Signature Stats: Pulse Rate

| January 18, 2013

Football statistics are great — but smart statistics are even better. Numbers can always lie, and they won’t ever tell you the complete story, but to use that as an excuse for never trying to better the raw data already to hand would be folly.

We have used yards per carry as a statistic for decades, breaking down a runner’s total yardage by his number of carries. Adrian Peterson ended the 2012 season rushing for an astonishing 6.0 yards per carry on his way to 2,097 — but what was he realistically averaging on normal runs?

We decided to take a look, and to try and weed out the breakaway runs that otherwise skew a player’s average. When we looked at all runs on the season, there was a leveling out point for the frequency of yardage gained. This figure stood at 10 yards. This means that most runs in the NFL gain 10 or fewer yards, and anything above that is a figure that potentially skews a runner’s average mark.

So instead of chasing a blunt average, we are going to look at a player’s rushing average using only runs of 10 or fewer yards, with anything that went longer than that given a capped value of 10 yards and still included in the average. This way his average isn’t dragged upwards by one run of 90 yards, but rather we get a more accurate view of what a runner is getting every down. In short, we get his Pulse Rate.

Reining in ‘All Day’

The first and most obvious thing to note is the demise of Adrian Peterson. The man they call ‘All Day’ may have been a threat to break it all game long, but in terms of what he was churning out every snap, he was far more feast or famine. When you look at the stats we already have on the site, that’s not really that surprising. Peterson gained 1,184 of his 2,097 yards (56.5%) on breakaway runs, or runs over 15 yards. That percentage is by far the most of any other runner, and his 40 runs of 15 or more yards is 16 more than any other back.

Peterson’s season was utterly ridiculous, but the Vikings weren’t necessarily able to rely on him having the best Pulse Rate in the league every game. His mark of 3.68 yards per carry ranks just 22nd, dropping him from his gaudy 6.0 raw average.

By contrast, running behind the league’s most dominant offensive line has left Kendall Hunter sitting at the top of our list from his 72 carries. Those carries went for 371 yards on the season, giving him a raw average of 5.2 yards per carry, and he was consistent enough that when we dampen the effect of those longer runs it falls only as far as 4.49 per carry, good enough to top the table.

The Workhorse Backs

Arguably the most impressive figure though is that from C.J. Spiller, a player most would have imagined to have been heavily reliant on longer runs to boost his average. When you look at the numbers, Spiller was indeed at the sharp end for those breakaway runs. His 16 runs of 15 or more yards ranked fifth in the NFL, and those runs accounted for 39.9% of his rushing total for the season, lower than only Peterson and Jamaal Charles, and yet his Pulse Rate remains one of the best. The explanation lies in the number of negative or poor runs he had. Spiller was stopped behind the line of scrimmage for no gain on just 15.5% of his runs, one of the better marks in the NFL. Peterson by contrast, adding further context to his score, was stopped on almost a quarter of his runs, 24.5%.

Marshawn Lynch tops the list of the workhorse backs, followed closely by Alfred Morris, who was aided by the creativity of the Washington offense on his way to a PR of 3.89 yards per carry.

 

Rank
Player
Carries
Pulse Yards
Pulse Average
1Kendall Hunter723234.49
2C.J. Spiller2078754.23
3Marcel Reece592414.08
4Andre Brown732974.07
5Lamar Miller512064.04
6Marshawn Lynch31512674.02
7Justin Forsett632503.97
8Ahmad Bradshaw2218743.95
9Willis McGahee1676503.89
10Alfred Morris33513033.89
11Frank Gore25910033.87
12Joique Bell823153.84
13Pierre Thomas1054033.84
14Robert Turbin803033.79
15Ben Tate652443.75

 

Behind the Chains

At the other end of the scale all three of Arizona’s runners prop up the foot of the table, casting more light on the poor performance of the Cardinals’ running attack and blocking up front. Darren McFadden lies perilously close to the foot of the table despite the other members of the Oakland backfield faring much better, lending more weight to the notion that McFadden just wasn’t a good fit for the Raiders’ scheme in 2012.

Most worrying for a team playing in the NFC Championship game this weekend though is the Pulse Rate the Falcons have been able to expect from Michael Turner this season. When the Falcons can get their blocking functioning, Turner can have still have big games, just as he did against Seattle last week, but when they can’t, which has been most of the year, Turner has been completely unable to get up to speed before being taken to the ground. The Falcons have been able to rely on him for an average of only 2.93 yards this season when his long runs are mitigated. That isn’t enough to move the chains on a consistent basis and has put Atlanta behind the count more times than they can afford to be against the 49ers.

 

Rank
Player
Carries
Pulse Yards
Pulse Average
60Isaac Redman1103483.16
61Chris D. Johnson2768693.15
62Peyton Hillis852673.14
63Jacquizz Rodgers942953.14
64Bryce Brown1163553.06
65Jonathan Stewart932843.05
66Toby Gerhart501523.04
67Alex Green1344053.02
68Michael Turner2226512.93
69Rashad Jennings1012702.67
70Darren McFadden2165632.61
71Jackie Battle952442.57
72Beanie Wells882112.40
73LaRod Stephens-Howling1102532.30
74Ryan Williams581332.29

 

Follow Sam on Twitter: @PFF_Sam

  • Kosmokenny

    If you are going to remove long runs due to them inflating the YPC average, you should also find a way to remove the deflation of YPC avg that touchdown runs inflict.  Peterson wouldve broke the rushing record if that darned end zone didnt keep getting in his way…

    • captainhuggyface

      Excellent point. A 2 yard run into the end zone is not the same as a 2 yard run at the 50.

      But how do you accurately measure distance of TD runs from, say, the 5 yard line? Sure, there are times when a RB bursts through a hole and runs into the end zone untouched from the 5. I guess the folks at PFF can review all TD runs of 10 yards or less and place a value of projected yards.

      • Martin

         I’d argue that the effect is near meaningless.

        Most workhorse running backs carry the ball on average 200-300 times per year and score between 5 and 10 rushing touchdowns per year. Several of those rushing touchdowns likely come from beyond 10-yards, so there’s no “deflation effect” from those runs with this signature stat.

        Therefore, for the average starting running back, we’re talking about 4-6 “touchdown-deflated” runs out of 250 total carries, so there’s almost no skewing going on. As for non-workhorse backs, sure, they have fewer carries, but at the same time, they have fewer TDs. This only really becomes a problem for running backs who carry the ball fewer than 100 times but score 5+ TDs – and there are very few of those.

    • Martin

       Out of 348 carries, Peterson scored 12 touchdowns. That meant that “deflation effect due to touchdowns” at most affected 3.4% of his runs.

      When you consider that he had **several** touchdown runs of greater than 10+ yards (meaning there’s no “deflation effect” since this analysis caps all runs at a maximum of 10 yards per carry), the effect is even smaller than 3.4%.

      Therefore, your proposed adjustment would have little to no effect on Peterson’s numbers, since it only would remove probably around 6 of his 12 touchdown runs from this analysis, and that’s only 1.7% of his runs.

  • beardown

    Sounds like a case for a Total Running Back Rating similar to the recent QBR.

  • river

    Though a running back who is averaging around 2 yards a carry during a game will see less men in the box as there is no threat, which in turn will allow him to increase his average.

    However someone who is averaging around 6 yards a carry will see more men in the box to limit the threat thus reducing their average.

    Maybe you could do a stat on yards per carry v average men in the box? 

  • roguepatriot

    I’m not surprised to see Chris Johnson on the “bottom” list.
    I noticed this phenomenon with the Cowboys, several years ago, when they had Julius Jones. Jones had some impressive runs, but did little in between. Will this stat have its own tab for the running backs section anytime soon?

  • LightsOut85

    I’d have to 2nd the variable of defenses faced. AD + Ponder is a much different defensive plan than (insert “bad” RB & good QB).  Also, the minimum-carries needed seems a little low. I’d feel better looking at minimum 100 carries.  Just looking at the top 15 I’d imagine AD would jump up much further in the rankings. AND, I’d say it’d be more prudent to look at the difference in “pulse rate” between the RBs, not just rank. That is, there could be .05 difference (statistically likely not significant) between 2 RBs, but 5 ranking spots, which makes the difference seem greater than it is.

  • G_Hunt

    CJ Spiller is surprising. I think of him as a big play back.

  • Alex

    I wonder if carries within the first ten yards say more about the oline than the RB. But when the elite RB’s get to the second level (10 yards) they are the ones to make the elite plays.  This explains Peterson to me.  You seemed to suggest this with you Kendall Hunter comment.

    • LightsOut85

      This is more along the lines of what I’d theorize too. If not the first 10, at least the first 5.  One theme of all the Elusive Rating articles over the years was that guys with low ERs, like Benjarvus Green-Ellis, were getting just what the OL opened up, but not much after. That you just have to have good enough vision to navigate the space the OL opens up, IF they open it up. But once you have room to run, that’s when you can show what you can do. Yes, your location and the location of other defenders at that exact moment is an important variable that’s unique to every play & you could argue it’s all a product of the random circumstances, but I think the most talented backs will increase the odds of those big plays happening.  

      I’d like to see what % of runs that were at least 5 yards, went for 15+, or 20+. 

  • Cito_03

    Parsing outliers can be a tricky business when looking at elite performance because the outlying data are a key feature of what makes a player elite.  An alternative approach would be to look at the median and modal values for any given player.  

  • Izach

    i find it interesting that the “good” pulse rating list is littered with RBs not only getting less carries but guys who were proably coming in to games fresh, in relief of a “starting” caliber RB. thus meaning defenses took it easy when they came in mixed with them being at 100% inflates their rating somewhat. maybe it just goes to show you how good guys like lynch and morris really were

  • mjp

    I understand the desire to get rid of averages, they are one of the most meaningless statistics in terms of describing moment by moment ability (90yd runs sure make a guy who has 15 2.5 yd runs seem great).  I know in baseball I want a guy who hits .300 every week of every month
    not a guy who hits .400 in May and June, .260 in July and then .200 in the heart of
    the pennant race.  However I have issues with your approach.

    I believe what you’re looking for is some measure of modes, or how many times did a RB gain at least 5 yds or 10yds, not what is his average YPC on run plays of 10yds or less (doesn’t the irony strike you that you’ve just formulated a stat that is yet another average, even if it’s just cherry picked to avoid long runs), unfortunately your smart stat doesn’t measure modes and thus isn’t very smart. 

    This is where I have an issue with your approach, your pulse rate doesn’t count modes of a certain threshold, you’ve just cherry picked runs and penalized a RB for being dynamic.  In essence what you’ve found is the best of the average RB’s, or guys who are good at only gaining a maximum of 10yds.  Peterson had 40 runs of 15+ or more yds which you are removing because you claim it skews his average.  So, did you remove all his runs of 2 yds or less, or 0yds or less?  Don’t those runs skew his average as well?

    On that particular point, don’t busted run plays of -2 to 2 yds, plays where essentially a RB is hit before he gets past the LOS and either falls sideways for -2 to 0 yds or forward for 0 to 2 yds (which I’m sure a RB experiences far more often than the gain busters of 75 yds), don’t those plays help reign in the effects of those handful of big gains, I’m sure they do.  One wonders if there is even a need to construct a new average statistic given this point of view.

    And what about those runs of 11, 12, 13 and so on?  I understand the desire to reign in an RB’s average which is skewed by 75 yd runs, especially mediocre RB’s with few carries or lots of below average gains, but how does a 15 yd run skew his average when he is consistently breaking 15 yd runs, that’s just how good he is (never mind the fact that AP had over 1,000 yds come after contact, he should have just gone down and repeated that process, then he’d fair a lot better in your pulse rankings).  Basically, it’s one thing to try and remove outliers but what you are doing is essentially skewing the spectrum of allowable runs which in turn skews the reality of what the numbers are and should be telling you.

    A true pulse statistic would consider multiple factors and it would try and count modes, runs of 5 yds, 10 yds, 15 yds and so on (perhaps 4 yds could be the lowest threshold since that seems to be a reasonably good gain).  The number of broken tackles or yds after contact should absolutely be weighted into the calculation as well.  You could then find a way to combine long runs (20yds seem like a reasonable place to start) vs short runs (3yds, mabye 2) into one variable or factor because they are in fact outliers, with so few you could; supplement this variable by using your play by play analysis to determine if the fault of a negative run play is on the RB or the O-line and whether or not a RB broke any tackles on a big run or was he running free).

    I would also suggest you give a weighted value for those runs that result in a 1st down giving more weight to longer runs that result in a 1st down and also carry no negative effect of runs under 5 yds that result in a 1st down and occur on 3rd and short or 4th and short situations (since the goal there is merely to gain 1-3 yds for the 1st down).

    I’m willing to bet Adrian Peterson would fair quite well in a statistic that incorporates the factors that I’ve tried to present.  Gerrymandering or cherry picking the spectrum of runs to create a new average is not a good idea. A useful statistic should utilize a more comprehensive set of guidelines that incorporate modes that are supplemented by other useful factors/statistics,  or in other words, create a stat that is comprehensive and multifactorial (Red Sox fans cringe).

  • DownsouthD

    I’d love to see the pulse rate and % of times tackled for a loss added to the RBs signature stats for all premium account holders to view.