≡ Menu

Don't worry, this will all make sense by the end. I think.

Don't worry, this picture's presence will make sense by the end. I think.

Two years ago, I wrote this post on running back aging curves. One conclusion from my research was that age 26 was the peak age for running backs, which was immediately followed by a steady decline phase until retirement. In that study, I only wanted to look at very good-to-excellent running backs in the modern era; as a result, I was forced to limit myself to just 36 players. I’ve been meaning to update that post, but wasn’t quite sure what methodology to use.

Last year, Neil wrote a very interesting post on quarterback aging curves. In it, Neil computed the year-to-year differences in Relative ANY/A at every age. While reviewing that post, a lightbulb went off. We can greatly increase the sample size if we only look at running backs from year-to-year, and not just the best running backs on the career level.

There are 723 running backs since 1970 who had at least 150 carries in consecutive seasons and who were between 21 and 32 in the first of those two seasons. For each running back pair of seasons, I calculated how many rushing yards the player gained in Year N and many yards he gained in Year N+1. Take a look:

Age#Rsh Yds Yr NRsh Yds N+1Diff

Just so we’re clear on what that table says, let me walk through an example. There were 43 running backs who were 22-years-old and had at least 150 carries, and also had at least 150 carries at age 23.1 Those running backs averaged 1,057 yards at age 22 and then 986 yards at age 23. While these numbers may be useful for reference purposes, if you’re like me, a table like this isn’t particularly intuitive; a graph would be much better at explaining what the data tell us.

If we assume that a running back will rush for 1,000 yards at age 21, and then gain or lose the amount of yards showed in the table above each season (i.e., will rush for 89 more yards at age 22, then 70 fewer yards at age 23, then 74 more yards at age 24, and so on), he would produce the following career curve:

rb aging

The dotted blue line tracks our fictional player who rushed for 1,000 yards at age 21. But the smoothed black line is probably the more useful one, which is presumably a better representation of the effects of age on a running back’s production. What’s interesting to me is that despite using several different variables and measuresh, this study comes to a pretty similar conclusion to the last one: running backs peak at age 26, and then begin a steady decline. It also suggests that for older running backs, some dropoff should always be expected from year-to-year, even if they have done very well in the prior season.

Do you know which running back’s career most closely resembles our set of averages? The answer: Buffalo Bills Hall of Famer Thurman Thomas. The graph below shows Thomas’s yearly rushing averages from ages 22 (when he entered the league) through 33: as you can see, the curve is shifted upwards because he was better than the average back, but the shape of the curves are pretty similar. If nothing else, may you remember that Thurman Thomas aged like your typical running back.


Thomas had a great five-year peak from ages 23-to-27 on which his Hall of Fame career was built. That was followed by a respectable but inferior three-year period; after that, his production fell off a cliff sharply, beginning after his age 30 season.

What can we take from today’s post? The survivorship issue2 has not been resolved, although I don’t think it’s a huge issue here. If anything, my guess would be that it understates the degree of magnitude by which older running backs decline. Another conclusion is that running backs don’t take very long to progress in their careers: a running back in the draft is not far behind, if at all, from a running back who is 25 or 26.3 This is your 1,283rd reminder that giving big money contracts to older running backs is one of the riskiest moves an organization can make. Three years ago, a 26-year-old Maurice Jones-Drew led the league in rushing. Over a week into free agency, the 29-year-old running back is still looking for a job. Last season, a 26-year-old Knowshon Moreno gained 1,586 yards from scrimmage and scored 13 touchdowns. He’s still looking for a job, too.

Update: For those curious about the choppiness at the beginning of the curve, the table below shows the results for the 22-year-old running backs:

Running backYearTmageYr N GYr N RshYr N RshYdYr N+1 GN+1 RshN+1 RshYdDiff
LeSean McCoy2010phi22152071080152731309229
Jonathan Stewart2009car2216221113314178770-363
Knowshon Moreno2009den221624794713182779-168
Ray Rice2009rav22162541339163071220-119
Rashard Mendenhall2009pit22162421108163241273165
Kevin Smith2008det221623897613217747-229
Maurice Jones-Drew2007jax22151677681619782456
Adrian Peterson2007min22142381341163631760419
Steven Jackson2005ram22152541046163461528482
Kevin Jones2004det2215241113313186664-469
Clinton Portis2003den22132901591153431315-276
LaDainian Tomlinson2001sdg22163391236163721683447
Edgerrin James2000clt221638717096151662-1047
Ron Dayne2000nyg221622877016180690-80
Ricky Williams1999nor2212253884102481000116
Fred Taylor1998jax2215264122310159732-491
Fred Lane1997car221318280914205717-92
Warrick Dunn1997tam221622497816245102648
Karim Abdul-Jabbar1996mia2216307111616283892-224
Mario Bates1995nor221624495114164584-367
Marshall Faulk1995clt2216289107813198587-491
Curtis Martin1995nwe22163681487163161152-335
Jerome Bettis1994ram2216319102515183637-388
Natrone Means1994sdg2216343135010186730-620
Rodney Hampton1991nyg2214256105916257114182
Emmitt Smith1991dal22163651563163731713150
Johnny Johnson1990crd221423492615196666-260
Reggie Cobb1990tam221615148016196752272
Barry Sanders1990det22162551304153421548244
Sammie Smith1989mia221320065916226831172
John Stephens1988nwe2216297116814244833-335
Thurman Thomas1988buf2215207881162981244363
Greg Bell1984buf2216262110016223883-217
Marcus Allen1982rai229160697162661014317
Tony Collins1981nwe22162048739164632-241
Andra Franklin1981mia22162017119177701-10
Joe Cribbs1980buf22163061185152571097-88
Ottis Anderson1979crd22163311605163011352-253
Clark Gaines1976nyj221415772414158595-129
Mike Thomas1975was2214235919132541101182
Walter Payton1975chi2213196679143111390711
Franco Harris1972pit2214188105512188698-357
John Riggins1971nyj221418076912207944175
  1. Careful readers will quickly recognize that this opens us up survivorship bias issues. Running backs who decline greatly from Year N to Year N+1 won’t have their failures recognized in this study if they fail to hit the 150-carry threshold. Unfortunately, I can’t quite think of a good solution right now, but perhaps one will come to me (via the comments?) by the time Part III is produced. []
  2. You guys always read the footnotes, right? []
  3. Of course, we’re focused just on production in the running game here, not things like pass protection. []
  • Sunrise089

    Missing word: “of you’re like me, a table like this ISN’T particularly intuitive.”

    Also, what’s with the choppy curve early? I’d think 40+ seasons would be enough to smooth it out. Is there anything weird going on with draft age versus skill?

    • Chase Stuart


      I posted the data on all the 22-year-old backs. I think the choppiness there is almost entirely a function of injury: as you can see in the average row at the end, the numbers are pretty similar on a per-game basis. I thought about using per-game numbers for all backs (while keeping the 150-carry cutoff) and then prorating every year to 16 games, but I didn’t since risk of injury is one of the things people are focused on with older running backs. Of course, maybe this is just a reminder that any running back can get hurt.

  • Sunrise089

    Oh hell, autocorrect strikes again. Typo in my typo correction above. “if you’re like me…”

    • Chase Stuart

      Typing ain’t so easy is it!!

  • Jp

    Chase, any well-known backs buck the trend?
    How was Barry Sanders’ production when he retired?

    • sn0mm1s

      Great – at least until he played injured the final 1/3 of the season. I suppose the season was still great – but not Barry Sanders great.

      First 10 games
      226 carries, 1126 yards 5.0 YPC
      Last 6 games
      117 carries, 365 yard 3.12 YPC

      Total 343, 1491, 4.3 YPC

      • sacramento gold miners

        It’s certainly possible Sanders was beginning to decline due to the aging process over that final 1/3 of his last season. But there’s another factor beyond the numbers which affects performance. Sanders saw no hope for winning in Detroit, disliked the head coach, and that led to his retirement. When a player has lost the passion, it’s going to show up on the field, and that 3.12 ypc wasn’t indicative of the great play earlier that season.

  • I’m sure there are exceptions, I’d argue modern day exception would be Matt forte ?

  • Nate

    Does the chart look any different if you use yards per carry instead of total yards?

  • I know you don’t want to include pass blocking, but what about including receiving production and using scrimmage yards from n to n+1? That may be a better way of describing usage rates, and expanding the threshold to 150 offensive touches would give you even more players to work with (then again, you’d have guys like Larry Centers who confuse things). I think a lot of backs have a similar dropoff in receiving production, but there are probably exceptions.

    • Chase Stuart

      My guess is that little changes, but a good idea for a future post.

  • Richie

    Two years ago, I wrote this post on running back aging curves.

    Can’t believe this site has already been up for 2+ years.

    • Chase Stuart

      Me too! Although the two year anniversary comes in June. I believe you lead the way with comments here, by the way!

  • Tim Truemper

    Two thoughts: What if total yards from scrimmage was used rather than just rushing yards; and what RB’s buck the trend, i.e. that they had relatively strong performance after age 30 when compared to their peak years.

    • Chase Stuart

      My guess is the numbers would be very similar if total yards from scrimmage were used. As for the other question, I think Part I answers that one pretty well.

  • Richie

    RB’s have recently come to be considered a position that is easy to fill by just about anybody. I always figured this is because so much of a RB’s success hinges on the offensive line opening up holes for him.

    But if it’s true that RB’s are heavily reliant on their offensive lines, wouldn’t that mean that RB production wouldn’t drop off so quickly and consistently as they age? After all, how much “worse” is a RB really going to get between ages 29 and 32, especially if his line is really doing most of the work?

    • Chase Stuart

      That’s a good point. I hadn’t thought of it like that before.

      • Richie

        Neither had I.

        Thinking about it some more. What skills do good RB’s have?

        – Vision. The ability to quickly spot holes in the defense, and tacklers coming their way. I wouldn’t think this is a skill that is lost at such an early age. I would think this is something that could last until a player is 40 or 50.
        – Juke. The ability to make quick changes to go where the defenders ain’t. This seems like the main skill that would deteriorate. It probably doesn’t take much difference in juke ability to explain the difference between a “good” RB and an All-Pro.
        – Power. The ability to break tackles and/or run over defenders. This is another skill that probably doesn’t deteriorate as quickly. Some of those strong man and physical fitness competitions have guys who are in their 40s. I think guys can keep most of their strength into their 40s or 50s.
        – Speed. The ability to outrun defenders. (As opposed to the quicker “juke” ability that is just about moving very short distances quickly.) Speed is something else that players probably lose at an early age. But I don’t think this is a critical skill for a RB. Yeah, it’s nice if a RB can outrun some defenders, but most of the true success of a RB comes from the ability to tack on a bunch of 5 to 10 yard runs over the course of a game or season. Though I would be really interested to see some sort of distribution chart of length of runs for good RB’s over the course of a season. Maybe a guy really does need to have about 10 runs per year that go for 20 yards or more in order to put up a 1,200+ rushing yard season.

        • sn0mm1s

          Speed is very important in regards to hitting a hole before it closes. I forget who interviewed Sanders but he asked (and I paraphrase) “What is it that you see that no one else does?” Sanders’ reply “I see the same thing you do – I can just get there.”

          Also don’t forget recovery time. It takes longer and longer to recover from the pounding a RB gets each game.

        • Rob Harrison

          Good points, but I think you’re wrong w/r/t power. From what I can see, I think power effectively deteriorates quickly, not because RBs lose strength with age, but because the more a back runs over/through defenders, the more beaten up he gets and the more injuries he accumulates.

    • sn0mm1s

      Oline is likely only important for the first few yards. The great RBs need less of an Oline and when they do get a hole they can do more than an average back. The reason that the position can be easily filled by any RB is that even the great RBs generally rush for 3 yards or less. The running game should really only be used for short yardage – and any RB can fill that role. Something I posted on another site:

      My point is there is no running game, nor has there ever been a running game (or RB), that rattles off 4-5 yard gains at will. QB teams don’t hit every pass – but the great QBs are hitting 60%+ of their passes. RBs don’t come close to getting 4-5 yards 60% of the time.

      Let’s look at recent “great” seasons
      Peterson – 2012, gained 4+ yards ~46% of the time.
      Tomlinson 2006, gained 4+ yards ~46% of the time.
      Faulk in 1999,2000,2001 gained 4+ yards 50%, 50% and 52% of the time.

      The closest you are going to get to 60+% (that I can look up) is Jamaal Charles 2010 (who split time). He averaged 6.4 YPC and got 4+ yards 57% of the time.

      Now, those career rushing bests, and all time NFL seasons.

      Tom Brady, who most agree he had one of his worst statistical seasons of his career last year, including sacks, gained 4+yards on passes on ~50% of his attempts. One of Brady’s worst years passing is as effective or better than AD, LT, and Faulk’s best years rushing. Brady’s best year has him gaining 4+ yards on 60% of his pass plays. However, that doesn’t tell the whole story since a pass plays will generally gain more yards on average to begin with.

    • I think that’s an awesome point. Here’s a potential explanation that still leaves RB easy to fill. Suppose a bunch of equally talented RBs get drafted at the same time. Who survives? The guy lucky enough to have a good line early on. Over time, though, his line ages and regresses (although not as quickly as him). So maybe the line is getting worse as RBs age since the ones that survive had better lines early on.

    • Kibbles

      It could be that RBs are relatively fungible as long as they meet certain physical minimums (i.e. at least X speed, at least Y explosiveness, etc). That would explain why RB production is damn near flat from age 21 through 27- those guys meet the minimum, so year-to-year production swings are mostly just noise. Then, starting at age 28, RBs start to decline physically, and as soon as they no longer meet the physical minimums, their production falls through the floor.

      Not saying that’s necessarily the best explanation, it’s just one possible theory that would explain both the “RBs are fungible” phenomenon and the “old RBs are less productive than young RBs, despite the fact that RBs are fungible” phenomenon.

  • Kibbles

    Chase, what was the distribution of the declines among older RBs? For instance, your chart says that between age 29 and 30, RBs decline by 100 yards on average. Now, that could potentially mean two different things. It could mean that all RBs decline by a little bit (approximately 100 yards). Or it could mean that 80% of RBs don’t decline at all, but 20% of RBs fall off a cliff and decline by 500 yards. Either would explain the observed decline, but both carry very different implications. The first implies that all old RBs are certain to decline a little. The second implies that old RBs are simply a heightened risk to decline dramatically when compared to young RBs.

    • Chase Stuart

      I’ll have to check.

      • Pierre NyGaard

        Did you find an answer? This seems particularly important.

  • James

    I read the footnotes!

    Unfortunately, there isn’t a perfect solution to the survivorship problem. One way to test how big it is would be to compare the YPC of the players in the study versus the YPC of the league as a whole. It’s almost certainly higher since we are only looking at the good, healthy players in N and N+1 and the average players that had a lucky, healthy N season and a healthy N+1 season, but presumably you care about aging for all RBs. The lucky N seasons might also mean you are OVERstating the decline, due to regression to the mean.

    Another way to determine how big survivorship is: 1. How many players at each age (or what percentage) failed to hit 150+ carries at N+1? 2. What if you did the same analysis, but only looked at players that did NOT get 150+ carries in the next season? That might tell us if it’s primarily due to injury (good performance but few games) or ineffectiveness (bad performance), or at what age the switch occurs.

    One option is to assume some sort of replacement level production from the players that fail to get to 150+ carries in N+1 and factor them in, and while you sort of assume your answer that way it might be most accurate. Another is to do away with the minimum carry limit and look at all players, but weight them according to the harmonic mean of carries for *both* seasons (harmonic mean where 1/x = 1/N + 1/(N+1)), but that minimizes effect of injury. You could also look at only players that had long careers, which will make a flatter curve, but give you a ‘minimum’ aging decline to contrast with the one above. You could also regress the N season performances and rerun the analysis, as the N seasons are likely good luck seasons (good luck seasons more likely to hit 150 carries in N and N+1 than bad luck seasons!).

    If you do some or all of those you should get a variety of similar answers, and then combine them for one overall solution. Especially since one method’s weakness will be covered up by anothers!

    • Chase Stuart

      Interesting ideas, thanks.

  • Pingback: The NFL Likes Running Backs Again. Kinda, Sorta. – FiveThirtyEight | Quick & Fast Sports News()