Behind the Numbers - Rushing Stats

Submitted by CollegeFootball13 on September 19th, 2009 at 11:52 PM
There are a lot of statistics floating around the world of College Football. Some intriguing, some important, some useless, some hilarious. Behind the Numbers is a semi-regular look at just a few of the statistics that you may or may not be aware of, with a little bit of a dissection of each. Enjoy.

Today's Focus: Rushing Stats (Rush YPG)

(Full NCAA Rankings)

Players of note: Ralph Bolden, Purdue (1st, 178.5ypg); Jahvid Best, Cal (6th, 140.5ypg); Armando Allen, ND (23rd, 105.5ypg); Caulton Ray, MSU (99th, 61ypg)

Why It's Important:

Because.. it tells you how many rushing yards a player has per game, on average. Pretty self-explanatory here. Generally the more yards a player rushes for per game, the better they are.

Why It's Flawed:

It just measures yards. A big, bruising back that gets the ball on third and short situations or inside the ten yard line can be just as valuable as a quick running back who gets big yards but can't break tackles. What would you rather a running back's stat line be -- 6att, 25yds and 3 TDs or 30att, 200yds and no scores? One gets you points, the other gets you valuable field position that can turn into yards.

Also, it doesn't take into account the number of rushing attempts. YPC does this, but you'd have to look into two or three different stat lines to really see the effectiveness of a RB.

ALSO, it doesn't take into account fumbles. 200yds in a game is all well and good, but if all that field position is wasted because he fumbled 3-4 times, it doesn't help at all.

So any one stat for a RB will be leaving out a lot of the story.

Applying this to Current Statistics

Ralph Bolden, Purdue: 178.5ypg (#1)
Definitely a great YPG average, good enough to be #1 in the nation after two games, but a look at his YPC tells a different story. Bolden averages 7.14ypc, still a respectable number, but not nearly #1 in the nation. In fact, second through sixth leading rushers in terms of YPG have a higher YPC average than Bolden. His 50 carries are the second highest in the top 10.

It's pretty obvious that between two equally talented rushers that have the same YPC average, whoever gets more carries per game will have the higher YPG average. Hence the flaw.

Robert Turbin, Utah St: 148.0ypg (#4)
Obviously an extremely small sample size here, as Turbin has only played one game so far (Utah), but he's listed here for another reason. That 148yds was garnered on only 13 carries, earning him a 11.38ypc average, the best of anyone in the Top 25 of YPG.

Reggie Arnold, Arkansas St:
104.5ypg (#25)
Arnold, while not dominant in either YPG or YPC (8.04), is extremely efficient in terms of points earned with his carries. He's had 26 carries thus far, and has scored 5 touchdowns. Almost 20% of the time this guy's had his hands on the ball out of the backfield he's been in the endzone.

So three different stat lines, all pretty damn good in their own way.

An Alternative

Along the lines of my Quarterback Efficiency Rating, I've come up with a Rushing Efficiency Rating (RER). It's much more than YPC or YPG, it's a combination of the major aspects of a running back's game that is contributes to their overall efficiency.

Here's the first draft of the formula:

                           (Yards) + (Touchdowns x 10) + (Fumbles x -10)
                                                          Attempts

So a big bruiser who might not rack up 8-9ypc but is solid with ball control and in the red zone who's usually good for a few scores:

10att, 40yds, 3 TDs (RER: 7.00)

Has an RER that's similar to a speed back who might rack up the yards, but is prone to a mistake here and there and might not always get the ball on the goal line:

28att, 170yds, 2 TDs, 1 Fumble (RER: 6.79)

Applying the RER to Last Season's Backs

Rank (YPG)
(YPC)
Player
YR
TDs
YPC
YPG
RER*
Rank
17
Donald Brown, ConnecticutJR18 5.68 160.236.17
6
25
Shonn Greene, IowaJR20 6.03 142.316.68
5
31
Jahvid Best, CaliforniaSO15 8.14 131.678.92
1
410
Javon Ringer, Michigan St.SR22 4.20 125.924.76
10
58
MiQuale Lewis, Ball St.JR22 5.39 124.006.07
8
66
Chris Wells, Ohio St.JR8 5.78 119.706.17
6
72
Kendall Hunter, Oklahoma St.SO16 6.45 119.627.12
2
83
Vai Taua, NevadaSO15 6.44 117.007.08
3
94
Tyrell Fenroy, La.-LafayetteSR19 6.08 114.586.92
4
109
LeSean McCoy, PittsburghSO21 4.83 114.465.51
9
*HUGE difference is made in the RER because I couldn't find fumble statistics. Any help there would be greatly appreciated and the RERs would be updated for more accuracy.

Quite the shakeup in the YPG rankings when the number of carries is taken into account, as well as the number of touchdowns. YPC numbers, on the other hand, are nearly identical. If the fumbles were taken into account, this would surely be a bit different, but until I can find those stats this is all we have to go by.

Thoughts? Comments? Fumble statistics? Let me know.

___________________
Behind the Numbers will be back soon with another look at a stat from the world of College Football. Any stats you want to be examined a little closer? Or even just a stat you've been interested in for a long time? Let me know in the comments and I'll do my best to get to it in the next few installments of BtN. Thanks for reading!
-CollegeFootball13

Comments

MaizeNBlu628

September 20th, 2009 at 12:03 AM ^

nice entry! just wondering, why is the multiplier 10 for both touchdown and fumbles (-10)? So in the equation you are suggesting that a fumble is as bad (value-wise) for the team, as a TD right? Also, is it just lost fumbles or total?

CollegeFootball13

September 20th, 2009 at 12:11 AM ^

In the equation, yes. Touchdowns are equally as valuable to your team as a fumble would hurt them. I was kind of on the fence about this, but so far I don't even have the fumble statistics so it doesn't really matter haha.

And I was thinking lost fumbles would be all that matters, because while all fumbles are dangerous, only the lost ones would actually hurt your team.

MCalibur

September 20th, 2009 at 12:32 AM ^

Just like Tebow v.McCoy can't be really settled without accounting for rushing stats (McCoy wins, btw), a RB's value must include his receiving contributions. Ultimately things would condense into a Playmaker Efficiency Rating or whatever. Relevant stats in the numerator with Passing Atts + Rushing Atts + Receptions in the denominator.

Ideally Receiving Targets should be used for the denominator but that's not a standard stat and it'd be impossible for a single person to take on a project like that. I think Football Outsiders is starting to chart college games but I'm not sure about that. My only other beef is with valuing turnovers on par with TDs; seems a bit extreme to me, maybe that Alton guy can give an assist here.

A Playmaker efficiency rating would be interesting and probably pretty fair b/c each position would collapse into its own relevant stat line while still being universally applicable. It's like the Physics Theory of Everything. Since the rating is based on the each player's specific involvement it would show relative impressiveness for each player relative to all players (who accumulate offensive stats).

Of course real player value needs to include subjective criteria like relative competition, big/rivalry game 'show-up-itude', and so on. But good luck establishing fair value for those.

MCalibur

September 20th, 2009 at 1:11 AM ^

Yeah, figured you were sand bagging a little. So, while you're in draft mode, here are a few more thoughts:

-The final draft for the playmaker rating should consider 'positive rushes' in the numerator as distinguished from total attempts. This would be analogous to completed passes vs. attempted passes and completed receptions vs. attempted receptions (aka targets). This difference is key to the Barry Sanders vs. Emmitt Smith debate (also Detroit's shitty offensive line play).

-How about including special teams play? Desmond Howard would never had won the Heisman without his contributions in the third phase. I would propose a 'yard is a yard' approach here since ultimately field position and scoring is what matters regardless of what game phase you're in. This is where Devin Hester's value was locked (until he became a mediocre receiver).

-I bet we could do pull off a charting project to get some of these non-standard stats such as fumbles, targets, 'positive rushes' (that category needs a better name). We could focus on the BCS conferences. I'd limit it there since its fairly self evident that stats accumulated as a matter of intra-league play in the MAC, MWC, Sunbelt, and C-USA are a little light in the pants compared to what Tebow, McCoy, & Co. have to do. Anyway, its a little daunting but not as bad as it seems at first blush. I single handedly put five seasons worth of data for the QB injury project I did together in less than 20 hrs. We can trade emails if you're interested ([email protected]).

Drederick Tatum

September 20th, 2009 at 3:30 AM ^

No, they're not in any top ten rushing statistic, especially when looking at last year - but in the future it would be nice to see where our running backs stack up with the best. Great post though!

matty blue

September 20th, 2009 at 8:11 AM ^

...but it seems awfully arbitrary to me. where do the 10x factors come from? what made you choose those numbers? i can see what you're doing, of course, but what if the fumbles were only a 5x factor? or a 15? seems like they could be, if you choose - and that would significantly change your results.

the best "new" stats (runs created, vorp, etc.) have some underpinnings that relate them back to the ultimate goal of wins and losses, in some demonstrable way...i'd like to see more of the background here.

teldar

September 21st, 2009 at 8:05 PM ^

I was thinking the same thing about the multipliers. Rather than tdx10, I would look at # of points. This would add for the 2 point conversion on a running play as well as the td. However, I suppose that's taking away from the runner. The running isn't any less hard because it's only for 2 instead of 6. But whatever.

bronxblue

September 20th, 2009 at 8:42 AM ^

Interesting post. As mentioned below, I would like to see the system that incorporates receiving yards. Also, I worry that this system does not account for the percentage of offense/runs a back has relative to his whole team. In other words, if your team is designed for you to be the workhorse because you are the only viable weapon (think Ringer last year for MSU), then you should not necessarily be punished for that even though your YPC and RER numbers are going to be down somewhat. Maybe it means a factoring of attempts and yards by a percentage of total offense you generate.

Because given enough of an offensive focus, a player's stats on a per-play basis will inevitably suffer. The homerun back who only runs a couple of times a game, like Reggie Bush was when he first started at USC, will have crazy stats but you doubt it can scale successfully to more carries.

Still, good stuff. I'm looking forward to seeing your next post.

Not a Blue Fan

September 20th, 2009 at 8:43 AM ^

I'm not a fan of YPC, particularly for guys who don't get that many carries. In addition, you've heavily weighted TDs, which may or may not favor good running backs. To wit, a few years back I recall Jerome Bettis opening the season as a short yardage only guy, and by week four he had a stat line with something like 10 carries, 8 yards, 5 TDs. Your metric would unfairly overvalue his contribution to his team.

In addition, the use of mean YPC implies that the distribution of yardage on a per run basis is normal (or at least close to normal). This is demonstrably untrue - there are many, many outliers. A mild improvement could be made by perhaps using the median, but honestly I'm not a stats maven and couldn't tell you if the median is substantially better than the mean in this case. The point being, the mean skews things by overvaluing a few long runs.

Also, why not consider first downs generated? A running back that keeps the chains moving is very valuable. You probably also want to consider situational stats (cfbstats.com has these); it's well established that defenses respond differently to different situations - a RB who routinely converts third and short is much more valuable than a guy who converts second and short (when the defense is likely looking for a pass).

Just some thoughts - what you've got here is a good first stab at a metric.

DeepBlue83

September 20th, 2009 at 9:56 AM ^

to be evaluating rushing stats this early in the season. There are always a few guys like Bolden who pile up big numbers against a few early season cupcakes, and then vanish once their teams get into conference play. Neither stats nor polls mean much of anything until we hit mid-October.

Eye of the Tiger

September 20th, 2009 at 10:38 AM ^

I like the idea of re-evaluating rushing statistics, but think you are overvaluing touchdowns, and perhaps also fumbles. This does not make sense:

10att, 40yds, 3 TDs (RER: 7.00)

Has an RER that's similar to a speed back who might rack up the yards, but is prone to a mistake here and there and might not always get the ball on the goal line:

28att, 170yds, 2 TDs, 1 Fumble (RER: 6.79)

Fact is, the second guy actually puts the team in scoring position throughout the game. The first guy does not. So why should the first guy be valued over the second?? One fewer touchdown and one more fumble shouldn't be worth more than 130 yards, particularly since the touchdowns are likely short-yardage carries.

I agree there is need for a better system than just yards/game, but I think you could tweak your system.

What if you changed this:

(Yards) + (Touchdowns x 10) + (Fumbles x -10)
Attempts

to:

(Yards) + (Touchdowns x 5) + (Fumbles x -5)
Attempts

???

Also, what about fumbles lost vs. fumbles recovered? Blocks? Etc.

MCalibur

September 20th, 2009 at 1:55 PM ^

I think this is a valid point but a single game comparison in not appropriate. A fumble is killer. So what if player B helps his team down the field if he just ends up giving the ball away? In fact, after reading some of the previous posts and comments in this series, I now think fumbles are under-penalized here. In the Quarterback Efficiency Rating post one of the commenters presented some information suggesting that a TD has the equivalent value of 20 yds whereas a turnover has the equivalent damage of -50 yds. For a single game that would kill an RB's relative contribution rating but over the course of a season it should simply ding it a little. If a player has a habit of putting the rock on the carpet, he wont be playing.

Finally there's always going to be a subjective component to player evaluation. A goal line back who gobbles up TD would never be more highly valued over a guy who is in on every play. Also, an every down back who gouges the sunbelt conference week in and week out will always have the stigma of playing against limp competition. Ratings and metrics are decision helpers, not decision makers.

jmblue

September 20th, 2009 at 3:06 PM ^

ALSO, it doesn't take into account fumbles. 200yds in a game is all well and good, but if all that field position is wasted because he fumbled 3-4 times, it doesn't help at all.

But how realistic is that scenario? Most guys who fumble regularly get buried on the bench.

Ali G Bomaye

September 20th, 2009 at 9:45 PM ^

but this method is sort of BS.

First, it overvalues TDs pretty significantly. Most touchdowns for running backs are simply a result of getting carries near the goal line. In Barry Sanders' final season with the Lions (it's easier to find old NFL stats than NCAA ones), he had 1491 yards and 4 TDs. Would anybody argue that Leroy Hoard, who had 479 yards and 9 TDs in the same season, was anywhere near as valuable as Sanders that season? Of course not - but this is pretty close to how full seasons of your examples above would play out.

Second, it gives no bonus for attempts/game. I know that production per attempt is one key factor to measuring a RB's value. However, it's much more impressive to have 40 carries for 200 yards and 4 TDs (5.0 ypc) in a game than to have 10 carries for 50 yards and 1 TD. The first guy carried his team's offense; the second guy was a nice change-up that probably wasn't the focus of the defensive game-planning. This is sort of shown in the example of Robert Turbin, above. He has the second-most carries on his own team, and most of his outstanding average comes from one 96 yard touchdown run in which all the work was done within 10 yards of the line. The length of this run was only due to the Aggies being backed up on their own 4 yard line; it would have been just as impressive if it was from the opponents' 10. Now, if he could keep up that production over a lot of carries, he would be something special, but as it is, he's a guy who got lucky in a small sample size.

I think this kind of stat has potential, but it needs a LOT of work to be anywhere near useful.

MCalibur

September 21st, 2009 at 9:42 AM ^

Fair points, but do you have any suggestions? The most common way to address the MVP-ness of a player is to apply a coarse filter based on a minimum number of plays (ex: QB Efficiency based on a minimum of 15 atts/gm).

I think another way is to 'un-normalize' the score by removing the denominator. The problem with this is that this then over values raw production. Javon Ringer being a case in point. He ran for ridiculous yardage totals but, his ypc was limp as hell. He only beat Jahvid Best by 100 yds-ish on almost twice as many carries!

Once again, math is a tool. No metric will ever remove subjective interpretation when it comes to unequal scenarios; it's a Quest for the Golden Fleece.

NoNon

September 21st, 2009 at 3:44 PM ^

I don't want to beat a dead horse but I would have to agree, there has to be a balance between total yards and touchdowns.

Would anyone argue that in a two back offense, say the Tennesse Titan's, that Chris Johnson is less important than LenDale White in a two back system (maybe not yesterday but you get the drift).

Let's say Johnson has carried 17 times in the game for 125 yards and no scores while White has carried 3 times for 14 yards and 2 touchdowns.

Assuming neither fumbled, under your formula, White's value would be 34/3 = 11.33

While Johnson has 125/17 = 7.35

It appears that White is the far superior back (with only 3 total touches) while Johnson recieved the majority of the carries by a wide margin and outgained White by a significant majority also - continue those trends for the year (give or take a few fumbles, td's yards) and White appears to clearly be the dominant back when Johnson is receiving the bulk of the carries.

This would seem to overvalue TDs and underrate attempts. It also does not factor in receiving yards and receiving TDs(in which Johnson would also have the clear edge in over White) which would further showcase a back's versatility.

MCalibur

September 21st, 2009 at 5:43 PM ^

This is all true, but the metric isn't meant to separate LenDale White and Chris Johnson; it's for separating Chris Johnson from Adrian Peterson. If a RB can't take care of business at the goal line or in the red zone, how good is he really?

Player ratings are a fine/medium filter. The user must filter out the LenDale Whites from the discussion at the coarse level b/c of lack of touches and what not; then applies the rating; then considers other, more difficult to measure stuff (pass protection, 1st down conversion, etc) to distinguish players at the fine level. There is no way to remove subjectivity from these discussions regardless of how nuts you get with UFR level inspection.

NoNon

September 21st, 2009 at 9:33 PM ^

Understood, but if it's an indication of the measure of a running back than why have the system at all if it just "filters out" all of the Jerome Bettis's and LenDale White type running backs.

The numbers are clearly slanted to White-type running backs under that formula and if they (White-type backs) are not supposed to be in the equation at all after being filered out, then clearly its not an effective formula to measure a running backs usefulness if it filters out all goal line runners.

And I agree there is no subjectivity, but I guess the closest formula would be fantasy points.

MCalibur

September 21st, 2009 at 10:06 PM ^

Overall, I think we're on the same page here, or at least the same chapter. My basic point is that you don't need an equation to tell you that Chris Johnson is more vital to the Titans than Lendale White; it is self evident. What's more difficult to determine is, who has the most absolute value as a RB, Chris Johnson, Adrian Peterson, LaDainian Tomlinson circa 2006, Barry Sanders circa 1996, or Walter Payton circa 1986? Lendale White and Jerome Bettis are nowhere near that discussion, you throw them out.

I guess it comes down to the question: what is the goal? If the goal is to determine which player is more valuable then you automatically throw away obvious 'not-the-answers' and focus on the players that make the basic cut . If the goal is to come up with a metric that deals with this issue for you then you're right, you need to face the issues you mention. But when you do that you're opening up a whole set of issues when you ask a different question: Who's better Drew Brees, Adrian Peterson, John Elway, Jerry Rice, or Emmitt Smith?

That question demands that you focus on win contribution and to do that you have to keep a yard-is-a-yard mentality. With respect to the mission of the Offense, the only thing that matters is getting closer to the goal line then, ultimately, crossing it by whatever means necessary. A passing yard is of equal value to a rushing yard is of equal value to a return yard and the same goes for TDs. Focusing strictly on a subgroup of the Offense is of limited value.

+1 for thoughtful engagement.

Eye of the Tiger

September 21st, 2009 at 10:56 AM ^

1. Instead of doing x10 for touchdowns, change it to x5.

Your current system way overvalues touchdowns, many of which are 1 yard handoffs, and as a result undervalues yards, which put the team in position to score touchdowns (and field goals).

2. Instead of doing -x10 for fumbles, change it to -x5

Same critique.

Now your two guys:

10att, 40yds, 3 TDs (RER: 5.50)

28att, 170yds, 2 TDs, 1 Fumble (RER: 6.25)

As I see it, that is a more realistic assessment of their comparative worth.