This is the 6th in a series of articles I am writing to analyze some common NFL statistics, focusing on how much value they have relative to team wins. I want to acknowledge the work of Brian Burke, Chase Stuart, and even our own Matt Grecco, who inspired this analysis and whose methodologies I have leveraged, as well as Pro Football Reference, Armchair Analysis, and NFL.com as the sources of my data.
I'll admit it. I have a problem. I like offense. And not just any offense; passing offense.
Keep your smashmouth games that grind to a 7-6 finish, I want to see the star wars numbers. Sure watching running backs juke and truck defenders is exciting, but its even more so when predicated by a hoodwinked defense off a screen pass.
But if passing offense is my addiction then the NFL is my enabler. Since 2000, 66% of all yards and 63% of all TDs have come via the pass. And in that time, 11 out of 17 MVPs have been quarterbacks. The bottom line is that passing is far more important to wins than rushing and that is reflected in the metrics.
That is why the most popular stats are all passing related. Passer Rating, Completion %, YPA, TD-INT ratio etc. There seems to be no end to how the passing game is judged.
But what about the run: how do we measure that? Yards . . . which sucks . . . and yards per carry (YPC). That’s about it.
YPC is a useful stat when comparing running backs, but does it help identify which team has a good “run game”? Does YPC tell the rushing story like yards per attempt does for passing? Let’s see.
In the Luck era, the Colts have only managed a 3.91 YPC, which is a pathetic 28th place.
But don't worry, that looks a lot better when compared against a defense giving up a 31st ranked 4.58 YPC.
Dear God. This better be a crap stat.
Usually efficiency stats predict wins better than the related volume stats but for rushing yards, that is actually not the case.
The red line is the minimum correlation for a good predictive stat and neither rushing yards (Ry) nor YPC are in any danger of meeting that criteria. In fact, YPC’s 0.06 correlation to wins is barely more predictive than flipping a coin.
From the chart above, you can see that the average pass has a far greater yardage impact than the average run. The most common rushing gain is 2 yards whereas it is 3 times as much for a completed pass. Notice also that rushing plays have a much smaller variance of outcomes making explosive plays far less likely than pass plays.
When including incompletions, the NFL average YPA is 7.0 yards. A team would have to perform at the 83rd percentile level in rushing just to match that average passing performance.
It is most definitely a passing league. But let me clarify that I am not saying rushing is not important for wins. It most very certainly is. Rather, I am saying measuring it's importance using YPC is stupid.
So what can be used instead? I usually try to provide better alternative stats that can be derived from standard box scores, but with rushing it is difficult. There just aren't any good game level stats to truly judge a rushing offense.
Here is a list of stats I tried and their predictive correlations:
- R: volume of carries
- R%: % of plays that are rushes
- Ry: total rushing yards
- Ry%: % of yards that are rushing
- YPC: yards per carry
- R_fst: volume of rushing first downs + TDs
- R_fst/A: rushing first downs + TDs per carry
- R_fst%: % of first downs/TDs that are rushes
None of these even come close to being acceptable. The first 5 are all related to yards and they are the worst of the bunch.
The last three are related to first downs and while they still aren’t good stats they are improved which provides an insight to a better measurement.
Not all yards are created equal. Gaining 1 yard on a 3nd and 1 is a better outcome than gaining 8 yards on 3rd and 10. So instead of measuring yards (and all its variations), the rushing game should be measured by how "successful" each of the individual carries are.
"Success Rate" is not a new idea. In fact it is one of the first ever ADVANCED STATS. The idea is that by adding context, success or failure can be judged for each play and then compared against all plays to determine an overall Success Rate.
The definition of "success" is more of a fluid concept than a strict formula, but let's use the very crude rule of thumb first written about in The Hidden Game of Football almost 30 years ago. A success on a play is any of the following:
- a TD
- an earned 1st down
- a gain of at least 40% of the "to go" yardage on 1st downs
- a gain of at least 50% of the "to go" yardage on 2nd downs
There are some tweaks to the above, but in general, them’s the rules(1) and the following chart shows the results when they are applied.
Muuuch better. By using play level data, Rushing Success Rate (RSR) creates a superior rushing stat. Quite simply, it helps explain how rushing is linked to wins better then yard based stats like YPC.
Applying RSR to the Colts 2012-2016 data significantly alters the narrative of the offense.
Hmmm, this method moves the Colts up to 11th place. Well that's not right.
I did say it was a very crude method, remember? I mean this was before Excel or the internet and they probably crunched the numbers on a PS/2 (no, kids, that is not a reference to a PlayStation).
I am not advocating this particular methodology, rather I’m just using it as an example to show that on average it measures the rushing game’s impact on wins better than YPC. Maybe not for the Colts specifically(2), but in general this is big improvement overall. Using DVOA, the Colts rank 19th in rushing and so improvements to the RSR methodology would likely move the Colts down from 11th(3).
If it makes you feel any better, RSR moves our defense up 4 spots to 27th which basically agrees with DVOA at 28th.
- YPC is somewhat useful for running back comparisons but has no value when attempting to measure the relative importance of a team's overall run game.
- Rushing First downs (including TDs) is not a great stat but is better than YPC.
- Success rate based stats are far superior to yardage based stats for rushing but they are very complex and require play level data.
1) The total number of plays exclude kneel downs and any rushes on 3rd and 5+ since the goal of those plays is not usually to get a first down. In the 4th quarter, the % thresholds are reduced 10% for teams with at least a 7 point lead (running out clock) and increased 10% for trailing teams (defense playing soft).
2) The biggest disconnect between the Colt’s RSR and DVOA was from 2014. Excluding that year yields a 16th place RSR which is fairly close to DVOA’s 19th ranking.
3) The currently most accepted methodology for Success rate is Expected Points Added (EPA), which is the basis of DVOA. Any play with a positive EPA is considered a success, which helps account for short fields or plays where the run is just being used to set up a field goal.