Running Backs Sometimes Matter

I’ve written about this topic before, but a recent thread in my Twitter feed prompted more thought about it. When someone purports to demonstrate, via basic stats, that intelligent, experienced, well-incentivized market participants are wrong, I’m usually skeptical.

Of course, coaches like Bill Belichick and Andy Reid might have been off base in thinking it permissible to use first-round picks on running backs the last half-decade — they were certainly wrong about the particular backs on whom they used those picks (Sony Michel and Clyde Edwards-Helaire, respectively!) But in my view it’s unlikely great coaches simply “don’t get the math” or fundamentally misunderstand the game in way that seems so obvious to an RBDM advocate with a spread sheet.

So let’s take the thread, claim by claim and examine whether that’s the case, and, if so, try to identify what’s missing. (I’ve done this kind of exercise once before for stats-based claims that struck me as interesting, but ultimately unpersuasive.)

I’ll also add (and this should go without saying, but it’s unfortunately necessary in the current social media environment): This is not an attempted “dunk” on anyone in particular making these arguments. We should be unsparing toward arguments, but respectful towards those making them. The idea is to find out what’s true, and to that end have a healthy back and forth. The idea is not to disparage or insult people with whom we disagree.

Tweet 1

The article argues NFL teams run too often, and that running, while necessary, is on average counterproductive, compared to passing. That’s fine, and there’s neither anything objectionable in the article, nor any reason to leap to the conclusion from it that running backs don’t matter.

Tweet 2

This article uses an anecdote about Belichick wanting Thurman Thomas to rush for 100 yards against the Giants in the Super Bowl (he did, and the underdog Giants won) as its jumping off point. As one of the commenters pointed out, it omits the Giants own run/pass split in that game:

Belichick’s defensive gameplan might have had more to do with the specific matchups, the element of surprise and the desire to shorten the game (more volatility against a seven-point favorite) than the implication that the running game, or specific running backs, is unimportant.

But let’s leave that aside, and get to the meat of the article wherein Feng shows that passing efficiency correlates much more closely with winning than rushing efficiency does.

A commenter offers the following counterpoint:

In other words, teams with leads not only run more, but they run predictably in an attempt to kill clock rather than score points. Predictable runs are typically less efficient than unpredictable ones, as defenses will have more personnel in the box geared to stop them.

Feng responded to the comment by saying he’ll look more deeply into it, and he might have at some point (this was in 2014), but I am only going by what’s linked in the thread as the purported QED as to running backs not mattering

So through two tweets, we see that passing is *more* important (though running acknowledged as necessary) and that *average* passing efficiency (as measured by yards per play) is far more correlated with winning than *average* rushing efficiency, but that rushing to kill clock might be confounding our measurements of true rushing efficiency.

Interesting points, and useful in establishing what I haven’t heard anyone deny in the last decade-plus — that passing is *more* important. But for running backs not to matter, the thread must show one of two things: either that rushing efficiency is unimportant to winning, not just on average, but even in cases when the team is built around running, or that to the extent (or on the occasions) it is important, who the particular running back is makes little to no difference toward that efficiency.

Tweet 3

This article is about how the 2018 Rams’ uncanny rushing efficiency was in large part due neither to Pro Bowl running back Todd Gurley, nor the Rams’ vaunted offensive line, but Sean McVay’s heavy use of “11 personnel” — three WR sets — that necessitated light boxes. It argues the number of defenders in the box explain rushing efficiency variance almost entirely.

But even if box-weight were the primary factor in predicting YPC, that doesn’t advance the argument that running backs don’t matter. In fact, if anything, it cuts the other way! To the extent a running back is good and featured, you’d expect defenses to stack the box against him more often, resulting in less efficiency for that running back than he’d otherwise have, given an average number of box defenders. The best running backs should in that case get punished, YPC-wise, for being good.

That also means some of a good running back’s value would be that by necessitating more box-defenders, he’s freeing up his pass-catching teammates. Steph Curry doesn’t have to take a shot to have an impact on the offense — just the fact he’s on the court makes the defense extend itself farther, opening up space underneath the basket. If one running back is getting defenses to fill the box more often than another, he’s providing a similar (though not as valuable as Curry’s) hard-to-quantify benefit to the offense.

While that kind of intangible might be difficult to measure, one would at least want to look at YPC by box-weight if what Hermsmeyer alleges is true. If the best backs exceed expected production per box-weight, they would still be difference makers even if box weight itself were more predictive of YPC generally.

Tweet 4

This article is behind a paywall, so I can’t really dive into it, but it seems to argue that play design is more important to rushing efficiency than the player carrying the ball. Even if that were true — and I have no reason to doubt it — it has the same problem as the box-weight assertion, namely that if a back were outperforming the expected gains per a given play design he would still be adding value even if play design itself was the biggest factor in determining the baseline expectation for that play.

And of course teams have to vary their running concepts to keep the defense off balance, so you can’t just run the most-efficiently designed play every time. Performance compared to baseline per run concept would be a starting point to measure efficiency in this case, just as performance by box-weight would be one for the last point. And which is it, box weight or play design anyway?

Maybe game-context-adjusted rushing efficiency per box-weight/play-design is where we’re headed, and you can start to see how complex it’s getting and how maybe Belichick or Reid might have an intuitive sense of something that’s not easy to quantify by these metrics.)

Tweet 5

This article has eight parts: (1) Quoting the Feng article from 2014; (2) Showing RBs drafted in the top-20 picks have fared no better in YPC than league average; (3) RBs have high bust rates; (4) RB evaluation is tough; (5) RB rookie contracts are bad values; (6) RBs are fragile; (7) RBs have short careers; and (8) RBs don’t matter.

We’ve already addressed (1), and (3-7) are arguments for why RBs are bad draft and contract values, but don’t support the essential thesis of the thread which is that even, healthy productive, cheaply-acquired running backs don’t matter. Point (8) is begging the question — simply stating running backs don’t matter to their franchises is the point the thread sets out to prove. The author states it, links to another article he wrote using three backs as examples and some proprietary stats to reach his conclusion. (Decide for youself whether it’s the QED that it and the thread purport to be.)

But let’s focus on claim (2) because that’s the most pertinent to our question, and it’s also the one the thread author cites: “that backs picked in the top-20 (since 2004) are no better than league average on a per-carry basis.”

Let’s leave aside the possibility of arbitrary endpoints on “top-20” that excludes high-efficiency first-round backs like Chris Johnson, Mark Ingram and Felix Jones — there are plenty of bad backs in that range too, so I’ll assume using “first round” yields more or less the same result.

And let’s also set aside that whether talent evaluators can pick the right backs isn’t exactly the same question as whether the right backs matter. I can more or less go along with the idea that if backs matter, then there should be *some* benefit to prioritizing them in the draft. But that the top-20 backs manage league average efficiency actually shows just that.

Consider that former No. 3 overall pick Trent Richardson got 614 career carries at 3.3 yards per pop. Would a late-round back with that paltry level of efficiency ever crack 100 career carries? It’s only because Richardson was drafted so early that he lasted long enough to bring down the average. All things being equal, you should expect later-round backs to be *more* efficient to win a large workload, and yet they are equal.

Moreover, backs selected in the first-round of the modern NFL almost by definition have to be workhorse-capable — to the extent anyone is still willing to use a top-20 pick on a back, it’s not going to be on a quick outside-running timeshare guy. That means, the first-rounders are usually early-down, short-yardage and volume guys, also a per-carry drag. So that they’re keeping pace with a bunch of fresh-legged, passing-down and quickness specialists who had to produce on the field to reap a continued opportunity is actually a point in their favor.

Tweets 6 and 7

This article focuses on the values of different types of targets, and for our purposes seems to show targets to outside receivers have a much higher expected value than targets to the slot, tight ends or in the backfield. But the efficiency difference among recipients of non-outside targets (tight ends, slot receivers and backfield pass-catchers) is small:

I’m not sure how targets to outside receivers being more efficient than those to slot or backfield receivers advances a case against running backs mattering. All it does is establish a baseline of expected success from those kinds of plays, and you could compare each receiving back’s success to the out-of-the-backfield-catch baseline, just as you could with each back’s YPC, given various boxes and play designs.

Moreover, as Adam Harstad points out:

The dump-off to the back isn’t usually the first read on a passing play, so comparing him to targeted players on the outside or deep down the field isn’t apples to apples. Surely, the efficiency you gain from completing a pass to a back as opposed to throwing the ball away or taking a sack is positive — and those are often the options a quarterback faces when no one else is open.

The thread has a few more links to pieces showing running efficiency isn’t necessary for play action to be effective or that a good running game keeping the defense rested is a myth. Even if true — and I didn’t really dive into them — I think those are tangential to the main thesis of the thread, which is that running backs don’t matter.

To establish that, the author needed to demonstrate either that running efficiency doesn’t matter to winning, or that even the best individual backs don’t contribute to rushing efficiency. Neither bar has been cleared.

At best his thread shows that passing efficiency is more important than rushing efficiency, something that was not in doubt, and that *on average* rushing efficiency, as measured in a simplistic way, seems to contribute only marginally to winning. It precludes neither particular NFL teams deriving significant benefit from their rushing efficiency, nor certain running backs on particular teams driving such efficiency.

Moreover, using league-wide statistical averages to demontrate all “running backs don’t matter” is like using the fickleness of kickers generally to conclude Justin Tucker doesn’t matter to the Ravens. If anything, it means Tucker is more valuable — in a sea of mediocrity, greatness stands out even more starkly.

And I suspect it’s that way for most positions. There are x percent that make a meaningful positive (or negative) difference, and the vast middle (100-2x) percent don’t significantly drive outcomes. I also suspect it’s not always obvious which backs are great, though playing in bad situations, or mediocre but putting up great stats in ideal ones. NFL teams are complex systems wherein the whole is not equal to the sum of its parts, or the outputs linearly predictable from the inputs.

The best we can do is come up with general rules like quarterbacks are typically more important to the system than running backs, or passing-game efficiency is generally more important to winning in the modern NFL under the current rules, with its common defensive personnel groupings. While blanket statements like “Running backs don’t matter” might drive engagement, they rarely hold up under scrutiny.