The Running Backs Don't Matter theory and the discourse surrounding it are very interesting to me, first of all because I love running backs. Their task -- to take the ball from here to as close as you can get to over there while eleven other dudes try to stop you -- is beautiful in its simplicity and in the variety of ways in which it can be attempted. Barry Sanders ran around defenders, Earl Campbell ran through them, Chris Johnson simply ran past them, while others have tried everything in between.
The second reason I'm fascinated by #RBsDM is because I like to understand things. Secondary to the aesthetic and visceral experience of watching the game, the strategy of football is one of the major reasons I love the sport, and given all of football's moving parts, how that strategy can be optimized is interesting as a field of thought. The idea that running backs don't matter is particularly interesting in this area because it flies in the face of conventional wisdom in a way that opens the door to the exploitation of strategic edges, and (perhaps most importantly) because it is backed up by compelling statistical evidence.
I'm not really here to recap the last five or ten years of research that has been poured into this topic (I'm certainly not even familiar with all of it), but a lot of it can be found here. Given that research and more, I find myself fairly sympathetic to the idea that running backs might not matter (much).
At its core, the #RBsDM movement/theory/cult is based on three basic tenets. First, that passing the ball is so much more efficient than running the ball in the modern NFL that players whose primary influence is on the running game are inherently far less impactful on the outcomes of games than players whose primary impact is on the passing game. Second, that, given the contributions of blockers and the efforts of defenders, running backs have a relatively small impact on the outcomes not just of games, but of individual running plays. And third, running backs get injured more often and have shorter shelf-lives than players at most other positions.
Tenets one and three are givens, and are relatively uninteresting to me as a result. Tenet two is fascinating, because it goes against everything I thought I knew about running backs after growing up watching Shaun Alexander and LaDainian Tomlinson win MVP awards and popping in old NFL Films VHS tapes of Jim Brown, Gale Sayers, and Eric Dickerson running roughshod through defenses completely geared up to stop them. So, my small and insignificant contribution to this discourse engages with tenet two.
In general, I am interested in quantifying the on-field performance of running backs, and I have developed several rushing efficiency metrics to that end. The specifics of those metrics is not vital to the analysis I'm doing here, but ease of communication probably requires that I define them anyway. More detailed explanations can be found here and here, but here they are defined on a basic level:
Box-Adjusted Efficiency (or BAE) Rating
Measures running back performance in terms of yards gained per attempt, adjusted for the context of the offensive environment he operates in and the relative difficulty of his carries given the amount of defenders in the box he sees on his rushing attempts.
Relative Success Rate
Measures running back performance in terms of the percentage of his carries that result in a successful play given down-and-distance, adjusted for the context of the offensive environment he operates in and the relative difficulty of his carries given the amount of defenders in the box he sees on his rushing attempts.
Composite Efficiency Score
Uses a player's percentile ranks in Box-Adjusted Efficiency Rating and Relative Success Rate to produce an overall score indicating how successful a player was on his rushing attempts given the context of those attempts, relative to other backs around the league.
Most of the inspiration for the analysis I'm doing in this article came from Ben Baldwin's research in this piece. Baldwin writes:
If rushing were valuable (it's not), a natural question would be the extent to which highly-drafted RBs are better at rushing than other RBs. I used Pro Football Reference to look at the average yards per carry of all RBs drafted in the top 20 since 2004, when a heightened emphasis on illegal contact increased the efficiency of the passing game. Since then, the 17 RBs drafted in the first 20 picks have carried the ball 18,991 times for an average of 4.2 yards per carry. NFL teams combined have rushed the ball 195,381 times for an average of... 4.2 yards per carry. Of the 17 players drafted in the top 20, only 7 have at least 4.3 career yards per carry. With an enormous sample size, there is no difference between these RBs and everyone else.
The conclusion here is plainly stated: there is no difference in rushing performance between highly-drafted running backs and other players at the position. Taken to its logical end (and ignoring the investment-specific implications given that I'm not interested in assessing whether or not running backs are worth first-round draft picks), the real conclusion is that there is no discernible difference between the performance of the running backs who are supposed to be the best players and the performance of all other running backs. In other words, random dudes can do what the supposed elites can do, and therefore, running backs don't really matter.
In the past few months, I've collected data and calculated performance in the previously outlined metrics for every running back who carried the ball in an NFL game between 2016 and 2021. Leaving out the contributions of backs who carried the ball fewer than 10 times in a given season in that timeframe, I have a database consisting of 689 individual running back seasons and a combined 67,057 rushing attempts. That's not as large as Baldwin's 2004-2017 sample (though his test group was only 17 players), but I think it's plenty big for my purposes here.
I have classified each one of those 689 seasons based on where each particular running back fell in their team's rushing volume "pecking order" for that particular season. Player seasons are classified as "1" if the player led their team in rushing attempts (among running backs), as "2" if they finished second on their team in rushing attempts, and so on. For the rest of this article, I will simply be referring to each of these groups using that classification: "1s", "2s", and so on. The sample sizes for each of those groups are as follows:
1s: 192 individual seasons
2s: 192
3s: 166
4s: 93
5s: 37
6s: 36
If Baldwin's yards per carry-based conclusions hold true with these box-count adjusted metrics over this timeframe, we would see that the backs who've earned the most volume (presumably also the most talented backs) are not more effective than the backs with less volume. Formed via my handling of this data in general, my hypothesis is to the contrary: that 1s are more effective runners than 2s, that 2s are more effective runners than 3s, and so on, and generally that the running backs who get the most work are the most effective. At its core, my hypothesis is really that there is a discernible difference in the performances of running backs, and that those who earn more volume will generally perform better than those who earn less volume, presumably because of the disparity in talent that determines which backs touch the ball the most.
Testing this hypothesis is pretty simple. I can just place players into buckets based on how much volume they got (which I've already done in one manner above) and then see if the average performances within those buckets are any different from each other.
Below is a chart that shows the average percentile rank of all the players in each bucket for their performances in all three of the key metrics I outlined earlier, in addition to raw yards per carry:
It's pretty clear here that, on average, the players who receive the most volume are also the players who produce the most efficiently and effectively on a per-carry basis. Notably, players who lead their teams in carries are the only group that runs the ball at a level that is materially above the 50th percentile. 2s are generally of league-average quality, while 3s, 4s, and 5s typically perform similarly to each other, with 6s far less effective than the rest of the population.
Also, because I've successfully created a bar graph in Excel for the first time, here are those results (excluding raw yards per carry) represented in that form:
One thing I find interesting with these results is that Box-Adjusted Efficiency Rating seems to most strongly correlate to where a player fell in his team's backfield pecking order. The drop-off from the average BAE Rating of the collective 1s to the collective 2s and so on was much steeper than the drop-off from Relative Success Rate or Composite Efficiency Score. The drop-off in raw yards per carry (again, not pictured above) was even flatter than for those two metrics, as 1s averaged a 60th-percentile score, 2s averaged a 53rd-percentile score, and 3s were in the 48th percentile.
I theorize that this disparity is due to BAE Rating's simply being a better, more comprehensive statistic than raw yards per carry is. The difference in box counts seen by 1s compared to 2s, 3s, and so on, is quite stark: 1s average 0.09 more defenders in the box on their carries than the rest of their teammates do, a 64th-percentile mark, while 2s average 0.05 fewer defenders in the box on their carries than their teammates do, a mark in the 45th percentile. The greater difficulty that 1s have in producing efficiently on their carries due to those heavier boxes is not evident in yards per carry, but that context is fundamental to BAE Rating. But I digress.
The above analysis does not make any distinction between team-leading ballcarriers who serve as the heads of flatly distributed committees and those who dominate touches in their backfields. One might hypothesize that the latter group performs even better relative to the collective rest of the population than the former group, as the talent that warrants a such a large share of their team's carries would presumably also allow them to produce better results on the field.
This hypothesis seems to be true. I don't want to place arbitrary thresholds on carry totals in order to place players into groups, but we can compare the correlative strength between a player's place in his team's backfield pecking order and that of his raw carry total. If the correlative strength of carries outweighs that of pecking order rank, then our hypothesis is likely correct. Here are the r-squared values representing the correlative strength of those metrics to our efficiency metrics:
Except in the case of raw yards per carry, carry volume correlates more strongly to rushing efficiency performance than does the place in which a player falls in his team's backfield pecking order. None of these r-squared values are very high at all, as there is much more to figuring out who the best runners are than simply looking at who the highest volume runners are, but it seems that a) running backs who get more volume than their teammates generally perform better than those that don't, and b) with slightly more reliability, running backs who get high volume generally perform better than those who get less volume.
These findings don't necessarily contradict the literal findings of Baldwin's look into the performances of first-round running backs vs. the rest of the population of the position, but I do think it contradicts the spirit of those findings. On a macro level, there is a quantifiable difference in the performance of runnings backs classified by volume. If running backs were completely replaceable, then low-volume runners would be just as effective and efficient as high-volume runners.
The other half of that interpretation of these findings is that I believe NFL coaches are doing a good job of identifying the most talented players on their teams and appropriately giving them the most work, at least at the running back position. An alternate interpretation could be that it is easier to run efficiently with high-volume, given that those players can "get into a rhythm." I don't reject the idea that rhythm runners exist, but I think that's more likely to be a trait that applies to a certain sub-species of running back than it is a universal truth that applies to all players at the position. I'm not sure I have any evidence to support that theory, but "the best players get the most carries and are the most effective with their carries" feels like it should be our base assumption given that it makes logical sense and has legitimate evidence in support of it.
Comments
Post a Comment