Writers of Pro Football Prospectus 2008

21 Sep 2006

FO on BSMW: Itty Bitty RB Committee

This week's column, named after my favorite fantasy football team name of all time, takes a look at how teams employing a running back committee perform as opposed to similarly-skilled teams who use a single back.

Posted by: Bill Barnwell on 21 Sep 2006

43 comments, Last at 23 Sep 2006, 10:50am by Bill Barnwell

Comments

1
by Diane (not verified) :: Thu, 09/21/2006 - 2:11pm

first .... and ten

2
by ToxikFetus (not verified) :: Thu, 09/21/2006 - 2:29pm

first... and 25

Unnecessary Roughness, Defense, #58: Stomping the nuts.

3
by B (not verified) :: Thu, 09/21/2006 - 2:30pm

I feel that judging whether the committee or single back approach is successful by the number of wins is imprecise. Looking at the average YPC I would think is a better way. Obviously the best solution is to look at DPAR of the top back vs that of the backs in the committee, but that would limit your sample size.

4
by CA (not verified) :: Thu, 09/21/2006 - 2:31pm

The 1983 St. Louis NFL team was the Cardinals, not the Rams.

5
by navin (not verified) :: Thu, 09/21/2006 - 2:33pm

He means the LA Rams.

6
by Bill Barnwell :: Thu, 09/21/2006 - 2:35pm

Correct. Los Angeles. It seems so long ago...

7
by Dennis (not verified) :: Thu, 09/21/2006 - 2:38pm

I think you asked the right question but went about finding the answer the wrong way. The question is if teams with RBCs have more effective rushing attacks or better overall offennes than teams that use featured backs. Comparing teams with similar rushing stats is really pointless and the result should be that they have a similar number of wins. I'm sure if your sample size was larger the average wins would be even closer.

8
by NoraDaddy (not verified) :: Thu, 09/21/2006 - 2:39pm

The one thing that the committe aproach has over the single back is injury insurance. If your single back goes down, so might your season. Also, having one back doing most of the work may be more apt to break down and get injured more so than the committee backs.

9
by ChrisFromNJ (not verified) :: Thu, 09/21/2006 - 2:40pm

I wonder what the numbers would be if you lowered the requirement to, say, 70 percent as many carries for the second back- 90 percent seems both arbitrary and unreasonably high.

For example, Tatum Bell only got about 72 percent as many carries as Mike Anderson in Denver last year, and everyone still called that a "committee". Perhaps the larger sample would provide more conclusive findings?

10
by ammek (not verified) :: Thu, 09/21/2006 - 2:43pm

When exactly do teams use a committee? When they have two good rushers (one of them usually a rookie) and want to split the carries (this year's Saints, perhaps); when they have two different types of rusher (Csonka and Morris, Thunder and Lightning); or when they don't have anyone good enough to carry the load (93 Packers). In other words, it's generally a solution to a problem, rather than an offensive philosophy.

Prior to about 1982, this wasn't so, as the traditional Harris-and-Bleier, Taylor-and-Hornung black-and-blue offense was still in full effect, as your chart shows (14 of 41 committees in four years, 1978-81, versus 3 in seven years from 1993-99).

And I find the correlation with wins bizarre and misleading. Wouldn't it have been better to compare teams who were similar in, say, passing offense and/or defense to show what difference (if any) the committee approach makes?

11
by CA (not verified) :: Thu, 09/21/2006 - 2:44pm

Re: 5, 6

Ottis Anderson played for the St. Louis Cardinals. I assume the 1983 comparison is between the San Francisco 49ers and the St. Louis Cardinals.

12
by Kachunk (not verified) :: Thu, 09/21/2006 - 2:57pm

I think you might get more meaningful results if you just looked for teams with similar numbers of rushing attempts, then look at the number of yards gained. Still pretty imprecise, but better than a sharp stick in the eye.

13
by Josh (not verified) :: Thu, 09/21/2006 - 3:01pm

No big surprise that similar production from an RB committee to a team with one clear starting RB produces similar results. But still interesting.
Maybe the more interesting thing is that teams that do RB by committee include both good teams and bad teams, though recently the good have far outnumbered the bad

14
by Rich Conley (not verified) :: Thu, 09/21/2006 - 3:03pm

TO be running back by committee, the backup has to have 90% of the carries as the starter? Thats way too restrictive.

I'd say anything where a single back doesnt have more than like 60% of the carries is RBBC.

15
by Josh (not verified) :: Thu, 09/21/2006 - 3:03pm

#2: you've got a 15 yard penalty on the D, that doesn't make it 1st and 25, it makes it 1st and 10

16
by Ian (not verified) :: Thu, 09/21/2006 - 3:04pm

Um... I don't get why you look for other teams that had similar rushing seasons to the "commitee" teams. How does that help you judge the relative quality of commitees? Shouldn't you just look at how commitees compare to non-commitees?

Ian

17
by Rich Conley (not verified) :: Thu, 09/21/2006 - 3:25pm

These articles just seem like hes trying to make some statistical claim off of the results of small sample size.

Ian is right, if you want to compare them, you have to do it across the board.

18
by ToxikFetus (not verified) :: Thu, 09/21/2006 - 3:28pm

15:
Whoops... The lesson here, as always, is that I'm an idiot.

Now, for something germane to this XP, it's worth noting that RB #1 in the 1988 PHI committe was none other than Randall "Ultimate Weapon" Cunningham. Did any other of the teams listed have a QB in their committe?

19
by DWL (not verified) :: Thu, 09/21/2006 - 3:31pm

Regrettably, most of the comments regarding the methodology of this article are correct; this article hardly seems to pass muster for FO. Over-restrictive selection criteria, matching for rushing performance (would have been a better dependent variable) and using wins as dependent variable inevitably resulted in a conclusion that getting rushing yards is associated with winning – a point that is made by almost every clueless pundit, but in a manner that is not understood by said clueless pundit as frequently pointed out on FO (run late to win).

Consider this article as data point number one for possible cause for concern that FO staff while gaining much well deserved fame and notoriety, may be getting stretched thin by the multiple demands? Hopefully this is not the case, b/c I eagerly devour every FO piece I can find and now that the season is here, the long dry spell of summer is over, with some days feeling like I can barely keep my head above the rising water of excellent FO articles.

20
by Jay (not verified) :: Thu, 09/21/2006 - 3:31pm

Why not take your "buddies" and then compare their results in the following year? That would seem to be more useful so that we can talk about wear and tear on backs and have another set of data.

21
by Bill Barnwell :: Thu, 09/21/2006 - 3:38pm

Few comments:

#18 - Nope. Cunningham was 3rd in rushes for the '88 Eagles.

The Rams thing was my screwup. The data's right, but I grabbed the wrong guy's stats off of pro-football-reference when I was focusing in on Ottis. It was actually Eric Dickerson.

The article was analyzing the aspect of how to employ running backs from a management standpoint as opposed to a purely statistical one; in that sense, the purpose of the article wasn't to compare teams with similar passing offenses, for example, to see if their rushing attack was better than the other. In that sense, which way do you guys think would be the best way to analyze RB performance with/without committees without using DVOA/DPAR?

I would also argue that 40+ teams is a decent sample.

I also played around with the different thresholds for backups and found that 90% worked best when it came to the groups of players that were being included along with the size of the sample for my own sake.

22
by Ilanin (not verified) :: Thu, 09/21/2006 - 3:52pm

18, 21 - of course, the 2006 Falcons look like their plan involves the QB as part of the RB committee, and it does also seem to be working.

23
by Ima Pseudonym (not verified) :: Thu, 09/21/2006 - 3:59pm

Re 19: I agree. Last week I worried that the FO on BSMW column was diluting the FO brand. These columns just don't seem on par with the rest of the FO material (sorry, Bill - nothing personal).

Since it seems that there are regularly very good suggestions in the comment threads about how statistics could have been used to much greater effect in connection with what you are discussing, why not link a draft of a future column here, and let some of the knowledgeable FO readers (among whom I certainly would not include myself) give you some feedback in time to make changes before the column actually appears.

24
by Dennis (not verified) :: Thu, 09/21/2006 - 4:04pm

Re 21: The article was analyzing the aspect of how to employ running backs from a management standpoint as opposed to a purely statistical one

Then I really don't know what you were going for, other than the salary stuff at the end. From a management standpoint you want to know the best way to maximize your rushing ability (however you want to measure that). Comparing teams that had the same rushing ability doesn't give any insight into that.

25
by B (not verified) :: Thu, 09/21/2006 - 4:07pm

21: Sticking with the "buddy" system, I would compare two teams that had similar total passing yards, wins and rushing attempts, and see which of the two had the higher YPC and rushing touchdowns. That way you're accounting for varying difficulties such as protecting a lead. Also, I would compare just the stats of the star running back vs the main components of the committee.

26
by Bill Barnwell :: Thu, 09/21/2006 - 4:16pm

Pseudonym - Of course, I respectfully disagree to an extent, but I wouldn't dream of putting myself on the same level as an Aaron or as one of the Mike's. I don't think the draft idea would happen, but I'm certainly going to be incorporating ideas into future work.

Dennis - the wins/rushing correlation is the over-arching concept that teams can achieve equal levels of success by developing several running backs as opposed to signing one star RB. Neither is more likely to produce a winner.

27
by Dennis (not verified) :: Thu, 09/21/2006 - 4:45pm

Bill, IMO it was pretty intuitive that teams with similar rushing totals would average around the same number of wins, so I don't see where any new ground was really broken.

The question facing a GM in the offseason is whether one back or a committee will be more effective. Comparing teams where they were equally effective doesn't help answer this question.

28
by Rich Conley (not verified) :: Thu, 09/21/2006 - 4:46pm

Bill, but I dont think you establish that. You took a bunch of teams that have the same rushing stats, and then looked at their records. I'd be willing to guess that either A) defence, or B)Passing offence Correlates to how those teams did much more than whether they were RBBC.

The best way, IMO, would be to compare all single back teams over the last 5 years against all RBBC teams in the last 5 years. And look at wins/losses/dvoa/whatever.

I just think by normalizing rushing yards, you're invalidating what you're trying to prove.

29
by Bill Barnwell :: Thu, 09/21/2006 - 5:08pm

Rich,

I may followup with that piece at some point later in the season, although it'd be a stretch to use "running back by committee" for a bunch of teams in the modern era who didn't really use one.

Of course, the real solution is to get the play-by-play database running back 20 or 30 years so that we can have DVOA over a larger timeframe. One day in 2012...

30
by jason (not verified) :: Thu, 09/21/2006 - 5:36pm

wouldn't it also make sense to compare the injury rates of players in rbbc vs. "one-horse" teams? It would be interesting to see an attempt to quantify the increase/decrease in injury risk between featured backs vs the components of an rbbc committee

31
by Mitch Cumstein (not verified) :: Thu, 09/21/2006 - 5:45pm

First, love the title.

Second, I was confused after reading the article and felt I didn't get much out of it. On a counterpoint, though, I did like the salary info at the bottom.

Here are some thoughts, of course, all of these may be limited by practical considerations.

I think there are different meanings to RBBC. The one I would be interested in is when a team, on a per game basis, actually has a relatively close distribution between RB1 and RB2 touches within each game--the best current example being the Patriots through the first two games.

Another is where a team switches the role of players because they stink, or because of injury. For example, the Jets may decide Barlow stinks after giving him several starts of the bulk of the carries, then decide to give Houston the bulk of the carries in a game, etc., and at the end of the year, you have a bunch of guys with similar amount of carries.

I would be more interested in the first category. I would endorse that this could be done along the lines of what Rich Conley suggests, looking at most recent 5 years and all teams.

You could look at individual game ratios of RB touches (rushes + rec) between RB1 and RB2 for each game, and have a team average for the season. You could then sort the teams into three groups: RB1 heavy, medium, or RB1/RB2 balanced, and compare those groups for things such as YPC, Wins, Year 2 Performance (to see whether wear and tear had greater effect on RB1 heavy teams or not).

There would be some biases here. Teams that run the ball alot because they are leading (and I mean alot!) would tend closer to medium to committee appearances, because the RB1 can only carry the load so much, very few RB's have averaged over 24 carries per game over a full season recently. San Diego, for example, uses Tomlinson to a large degree, but Turner has gotten alot of touches late because of substantial blowouts. But overall, I think it could be more informative.

32
by Mitch Cumstein (not verified) :: Thu, 09/21/2006 - 6:27pm

Just to follow up on what I posted above, here are the teams whose top RB's have averaged 90% or more of the rush att compared to RB2, through 2 games of the 2006 season:

SF-96%
ARI-96%
CIN-95%
PIT-93%
MIA-91%
MIN-90%
DET-90%
JAC-90%

And the RBBC, where the ratio of RB1 to RB2 carries through 2 games is at or below 65%:

NO-63%
WAS-62%
TEN-62%
IND-60%
DEN-57%
NE-54%

33
by Larry (not verified) :: Thu, 09/21/2006 - 6:40pm

The buddy analysis seems designed to completely obscure any difference between the RBBC and Stud approaches. The only way it might not is if having backs with specific roles allows you to gain yards more effectively tuned to the situation, but that seems likely to be a very small effect and not the way that RBBC is supposed to give advantage. I'm curious how Bill James ever used this anlysis concept, to be honest.

34
by Subrata Sircar (not verified) :: Thu, 09/21/2006 - 9:15pm

My take on the article is that you essentially provided evidence that teams with similar rushing attacks have similar win totals. I think that this is due to the vagueness of the question: "do teams that use multiple running backs perform better than teams that don't?"

What do you mean by "perform"? In no particular order:
a. Wins are so dependant on everything else that it seems foolish to use those.
b. Rushing yardage is easily measured, but doesn't seem to have a good correlation to winning.
c. Rushing points has the same problem.
d. Overall offensive efficiency (points/possession? 1st downs/possession?) might be interesting.
e. Total rushing DVOA might be interesting as well.
f. DVOA/running-back-salary i.e. "marginal DVOA" might be most interesting of all.

Anyway, frame this as a hypothesis, and it becomes much easier to understand:

"Teams that employ multiple running backs can get equivalent rushing-contributions-toward-winning for a lower total cost, than teams that employ a single running back for the bulk of their carries."

Define "bulk" at your 70/90 percent thresholds, explain how you're measuring rushing-contributions-toward-winning i.e. DVOA, rushing yardage/touchdowns, etc. and then look at your data.

35
by Peter (not verified) :: Thu, 09/21/2006 - 11:42pm

25: You shouldn't exclude the 10% backup from the stats... while the 50% RBs would presumably get 100% of the carries, you're losing carries and ignoring part of the single back situation by not including them. If you have to resort to a crappy back to spell your workhorse, it's worth noting whether that helps (because they're super-fresh) or hurts (because they're really bad).

Not to pile on, but I also wasn't sure what the point of this article was at the end. I did like the salary section; I was very surprised to learn that RBBC is more expensive.

It seems to me that a simple way to preliminary compare RBBC to workhourse would just be a big graph... the independent variable would be % of carries to RB1, so this year, it would go from 96% San Fran down to 54% NE. The dependent would ideally be DVOA or DPAR, but since that's not possible, maybe yards per carry.

Of course the issue with this is injury, where a team that wished to go 90/10 now goes 50/50 because it's 2nd and 3rd stringers are equally good (or bad). I'm not sure how easy it would be to correct for that.

36
by Fred Taylor\\\'s groin (not verified) :: Fri, 09/22/2006 - 1:10am

35:
To account for your injury problem, maybe you see who was the RB1 for the first 6 games, then only calculate the ratio for any games where he carried at least once, and classify teams by that number. Maybe that is way too difficult to set up, given the databases, I have no idea. But it would then correctly classify (as 90/10) teams that went 90/10 for a few weeks, but were forced to 50/50 by an injury and equally sucky backups.

37
by J (not verified) :: Fri, 09/22/2006 - 7:47am

I agree with #8 NoraDaddy. The RB-by-committee approach is a result of teams limiting their risks due to injury.

RBs are the most likely to be injured postion. OL (behind QB and WR) are the 3rd least likely to be injured. Therefore, investing in the O-line with mulitple RBs seems to be the way to go. Certainly, single-back teams have had success as well, but their success depends on the health of their single-back.

What would SD's year end up like if LT goes down for the year or Alexander with Seattle? Worse yet, what if Alexander had an injury bad enough that he was never the same RB - Seattle's salary cap health would be hurting...due to his relatively large signing bonus.

If I were a GM in the NFL, I would not have a single-back - esp. not a highly paid one.

This, IMO, is one of the reasons the Texans passed on Bush. Bush is a unique case though, if managed properly (11-14 carries/game with 6-8 passes thrown his way per game -which the Saints seem to be doing) should help reduce his risk of injury and prolong his career...Bush - managed properly - would be the exception.

38
by DavidH (not verified) :: Fri, 09/22/2006 - 10:42am

But if you only hand it to Bush 11-14 times per game, you need a 2nd back. In which case you're not going to want to pay him "Reggie Bush money." I'm not sure I get how he is an exception here.

39
by Tracy (not verified) :: Fri, 09/22/2006 - 11:47am

I think this analysis verifies an assumption that most of us probably already held: Rushing performance matters more than the way in which it is achieved. It's always good to check your assumptions, but the bulk of the complaints here seem to be of the "tell me something I don't know" variety,

I was hoping to read that one use pattern can be expected to produce more effectively than the other. But the more I think about it, the harder it seems to set up that experiement. The only method I can come up with to answer that question is to find teams that used a featured back in one season and used that same back as part of a committe in another. Any other approach would fail to control for the quality of the running backs and offensive lines being compared. But I suspect that my approach will yield too small a sample to analyze. Maybe you realized this, too, Bill, and that's why you didn't go down that road.

40
by Mitch Cumstein (not verified) :: Fri, 09/22/2006 - 12:42pm

RBs are the most likely to be injured postion. OL (behind QB and WR) are the 3rd least likely to be injured. Therefore, investing in the O-line with mulitple RBs seems to be the way to go.

What is your source for these claims? I am skeptical.

41
by LeCharles Bentley, Steve Smith, and Trent Green (not verified) :: Fri, 09/22/2006 - 12:43pm

We are too.

42
by Mitch Cumstein (not verified) :: Fri, 09/22/2006 - 1:14pm

In another way of arriving at a similar conclusion, I used the pro-football-reference.com 2005 games database (link to it in my name) to sort the teams by RB1/RB2 per game ratios.

I dont have the data in front of me (maybe I can post later if anyone wants to see the specific team breakdowns), but there was virtually no correlation between the percentage of carries the top running back got relative to the second one, and wins.

The methodology used was a bit different. I didn't sort by specific running backs, I just listed the RB with most rushing attempts as RB1 each game, and the RB with 2nd most attempts as RB2, averaged 16 games for each team, and got the percentages. Thus, if Travis Henry got 2/3 of the carries in one game, and Chris Brown got 2/3 of the carries in another, each was listed as RB1 in the respective game.

It seems to me the RB usage ratios is part philosophical, and part dictated by personnel based. For example, New Orleans went from a heavy RB1 team to a balanced RB1/RB2 team based on McAllister's injury, while teams like Houston, Oakland, and Green Bay really did not change their stripes, even though the identity of personnel playing the part of RB1 and RB2 did.

And the ratios went the other way. KC went from a balanced team (around 64% RB1) for the first 7 games, to a very skewed RB1 (around 96%) in the games after Holmes' injury. This is what I would expect would happen (maybe not to that degree) if either Dillon or Maroney missed some games to injury and the other was healthy. Denver, on the other hand, did not change its pattern, nor did a team like Dallas or Jacksonville, even though personnel making up RB1/RB2 changed due to injury or other factors.

What would be interesting is to take the 2005 list, and cross reference it to 2005 rb salaries. Then we could look at RB1 Heavy/High Salary; RB1 Heavy/Low Salary; RB1/RB2 Balanced/High Salary; and RB1/RB2 Balanced/Low Salary. Maybe there would be some info there.

I was thinking one of the reasons the similar producing committees had higher salaries than single backs was because, philosophically, they were not committee teams, but after overpaying for a player, they realized that the player was limited and it became a committee. Just a thought, because if someone had a philosophy of RBBC (Shanahan?) they would not over-extend their resources for players at that position.

43
by Bill Barnwell :: Sat, 09/23/2006 - 10:50am

Here's the thing - comparing similar passing attacks to determine whether a RBBC is better than a lone RB doesn't really make sense, since the lone RB could very easily be worse than one or both of the backs in the RBBC. We're not comparing, then, similar things anymore, and the comparisons aren't valid.