets.
Most linebackers either don’t change their score with this addition or get closer to 100, with a few exceptions. Here are their scores:
The biggest mover is Davis, who jets up from fairly bad to average. After that, Anzalone makes a big move up, but probably not enough to save his evaluation, while Riley moves all the way up to rare territory.
Walker and Reddick make small moves up while Beckwith barely moves. On the way down by inconsequential amounts were Foster, McMillan and Lee. Bowser and Cunningham make the biggest moves down, but Bowser’s move from below-average to bad is an issue, while Cunningham’s move from very slightly above average to very slightly below average is less notable.
I would trust the run grades far more than the combined grades, but those combined grades do give us an interesting look.
This should give us an interesting field of linebackers to watch as we head into the combine; if Anzalone and Reddick test as athletically as reputed, then they’ll be big hits in the final scores. McMillan and Beckwith may lose their luster, too.
We’ll move down the list of positions before the combine to see which players provided the most meaningful production for their position.