I recently wrote a piece on how Jerick McKinnon gets unfairly set into a particular role because of his size—a trait that doesn’t even correlate to the role he supposedly can’t fill. McKinnon’s size falls well within the range of players who can fulfill every-down roles, and we can confidently say that players like 198-pound LeSean McCoy and 200-pound Jamaal Charles aren’t outliers; good running backs can range from 195 pounds to 245 pounds.
In that piece, I off-handedly mentioned that McKinnon doesn’t see a lot of third-and-short time, and that his yards-per-carry could be inflated by his expected role.
That seems particularly relevant to the Vikings not just because McKinnon’s efficiency numbers are a little different than what “normal usage” for a running back should be, but because Adrian Peterson’s are as well.
Peterson led the league in rushing last year and did so with a respectable 4.5 yards per carry, but it’s also well-known that the Vikings ran it more often on first and ten than any other team. On first and ten, it’s a little easier to get yards. The league average is 4.31 yards per carry in that situation, which is above the league average on non-first-and-ten runs, 3.94 yards per carry.
And if we do this for Peterson, we may as well do it for all running backs that we have relevant data for.
For this, I looked at the last four years—2012 to 2015—because I think that with how quickly the league is changing, it’s difficult to go too far back and use that data while projecting it onto the current NFL.
Here are the yards a player rushing the ball should expect in the following situations:
There’s not a ton of runs in other situations, so I didn’t use data from other down-and-distances in this exercise. I did a little bit of smoothing for the data in order to create more representative averages—players ran further on 2nd-and-six than they did on 2nd-and-seven in the data, but there were a few individual runs that skewed the data.
By subtracting the actual yards gained in each situation from the yards an average running back did in the same situation, we can get a “yards over expected” in each situation. By summing up all of the additional yardage, we can get a total yards over expected. Here are the leading yards-over-expected in the data I selected among the running backs with the most attempts:
|Player||Yards over Expected|
lol Trent Richardson
Adrian Peterson has been pretty good! The fact that he and Jamaal Charles lead the pack in the 2012-2015 data is pretty astounding given that Charles missed 12 games in that span and Peterson missed 17.
Because this is cumulative, it rewards backs with more carries. We can also divide by attempts and add that number to the league average (4.21 yards per carry) to get an approximate “true average” that provides a number that looks familiar to us but with the context of the above process:
With Peterson dropping to players like Justin Forsett and C.J. Spiller, this process may seem flawed, but this exercise isn’t to find the best running back, it’s to find the runner who was best at what he was asked to do. It is probably the case that Forsett and Spiller were asked to do a particular subset of things that they were excellent at and didn’t have to do things they were not particularly good at, while Charles and Peterson had to perform in every situation.
Also, remember that Forsett has averaged 5.1 yards per carry in the last four years. Weird, right?
So how do we judge a running back’s every-down-ness? I decided to look at how often their usage in those 16 down and distance situations matched their peers. For example, the average running back in the set had 59.4% of their runs come on first and ten, but Mark Ingram had 65.3% of his runs come in that situation—two standard deviations more than the average back.
I found the variance for each back and created a scale to see which backs were the most “every-down” backs out of the ones in the set. This is what we have, with the lowest numbers representing the most typical usage pattern for a back:
Spiller and Forsett weren’t quite Ingram or Giovani Bernard, but they weren’t really used as a typical bellcow like Peterson, Charles and Demarco Murray.
I also value success rate as a metric, which is simply how often a running back is successful when running the ball.
It turns out that running correlates with winning to a much higher degree when we look at it using Success Rate (SR).
SR is a very simple concept in principle and has been around for decades. Each play is graded as either a success or not based on its outcome. For example, if a play gains 3 yards on 3rd and 2, that would be a success. But those same 3 yards would be a failure if the situation were 3rd and 4. In the seminal book from the 1980s Hidden Game of Football, the authors devised a simple rule of thumb based on their intuitive sense of football success. A success would be: On 1st down–a gain of 4 or more yards; on 2nd down–a gain that at least halved the distance to go; and on 3rd down–a conversion for a new set of downs.
I’d like to combine that along with yards above expectation and the “typicality” of a back to create a final measurement that I believe is a much more accurate gauge of a running back’s ability to be a bellcow. I think success rate/consistency is the most important feature of a running back, but it is to some extent priced into the “yards above expectation” metric, so it won’t be weighted more—success rate, Typicality and yards above expectation over average can be weighted equally when looking at a player’s “Bellcow Yards.”
I turned those scores, which are basically unit-less numbers, back into familiar yards-per-carry numbers to give an equivalent understanding of how “good” a running back has been at being a bellcow given his actual usage, consistency and ability to generate yards over the average back.
One thing to keep in mind is that these rankings gauge a running back’s ability to be a “typical” bellcow back, so they are not talent rankings (yards over expectation is a better representative of that).
Still, I didn’t really expect these rankings; I assumed that Le’Veon Bell would be higher and that Stevan Ridley would be lower. That said, if all made-up statistics (this one has not been tested for any relationship to team or individual success) matched our intuitive expectations, there would not be much use for them.
I have thought that Ridley has been underrated but I also think that about Doug Martin, who ranks low. DeMarco Murray’s body of work over the past four years does not lead me to believe it has been better than Peterson’s, either—but that could be recency bias. His 2012-2014 were fantastic (4.7 YPC) and could overshadow his dismal 2015.
Perhaps that first list of total yards over expectation is more useful, but I definitely think this kind of analysis is interesting if not meaningful. It tells us that while Trent Richardson may not be as good as Bernard, McFadden or Rice, he is slightly more “fit” for a bellcow role than the three of them because he is somewhat more consistent and produced his (awful) yardage in more typical running back situations.
I also think it underscores the fact that while Peterson was an enormous asset in 2012, his work in 2013 and 2015 has been a little overrated as he has consistently been put into better positions to succeed—which is not to say he hasn’t played well (he’s ranked third in the metric, so clearly he has) but that we tend to overstate his impact.
If we look at 2012 numbers alone, Peterson ranks first by quite some margin.
Of these four metrics (Yards over Expectation, Yards over Expectation per Attempt, Typicality and the overall Bellcow Fitness Metric), I am most suspicious of the usefulness of Typicality as it mostly provides some context to Yards over Expectation and isn’t a value judgment in itself. That means I’m also a little hazy on the usefulness of the fitness tips I’ve been receiving, but the final metric but I do like the first two metrics quite a bit.
Maybe pure yards-per-carry is better than the Bellcow Fitness Metric above, and I’m certainly open to the argument that YPC is better at identifying every-down backs than the quick work above, but I suspect the above metric is a slightly better organizing principle.
I did this mostly to get context for what McKinnon’s numbers should look like given his last two years, so here it is:
|Player||Yards over Expected||YExp/Att||Typicality||Bellcow YPC|
|Jerick McKinnon||104.18 (19th)||4.94 (6th)||185.69 (37th)||4.22 (18th)|
The parenthesis indicate how McKinnon would have scored compared to the backs featured up there. It looks like he would be between Arian Foster and C.J. Spiller/DeAngelo Williams in Bellcow YPC, which isn’t bad considering that his Typicality score is awful; the Vikings rarely used him in short-yardage situations and hesitated to use him in intermediate situations.
He had 65.7% of his evaluable runs come on first-and-ten, the most of any of those running backs up there—”beating” Mark Ingram by 0.4 percentage points. There were only three down-and-distances he ran below expectation: 2nd and 9, 2nd and 7 and 2nd and 4. On the nine other situations he ran in, he exceeded the expected yardage.
If he ran the same average yards over expectation in the same situations, but ran in those situations exactly as often as the average bellcow back, he’d have 131.6 total yards over expectation and an adjusted yards-per-carry of 5.13 yards—good for third on the list.
What this tells me is the following: Jerick McKinnon has so far proven to be excellent at what the Vikings asked him to do, outperforming even the more generous statistical thresholds his relatively light assignments gave him. Though running backs who run more often on first and ten should have a higher yards-per-carry, his stellar performance exceeded even that allowance.
But the Vikings haven’t tested him very often in many other situations. In the few occasions he’s had, he’s been exemplary but we don’t have enough evidence to say, statistically, if he can perform the function of an every-down back.
The risk-averse approach above suggests he would be an above-average bellcow back—ranking 18th in the above group is still above average because some of the backs above him have retired or will retire soon, and there are a number of bellcow backs who did not qualify for the list and will not because they weren’t good enough for teams to continue giving them carries.
Instead of giving penalties for having an atypical split, like we do with the Typicality Index and Bellcow Score above, we can just re-weight all running back carries so they match the normal workload of a bellcow back.
If we projected all running backs to have the same carry split as the average bellcow, Jerick McKinnon would crush:
If McKinnon does what he’s done so far, he has the talent to be borderline elite. The Vikings have every obligation to themselves to put McKinnon into as many diverse situations as possible—set him up to fail, even—to see if he can be their long-term solution. Simply put: