This interesting thread over on the message boards (which gave me the inspiration to finally crunch the numbers for this article -- something I've been meaning to do for a long time) points out a large problem with the above. Namely, there is a whole lot of statistical noise in a given player's YPR. It is clear that a better back will gain more yards with the same number of carries all else equal, but the problem is that all else is never equal. In this case, it may be so far from equal that whatever YPR is trying to tell us about talent is completely lost in the noise. It may be that YPR is just too distorted to tell us anything meaningful. This article will examine whether or not that is true.

The first step is to understand what factors, besides the player's talent, are influencing a player's YPR. These are the factors we'd like to get rid of if possible. Two immediate possibilities come to mind:

1. Surrounding offensive talent: I don't think anyone is going to dispute that an RBs numbers are affected by the quality of the five fat guys in front of him. The quarterback and receivers may have some impact as well, but that's less clear.

2. The situations in which the carries take place: if an RB runs mostly draws on 3rd-and-long, when the defense is more than happy to give up 6 or 7 yards, he figures to post a high YPR. If he runs mostly in short-yardage situations, when the whole world knows the plunge is coming, his YPR is going to be lower.

[There is actually one more factor that I believe has a huge impact on YPR, but I'll save that for the end.]

This article will focus primarily on the second item, as I frankly have no idea how to deal with the first. To get a feel for how I'm going to handle things, let's concentrate on two runners from opposite ends of the spectrum: Shaun Alexander and Michael Bennett. Alexander got his team's full allotment of short yardage work, while Bennett was often yanked in short yardage situations. Bennett's YPR (5.1) was also a lot higher than Alexander's (4.0). How much of that was due to the differing contexts?

Let's first consider goal line carries. The simplest way to account for the difference in that department is to just take them out. Alexander had 32 carries inside the opponent's 10 yard line last year, while Bennett had only 17. Let's forget about them -- pretend they never happened and recompute the YPRs. Here's what you get:

YPR with goal line carries removed

Alexander's YPR improves by .26 without the goal line work, but Bennett's jumps by just as much (more, actually). Let's break this down a little bit further. Here are the league averages for 2001-2002:

About 10.9% of Alexander's carries came inside the 10 last year while only 6.7% of Bennett's carries did. Put that together with the above table, and it amounts to this:

Given the allotment of carries that Shaun Alexander had last year, an average RB would have had a 4.02 YPR. Given the allotment of carries that Michael Bennett had last year, an average RB would have had a 4.13 YPR. Based on this, I think it makes sense to say that, of the difference in their YPRs last year, only about .1 of that was due to the fact that Alexander got more goal line carries.

If you look at carries inside the 20 rather than inside the 10, you get similar results.

Here is a list of all backs with at least 200 carries from last year, listed in order of how tough their carries were. For example, at the top of the list is Tiki Barber. The LGAVG column tells you that a league average runner would have gotten 4.00 yards per rush with Tiki's carries. The other columns tell you that Tiki actually got 4.56 for a difference of +.56.

 

Player	LGAVG	YPR	DIFF
Tiki Barber	4.00	4.56	0.56
Warrick Dunn	4.02	4.03	0.01
Shaun Alexander	4.02	3.98	-0.04
Stephen Davis	4.04	3.96	-0.08
Garrison Hearst	4.05	4.52	0.47
Priest Holmes	4.05	5.16	1.11
Travis Henry	4.05	4.42	0.37
Eddie George	4.06	3.40	-0.66
Clinton Portis	4.06	5.52	1.46
Ricky Williams	4.08	4.84	0.76
LaDainian Tomlinson	4.08	4.52	0.44
Deuce McAllister	4.10	4.27	0.17
Corey Dillon	4.10	4.18	0.07
Marshall Faulk	4.10	4.50	0.39
Antowain Smith	4.11	3.90	-0.21
Duce Staley	4.12	3.83	-0.29
James Stewart	4.12	4.42	0.30
Jamal Lewis	4.12	4.31	0.18
Michael Bennett	4.13	5.08	0.95
Ahman Green	4.13	4.34	0.21
Edgerrin James	4.16	3.57	-0.59
William Green	4.16	3.65	-0.51
Fred Taylor	4.17	4.58	0.41
Anthony Thomas	4.17	3.37	-0.80
Curtis Martin	4.17	4.19	0.02
Lamar Smith	4.18	3.53	-0.65
Michael Pittman	4.20	3.52	-0.68
Emmitt Smith	4.21	3.84	-0.37

So it appears that goal line carries account for at most a difference of about .2 in YPR. At least that's true among full time backs -- guys like Zack Crockett who get nothing but goal line carries will show more extreme results.

But, as always, there's more to it than that. Much more. Not all carries inside the 10 are really short yardage situations and, more importantly, not all short yardage situations are inside the 10. In addition to those 32 carries Alexander got inside the 10, there were also several carries on 3rd-and-one and 4th-and-one and 3rd-and-two and so on. How much difference do those make?

We'll do the same thing, but this time we'll break down each runner's carries into down-and-distance groupings. Here are the groupings, along with the percentage of all running plays that were of that type, and the league average YPR for each:

 

Situation	% of run plays this type	YPR (lg avg)
1st and 0-2	0.9%	0.56
1st and 3-7	2.5%	2.34
1st and 8+	51.1%	4.28
2nd and 0-2	5.8%	3.40
2nd and 3-7	14.4%	3.86
2nd and 8+	14.8%	4.48
3rd/4th and 0-2	6.0%	3.39
3rd/4th and 3-7	2.5%	4.91
3rd/4th and 8+	1.8%	5.38

If you add the 1st-and-shorts to the 2nd- and 3rd-and-shorts, you find that 12.7% of all rushing attempts last year could fairly be called short yardage attempts. About 15.6% of Shaun Alexander's carries were for short yardage, while only 5.9% of Bennett's were.

We'll play the same game: figure, for each back, what kind of YPR a league average RB would have compiled with the same collection of carries, and compare it to the player's actual YPR. Below you'll find a chart that contains the same info as the one two charts up. The runner who faced the toughest down-and-distance circumstances last year was Tiki Barber. The easiest belonged to Michael Pittman.

 

Player	LGAVG	YPR	DIFF
Tiki Barber	3.99	4.56	0.57
Clinton Portis	4.00	5.52	1.52
Stephen Davis	4.01	3.96	-0.04
Warrick Dunn	4.02	4.03	0.01
Ricky Williams	4.03	4.84	0.81
Antowain Smith	4.05	3.90	-0.15
Priest Holmes	4.06	5.16	1.10
LaDainian Tomlinson	4.06	4.52	0.46
Travis Henry	4.07	4.42	0.36
Duce Staley	4.07	3.83	-0.25
William Green	4.08	3.98	-0.10
Shaun Alexander	4.08	3.98	-0.10
Eddie George	4.09	3.40	-0.69
Anthony Thomas	4.10	3.37	-0.73
Michael Bennett	4.10	5.08	0.98
Marshall Faulk	4.10	4.50	0.39
Jamal Lewis	4.12	4.31	0.19
Ahman Green	4.12	4.34	0.21
Corey Dillon	4.13	4.18	0.05
Edgerrin James	4.13	3.57	-0.56
Emmitt Smith	4.14	3.84	-0.30
James Stewart	4.15	4.42	0.27
Garrison Hearst	4.15	4.52	0.37
Deuce McAllister	4.15	4.27	0.12
Curtis Martin	4.19	4.19	0.00
Lamar Smith	4.19	3.53	-0.66
Fred Taylor	4.19	4.58	0.38
Michael Pittman	4.21	3.52	-0.69

The general conclusions are similar: among full time runners, the differing variety of situations faced by each seems to account for at most a difference of .2 in their YPRs. And remember, that's the difference between the top and bottom of the list. Most full time runners are separated by virtually nothing.

We're still not capturing everything here. Third-and-7 counts the same as third-and-3 according to this scheme, and it's possible that Bennett had a great deal more of the former while Alexander had a great deal more of the latter. Possible. And there are other circumstances, like offensive formation or score at the time or playing surface, that might affect a runner's YPR. But my guess is that short-yardage situations and goal line carries are the biggest issues. I've seen enough here to convince me that, except in extreme cases, two featured backs will, over the course of a full season, see fairly similar distributions of carries. And even in the extreme cases, it doesn't appear to make too much difference in the YPR.

One more quick item to point out: there is no such thing as a featured back who fattens up his YPR by running a lot of draws on third-and-long. Of the backs who carried the ball over 200 times last year, none of them had more than 5% of their carries on third-and-3+.

At this point you might be expecting me to conclude that YPR is indeed a decent barometer of running skill -- that the talent really does, to some extent, show through the noise. Not quite. There's still the offensive line issue. But even setting that aside, there's another factor that I believe causes massive distortions to YPR.

Luck.

I have some mildly scientific backup for saying this, but I'll first offer an intuitive argument.

A half a yard is a pretty big difference in YPR. But what does it take to add or subtract half a yard. For a 300-attempt featured back, that's 150 yards over the course of a season. Sometimes -- not all the time, but sometimes -- sheer dumb blind luck can be the difference between a zero-yard gain and a 40-yard gain. Or between a 20-yard gain and a 60-yard gain. If that happens four times in a season -- just once every four games -- you've added .5 to your YPR. It seems easily believable to me that luck can swing fortunes even more than that.

Think about it another way: if you could rewind life and replay the 2002 season a thousand times under the same conditions, Michael Bennett wouldn't have a 5.1 YPR every time. Sometimes, he'd have a 4.7, sometimes he'd have a 5.3, and so on. Which do you think is more likely, a 6.1 or a 4.1? A 5.7 or a 4.5? If you're answering with the lower number each time (as I am), then you're agreeing that Bennett got some luck last year. How much? .1? .9? I have no idea, but I'm confident he got some.

Here's the more scientific argument: if you look at all RBs for the last 20 years who have gotten 200 carries for the same team in consecutive years, you'll find that the vast majority of the time, RBs who have a significantly higher than average YPR see it return toward average the next year. If you run a regression on it -- essentially telling the computer to find the best formula for predicting next year's YPR from this year's -- you'll find that the "best" predictor for next year's YPR is 4 parts last year's YPR plus 6 parts league average YPR. In other words, it's typical for a good RB to see his YPR move more than half way back toward average. To give a concrete example, even though Clinton Portis's YPR was an otherworldly 5.5 last year, a fair over/under for this year would probably be closer to 4.7.

On the other hand, it is undeniably clear that YPR is not random. This year's YPR is a useful predictor of next year's YPR. So I think that YPR does capture some information about a runner's talent.

We end this article where we started it: asking whether or not YPR matters. In the end, I can't see how it's going to help the fantasy player much. YPR is driven by big plays. Big plays happen relatively infrequently and are subject to a lot of random variation, much like touchdowns. When a running back has an abnormally high YPR for a year, it's very likely that it'll come back down to normal. So I do not advise using YPR to try to uncover hidden gems. The YPR of any gem with few enough carries to still be hidden is based on too small a sample for me to have any confidence that it's legit.