Tom and Mike have climbed to the top of the Hill of Nonsense and unleash another wave of insane wagering on the world.
10 Apr 2009
Guest column by Mike Horn
"In our league, you're seeing less and less of that one back, that [one] guy doing it all the time," Sparano said. "I mean you can look around the league and really see that's something that doesn't last very long because of injury, and because of how good the other people are."
-– Miami Dolphins coach Tony Sparano, quoted in the South Florida Sun-Sentinel, April 2, 2008
Every so often, some talking head or coach will go on about the increased use of a "two-back" system or some other method of employing running backs. Usually these comments center around the way a few teams that have just made the playoffs use their running backs. The commentator claims this is now a "trend" without any real analysis of what has gone before. I'll wonder, are there really more teams using "two backs" now than there were? Are fewer teams using a committee approach or one heavily-used runner? Can the evolution of NFL running back usage be quantified? This is my attempt to answer those questions.
I decided to begin my study with the 1960 season. Like any start date, this is arbitrary. The founding of the AFL that year and the NFL's reactionary expansion marked a clear change from the small number of teams playing in the '50s to the much larger major league pro football of the '60s and beyond. I felt therefore that 1960 was a good point to start analyzing running back usage.
There are other key years that could have been chosen: the full NFL-AFL merger in 1970, the rules changes to increase offense in the late '70s, the start of the 16-game schedule in 1978, adoption of full free agency in 1994, etc. And there are the strike years of 1982 and 1987 to factor into the story. As necessary, I'll address all these dates, but I opted for the earliest date in order to maximize the length of time studied and to show how the game evolved from the '60s into the rushing-heavy '70s and beyond.
I'd love to use advanced statistics to tell the story. Unfortunately, FO's DYAR and DVOA don't go back as far as 1960. I decided to use rushing attempts and running back receptions as the key statistics to analyze usage patterns. Perhaps success-related numbers like yards or yards per attempt would have been better than activity numbers, but I stuck with carries and receptions as more indicative of usage.
There are lots of terms that get thrown around when discussing running backs. Two traditional terms are "halfback," which has fallen out of use, and "fullback," which persists. But the 21st-century fullback bears no relation to the fullbacks of the 1960s. Perhaps the greatest running back of all time, Jim Brown, has "fullback" listed in his statistics chart at Pro-Football-Reference.com. A history of the fullback position and its evolution might be interesting, but that's not what I'm writing. I'm going to treat both fullbacks and halfbacks as running backs for this analysis. I'm also going to consider I-backs, split backs, H-backs, and any other term descriptive of where players line up in formation simply as running backs.
I will use the term "lead back" to indicate the running back who had the most rushing attempts on a team in a given season. I will also refer to the lead back as the RB1. "Second" running backs, or RB2s, will likewise be the running backs with the second-most carries on their teams. Note this is a purely statistical description; often a team's RB2 will have been its intended starter or will have started some number of games, but injury or ineffectiveness caused him to get fewer carries than the "lead back." And in the '60s, at least, the RB2 may have actually played a different backfield position than the lead back and not been his alternative or backup at all. I am not trying to figure out intent, which would be a much longer and more detailed study, just actual usage. Similarly, "reserve" running backs will be those with the third-, fourth-, fifth-, etc., most carries -- RB3, RB4, RB5, and so on. The 1970 Saints had an RB13, Claxton Welch, although he had zero carries, as did Gary Lewis, the RB12 by virtue of the alphabet, while the RB10, Bill Dusenbery, and RB11, Earl Gros, had four rushes each.
I'll introduce and define a few more terms later, but for now let's look at how the lead back's role has changed.
First, running back rushing attempts per team in 2008 (24.4) are not that different than they were in 1960 (26.3) or 1961 (25.2). Obviously, the graph below shows there have been a lot of changes in between to get back to this point.
From 1961 on, there was a relentless increase in running back rushes per game until they peaked in 1977 (34.4), when the league began to change the rules to favor passing more. This had the desired effect, and the league returned to 1961 levels of running back rushing attempts by 1989 (discounting the 1987 strike year low point; for some reason, the strike years both fell below the trend in RB rushes, although not significantly). In 1990, the NFL dropped below an average of 25 running back rushes per team per game and has not gone above that mark since. It has stayed generally flat around 24.5 since then, with a low point of 23.5 in the 2000 season. How have those carries been divided up?
The next chart shows the breakdown of overall NFL running back carries. For each season, the RB1 rushes were summed and then divided by the total number of RB attempts. While this hides a lot of individual differences across teams, it should capture general trends. And it clearly shows the increased reliance on one lead back to carry the load from about 1980 on.
Meanwhile, second RBs saw a gradual decline in their share of the carries from 1970 to 2000, with a slight upward trend since then. RB3s have half the share of rushes (8 percent) that they had in 1960 (16 percent). And the rest of the backs on NFL rosters, who for a while from the late '60s to 1980 collectively carried the ball as often as RB3s, now make a token 5 percent of the league's rushes.
The next table breaks the data into three periods, with the minimum and maximum percentages of the total running back rushes that each of the various types of RBs received in those time spans:
|Ranges of RB Rushing Shares|
|*Omits the 1987 strike year share of 48%|
From 1960 to 1980, lead backs as a group never got more than half the league's rushes. Since 1980, they've never gotten less than half (except in the strike year of 1987 when the use of replacement players naturally spread the carries around more). From 1981 to 1993, RB1s never handled more than 60 percent of the total running back carries. And since the dawn of full free agency in 1994, RB1s have never seen their share fall below that mark (although in 2008, they dropped to that level).
Meanwhile, RB2s have gone from getting 25 to 33 percent of all carries from 1960-1980, to around 25 percent in the middle period, and then to 20 to 25 percent after 1993. They have only climbed back above 25 percent in the last two seasons, with 27 percent in 2008 -- not coincidentally, when the RB1 dropped to 60 percent. Reserve backs (RB3 and Other) also made 25 to 33 percent of the rushes (about the same total share as second RBs) in the '60s and '70s, then fell to less than 25 percent in the 1981-1993 time frame and to no more than 15 percent from 1994 on.
Basically, when NFL offenses had two true running backs in the backfield, one got most of the carries, another got some, and their backups shared the rest. But as offenses decreased the rushing responsibilities of one back (the fullback) and the run-blocking role of the other back (halfback, tailback, etc.), the rushing share of the lead back increased. The RB2 became more of a backup or change-of-pace player who substituted for the lead back instead of being another runner on the field at the same time as the RB1. In turn, this meant the "reserves" (by my definition in this study; the blocking back who got a few carries was of course also a starter in the actual line-up), instead of sharing the carries destined for two runners in the formation, were now getting only a few leftovers plus some opportunities when the top two backs got hurt.
A turning point may have been reached in 2002, when lead backs made a fraction under 70 percent of all the rushes in the NFL. For the next four years their share hovered around 65 percent before dipping to 61 percent in 2007 and 60 percent in 2008. It remains to be seen whether this is a true change in running back usage or just an anomaly, but the fact that it has happened two years in a row may mean that change has come.
Also, it's important to understand that while lead backs have seen their share of carries increase, their total number of carries per game played (which factors out schedule length and number of franchises) hasn't changed as dramatically:
Lead back usage in the '60s was flat, around 12.5 carries per game. Then it climbed to a plateau around 14 per game in the running-dominated 1970s and to around 15 in the early 1980s. After a drop below '70s levels from 1986 until the free agency period, it has held steadily above 15 until the past two seasons. But the real change has been more the decline in total running back carries since 1977 than an increased reliance on RB1s as a group. Looked at this way, 2002 doesn't stand out as a turning point in lead back usage, but 2007 might be.
What about those runners who carry a very heavy load? I think everyone knows they are a lot more prevalent than they used to be, but has that shifted at all in the last few years as Coach Sparano seems to claim?
First, I want to define how I'll use the term "workhorse" back. Obviously, 370 rushes was going to be too many. Not many running backs reach that workload, and the Curse of 370 ensures that not many do it very often. Because of the different season lengths in the period studied (12, 14, and 16 games, plus the strike years), I decided on a per-game count rather than a season total. Twenty carries per game was a nice round number that seemed like a heavy load week-in and week-out but didn't approach the accursed 370 for a 16-game season. This study will consider a "workhorse" back to be one that got 20 carries per game.
The first NFL workhorse was Steven Van Buren, of the 1949 Eagles, followed by Eddie Price of the 1951 Giants. Then there was a lull of several years until Jim Brown did it in 1958. The next chart shows the percentage of teams (rather than number of teams, to factor out expansion) that have employed a workhorse back in every season since 1960.
Brown was the first back to be a workhorse for an extended period of time, notching six total workhorse seasons in his career. Jim Nance of the '66 Patriots did it the year after Brown retired, and then there were no workhorses again for several years. The '60s (the red line on the graph) were still the pre-workhorse era.
The 1970s saw the dawn of the first workhorse-back era. Matt Snell ('70 Jets) broke the 20 carries per game barrier, but in only three games. There then was a gradual increase in workhorse backs over the next decade that peaked in the early to mid-'80s followed by a decline until 1990 when there were none, ending the era (blue line).
In 1990, it would have been easy to think that the day of the workhorse back had ended, probably because it was too hard to keep them healthy. In fact, it was the start of what looks like the golden age of workhorse backs.
The second workhorse back era started in 1991 (green line), when Emmitt Smith met Norv Turner, Barry Sanders had his first workhorse season, and Gaston Green had his one-time experience with a heavy workload in Denver. From that year on there was a steady increase in teams using workhorse backs until 2003, when more than 40 percent of teams employed one. Now the league appears to be moving away from that method of using running backs, although obviously it's too soon to tell if this is a permanent change.
The next chart takes the data above and plots it as trendlines (second order polynomial). If the current trend is consistent with previous examples, the use of workhorses will decline for a few to several years and then revive. It might even peak at a higher level next time.
There have been 89 backs who have combined for 200 workhorse seasons since 1960. But almost half of those have been produced by only 17 runners. There were 47 running backs, like Gaston Green or Jim Nance, who only sustained this type of heavy workload for one year.
|Distribution of Workhorse RB Seasons|
|Career WH Seasons||No. of RBs||Total WH Seasons|
|4 or more||17||93|
The "true" workhorses, those with four or more seasons with that type of usage, started with Jim Brown. There were four in the first workhorse era and twelve in the Golden Age:
|Workhorse RBs by Era|
|1960s||'70s and '80s||'90s to present|
|Jim Brown (6)*||Walter Payton (8)||Edgerrin James (8)|
|Eric Dickerson (7)||Emmitt Smith (7)|
|Earl Campbell (5)||Curtis Martin (7)|
|O.J. Simpson (4)||Eddie George (6)|
|Barry Sanders (5)|
|Jerome Bettis (5)|
|LaDainian Tomlinson (5)|
|Clinton Portis (5)|
|Stephen Davis (4)|
|Shaun Alexander (4)|
|Priest Holmes (4)|
|Ricky Williams (4)|
|Number of Workhorse seasons in parentheses; *Jim Brown includes workhorse seasons prior to 1960.|
In fact, it has been the number of true workhorse RBs that has set the "Golden Age" apart from the '70s and '80s. If those backs are taken out of the data, the period from 1970 on looks pretty much the same: About 10 percent of the teams in any one year will employ a workhorse back if they don't have one of the backs named in the table above. That is shown by the dotted black line in the next chart, which excludes the true workhorses. There was a small increase in 20-carry-per-game seasons among other backs from 2003 to 2006, but the fact that eight of the 17 true workhorse runners were posting high workloads in that span is also a factor in the spike during that period. Excluding the true workhorses, 2007-2008 look to be more in line with the running back usage pattern of the entire period from 1970-2006 -- in other words, it may be a regression to the mean. Or it could be that several teams tried the workhorse system in 2003-2006 and found that it didn't work without a true workhorse -- which might be hard to find. This would support Coach Sparano's statement.
Norv Turner broke into the NFL coaching ranks with the Los Angeles Rams in 1985 as a receivers coach. He got to see first-hand one of the foremost workhorse backs in NFL history, Eric Dickerson, the Curse of 370 exceptionaire. Norv apparently was strongly affected by this. When he became an offensive coordinator, and later a head coach, he became the foremost user of workhorse backs.
|Norv Turner's Lead Backs Per Game|
So far, it looks like in recent years there have been drops in the share of the rushing load carried by lead backs, and in the number of workhorse backs. Those declines may be temporary, or they may be real changes in the way running backs are used, but they have occurred. It is often suggested that teams are adopting a two-back system as an alternative. First, what is a two-back or dual-back system? And second, has it become more prevalent throughout the league?
I developed three criteria to describe a "dual RB" team. First, the RB1 and RB2 on a team had to combine for more than 85 percent of the team's runs by running backs. This indicates that a team relied on two players as opposed to a committee for the vast majority of its running game. Second, the RB2 had to have at least a third of the team's total RB carries. This screens out teams where the RB1 got a huge percentage of the carries and the RB2 got very few. Earlier we saw that the league average share for RB2s has never been more than 31 percent, so this is a pretty high number, but it does allow some flexibility in letting teams qualify. Finally, the sum of the RB1 and RB2 games played had to be more than 80 percent of their possible appearances; for example, in a 16-game season, they had to combine for 26 game appearances out of a maximum of 32. In combination with the other criteria, this screened out almost all times when a team had an RB1 get hurt and just used his backup as a surrogate RB1 as opposed to sharing carries between two main rushing options.
I then identified recent teams that struck me as employing dual RBs: the Vikings, Jaguars, and Panthers in 2007; and the Saints, Colts, Bears and Patriots in 2006. When tested against my definition, all passed. So I felt comfortable that this was a reasonable if not perfect description of teams that used dual running backs, although it probably excludes a few teams who were genuinely trying to use two runners.
The chart below shows that dual running backs have been used throughout the period since 1960. Their heyday was from 1964 to 1972, then this usage fell out of favor for more than a decade. (Even the dual-back system of the champion 1976 Steelers, with two 1,000-yard rushers, did not inspire a revival, although the rule changes in the late '70s probably affected that.) From 1985 on there has been a general upward trend in the percentage of teams employing this system (the black line shows the trend across the entire period). The strike year of 1987 was the last time there was no dual-RB team.
Of course, the modern dual-RB system basically consists of alternating tailbacks, only one of whom is in the lineup on a given play, while the '60s version was a fullback and halfback on the field at the same time, either of whom could get the ball. This required the halfback to be able to run-block to some extent, a requirement lacking for the modern tailback. Which is why for all the talk by coaches employing dual-RB systems of playing both backs simultaneously, there rarely is a time when two tailbacks are in the backfield together. Which one would block for the other? So while the net effect of two backs sharing rushing attempts has come back, the old methods have not.
There has been a definite spike in dual-RB teams the last three seasons, as 2006 to 2008 saw the first three seasons where more than 20 percent of teams used dual running backs. In fact, as the next table shows, there were more than twice as many dual-RB teams from 2006 to 2008 as in the 11-year period from 1974 to 1984. The minimums and maximums in the table refer to the lowest and highest percentage of teams using dual running backs in a season during the relevant period.
|Distribution of Dual-RB Seasons|
|Years||Teams||Dual RBs||% Dual RBs||Min||Max|
Aside from 1974-1984, and the last three years, the other three periods shown in the table were generally consistent with each other. It remains to be seen whether 2006-2008 marks the start of a new period with much wider use of a dual-RB offense or whether the league will regress to its long-term mean of 10 to 12 percent of teams employing two backs, but the recent trend seems very strong.
Over the entire period, there is almost no correlation (0.07) between the percentages of teams using workhorse backs vs. dual RBs. From 1970 on, when the workhorse system began to be used by teams not employing Jim Brown, there actually is a slight positive correlation (0.20) between the two methods. In the free agent era (1994 to present), there has been a more noticeable negative correlation (-0.44), indicating that the there may be a competition between the two systems. But historically there has been room for both in the league. In fact, over the last 38 years, the combined use of those two systems has been increasing:
Well, if those two usage methods have gone up, what has gone away? That will be addressed next week in Part II of this essay.
Mike Horn writes a stats-based column for FantasyGuru.com. In real life he is a retired Army officer working as a military defense analyst in the Washington, DC area. He occasionally comments as MRH on Football Outsiders discussions and has written two previous guest articles on tight end usage and how teams used draft picks over a five-year span. If you are interested in writing a guest column for Football Outsiders, please send an idea or rough draft to info-at-footballoutsiders.com.
38 comments, Last at 16 Apr 2009, 9:24am by mrh