Passer Rating Differential - Is There Correlation With Wins?

Submitted by ForestCityBlue on

I read an interesting article posted at SI.com by the folks from Cold, Hard, Football Facts:

http://sportsillustrated.cnn.com/2011/writers/kerry_byrne/06/23/most.important.stat.passer.rating.differential/index.html?sct=nfl_wr_a1 

I am curious whether there is any of the stats minded guys (or gals) around here who could be cajoled into seeing if the Passer Differential Rating correlation between winners and their ability to pass the ball and defend the pass transfers from the pros to college football.  I find it fascinating that the single biggest predictor of success in the NFL is a team's ability to both pass the ball and defend the pass.

With the greater diversity of offenses in the college game that include both run first offenses as well as pass happy attacks there may be less predictive value for Passer Differential Rating in the NCAA, but it might provide anothe point in the debate over which offensive system is the "best." 

Important also is the defensive component.  Dominant defenses may be able to make up for a lack of a world beating pass attack.  What is most interesting about this stat is that it is a team's ability to pass the ball and defend the pass that is most important (think recent Michigan secondary doom) predictor of success.  Its not about running the ball or stopping the run, but rather passing the ball and stopping the pass.  Its not overall offense or defense that is most important as a predictor, but just passing success and passing defense.

I wonder if there is someone here willing to take up the challenge of seeing if this correlates at the college level... 

Roger Mexico

June 27th, 2011 at 6:52 PM ^

I don't have the statistical insight, but I always thought running (and stopping the run) would be a better predictor of success, under the following logic:

Running is a low variance strategy in that it's easy to get a small number of yards, and hard to get a big number of yards.  You can run for 2 or 3 yards every play, and it's rare to lose more than a couple, or gain more than 5-10 (unless you are Denard).

Passing is high variance.  A completed pass frequently results in a lot of yards, and sometimes loses you a lot (like a sack or interception).

Therefore, if you can run for 4.5 or so yards every play, you are almost impossible to stop, because you never need to pass.  If you can't run for a decent number of yards, you're forced to pass, which may bring success, but will eventually fail.

ForestCityBlue

June 27th, 2011 at 7:00 PM ^

This is what Cold, Hard, Football Facts is all about.  You would think that the common wisdom of running the football and stopping the run would be the most important predictor of success, but in the end, the most reliable predictor of success in the NFL is how well you pass the ball and how well you defend the pass.  In fact, defending the pass may actually be the most important of the two.  The stats all point to the passing game as the most important predictor of success.  My question is whether or not this is the same in college football.

treetown

June 27th, 2011 at 8:39 PM ^

Agree with your hestiation to completely jump on this bandwagon. Passing is very important and stopping the pass is as you note maybe even MORE important. In large part this is due to (1) relative ease most teams have a stopping the run (2) team consistency from year to year and (3) huge pool of data against relatively closely matched teams. The last point can't be underestimated. It is hard to judge the value of strategies or plans if one side is overwhelmingly superior in sheer size, strength and speed. Mismatches are far more common in college ball.

Zone Left

June 27th, 2011 at 7:45 PM ^

You can look at it in another light. The NCAA passer rating (called passer efficiency) formula is:

(8.4*yards + 330*touchdowns + 100*completions - 200*interceptions) / attempts

Since it's divided by attempts, it basically attempts to measure how good a team/player is per pass, not overall.

A team that runs really well, like Michigan did against UConn may not not have to pass much and may get more "bang for the buck" as it were per pass (Michigan ran about 73.5% of the time that game). Michigan had a positive 79.69 efficiency differential in that game, partly because UConn got caught seriously out of position several times due to the effectiveness of Michigan's run game.

Really, if teams can do what they want to do, they'll be successful (obvious, right). In a really simplistic view, last year's Michigan team wanted to run well to set up huge passes. Texas Tech under Mike Leach wanted to pass and then run as a change of pace. Both philosophies can be successful, it just depends on who's playing who. 

Texas Tech was able to pass for relatively short distances per play, but were almost impossible to stop. For them, it was a relatively low variance strategy that had relatively low risk, but low reward too.

ForestCityBlue

June 28th, 2011 at 9:40 AM ^

This is what facinates me about this stat and its applicability to college ball.  In the pros there is much less offensive variation, not nearly as much as college.  But from Michigan stats below, even a run first, run heavy team like we were last year, the correlation between passer effecieny differential and wins is startling.  Could it be that regardless of what percentage of your plays are passing plays, how effecient you are when passing (and in defending the pass too!) is an extremely important predictor of success, perhaps the most important predictor.  It is the variance in offenses run in college that makes this a much more interesting question than the pros where the pass is all important.  Is it as reliable a predictor of success as it is in the NFL?  Given the data below from Michigan's last year, I would venture a hypothisis that if someone were to run the data, they would find that it would be a very reliable predictor of success across the NCAA and irregardless of the offensive system used.

Zone Left

June 27th, 2011 at 7:07 PM ^

Michigan's passer rating differential from last year:

 

Game Michigan Opponent Diff Game Result
UCONN 172.3818182 92.68421053 79.69760766     W
NOTRE DAME 119.49 121.8272727 -2.337272727     W
UMASS 248.8857143 156.0275862 92.85812808     W
BG 196.2307692 127.0451613 69.18560794      W
INDIANA 259.8117647 249.3777778 10.43398693     W
MSU 111.5862069 178.4923077 -66.9061008     L
IOWA 133.5 198.8833333 -65.38333333     L
PENN STATE 121.04 147.5 -26.46     L
ILLINOIS 173.5794872 172.4592593 1.12022792     W
PURDUE 118.048 72.99393939 45.05406061     W
WISCONSIN 155.3866667 192.56 -37.17333333     L
OHIO STATE 85.44242424 152.1481481 -66.70572391     L
MISS STATE 129.1121951 215.2347826 -86.12258749     L

Pretty wild, huh? Michigan had six positive passer rating differentials and seven negative differentials. The team that won the passer rating differential won the game 12 of 13 times, or 92.3% of the time.

Even more bizarre, the differential generally decreased as the games got closer. Michigan's two closest games, Notre Dame and Illinois, had pretty small differences and the games with wide score margins generally had wider ratings differentials favoring the winning team.

The math may be a little off as I didn't double check, so please forgive me if I made a couple mistakes.

 

Zone Left

June 27th, 2011 at 7:56 PM ^

If you look at the formula I posted in another post in this thread, yeah, big plays (good and bad) make a huge difference. Touchdowns and picks are the most heavily weighted items.

ForestCityBlue

June 28th, 2011 at 9:03 AM ^

It is cool to see this play out on a game by game basis.  With the differential being the key, you either have to do one (pass efficiently or defend the pass) or both well enough to create a possitive differential. 

Hopefully we can improve enough this year in pass defense to give the offense enough time to overcome teething pains.

I would be very interested to see if this bears out across the NCAA and whether or not the eventual champions (both conference and national) are all at the top of the statistical heap in terms of passer rating differential.

meechiganroses

June 27th, 2011 at 7:19 PM ^

I think this is beneficial to understanding our new offense.  I think the returning secondary steps up this year (because of more comptent LB play)--and Denard's rushing threat keeps our passing numbers about the same as last year with fewer TDs and INTs

justingoblue

June 27th, 2011 at 7:45 PM ^

I'm really hoping (Hokeing?) the TD's don't come down this year. Borges/Hoke put up 35ppg (we had 32.6) last year. Granted, they don't have their personnel all in place, but they do inherit Denard. We could easily be in the black for point differential between 2010/2011 if we can find a feature back and Denard cuts a few of the interceptions like the MSU game.

Zone Left

June 27th, 2011 at 7:51 PM ^

Hoking! I get it!

I think he meant see the passing touchdowns and interceptions come down a little. Last year, Denard had 18 touchdowns against 11 picks. I'd like to see that stay about the same, but cut down on the fumbles and find a non-Denard running game. That, and some actual defense, would put Michigan in a great position to improve on its record from last year--even if points come down some.

justingoblue

June 27th, 2011 at 8:06 PM ^

Going by 2010 offenses, we might be in for a good deal more passing (~50yds/game) so I really wouldn't know what to expect out of Denard's passing stats this year. Then again, I doubt even the coaching staff know what they'll be running next year, so who knows. One thing I will say is that I think the ppg need to go up for us to improve unless Gerg was even more incompetent than we all thought (I'm not convinced this defense can jump 40-50 places in one offseason, especially with only half a Mattison class on campus).

Zone Left

June 27th, 2011 at 8:29 PM ^

I do hope that the staff has a pretty good idea of what it wants to try next year. It will definitely be a work in progress, but I think we'll see a lot of quasi-spread stuff designed to 1) get Denard in space with two-ish passing reads followed by an auto scramble and 2) to use more of an H-back out of the shotgun to mimic i-form MANBALL.

As for the defense, last year seemed to be a perfect storm of injuries, zero depth, and terrible coaching. Injuries should be a little less damaging as the depth on paper looks better and the coaching has to be better. Woolfolk will be back, Avery should be better, and the safety play should be better, which will help the secondary and a better scheme should really help a front seven that looked totally lost.

As for PPG, Michigan had a pretty distorted PPG with the 67 point multiple overtime game against Illinois that added a couple PPG just from the overtime session. They really need consistency and to stop turning the ball over close to the goalline. Consistency is going to be tough, but a more seasoned Denard and a really experienced line should (hopefully) help with the turnovers.

No one knows what will happen, but there is definitely reason to believe in a return to traditional form on paper. This will be helped by a schedule without Penn State and Wisconsin. Notre Dame, MSU, and Nebraska seem like the really tough games, OSU is a total unknown, Iowa and Illinois lose a lot, and Northwestern may have a severely diminished Dan Persa.

The schedule also sets up well for the team to grow as the season goes on without risking many losses early. Notre Dame is a really tough second game, but no one else before Northwestern in Week 6 has a great shot at beating Michigan.

All that said, this is a year of hoping pieces that should fall into place do fall into place. Unfortunately, there are a lot more unknown pieces than there are in most seasons.

justingoblue

June 27th, 2011 at 9:09 PM ^

I completely agree, but there are a couple of your points I'd want to comment on:

First, I think we ended up with less points per game than our yardage total would suggest, so I'm not sure how much of an effect the stupidly high scoring Illinois game had bringing our total points up vs. Denard throwing two picks in the end zone against MSU and the rest of the drives ending in something other than points had on bringing that total down.

Also, I do agree that the coaches know what they would like to run, but I doubt we see what they'll actually go with until ND (assuming it's a close game like we're expecting). It's really easy to sit in April and say we're bringing in a bunch of new stuff, but if ND is up by 7 after a half of under-center manball, we might be running the same offense as last year in a hurry.

Zone Left

June 27th, 2011 at 10:10 PM ^

I only included the Illinois game because the extra time added almost two points to Michigan's season average. The stuff you mentioned definitely dragged Michigan's PPG down quite a bit. Not finishing was a big issue the last couple of years.

I don't think this staff will be very wedded to its system. I think they plan to bring the new stuff in slowly. We probably aren't going to see the QB off tackle play that much, but a lot of it will look pretty similar. If they come out against Notre Dame and try to go back to running the Mike DeBord counter 35%+ percent of the time I'll eat my hat.

Sambojangles

June 27th, 2011 at 8:49 PM ^

I am 99% sure this has been posted, and refuted, before. Among the potential explanations for the correlation (not causation):

  • Passing effectively is hard, so teams that do it well do better. A low passing efficiency basically means a lot of dropped balls and other incompletions, which obviously means ineffective O.
  • Interceptions are highly correlated with both losing and low PER. That is part of the cause. not the rating itself.
  • Running the ball sets up the pass, and vice versa. You cannot separate one from the other so simplistically.

I think the OP goes overboard in saying, "Its not overall offense or defense that is most important as a predictor, but just passing success and passing defense." No, not exactly. There is too much noise to come to a cause/effect conclusion from the data.

mikoyan

June 28th, 2011 at 9:57 AM ^

As others have pointed out, teams with a higher passing efficiency tend to do so because they can pass when they want to and not when they have to.  They can pass in situations where it is generally safer to do so and don't run into the problems of interceptions or whatever.  Teams that can defend the pass well can likely defend the run well as well and when you take out one aspect of a team's game and turn them one dimensional, you are more likely to succeed.

ForestCityBlue

June 28th, 2011 at 10:03 AM ^

But what you are saying makes the case for the value of the stat.  If they are efficient because they pass when they choose and thus are more likely to have a high passer efficiency rating, then it is an indicator of overall offensive success.  The same can be said for defense.  If everything you are doing is making it easier for you to defend the pass, then it is an over all predictor of success.  Although they never state it directly, the people at Cold, Hard, Football Facts are implying that if you are looking for one stat to predict how successful a team is over all, that Passer Efficiancy Differential is that one stat with the highest predictive value.  I find the argument compelling and interesting.  They make a persuasive case, especially if you read the longer atricle on their website that they link to in the SI.com article.

ForestCityBlue

June 28th, 2011 at 1:33 PM ^

Actually, the case that the people at Cold, Hard, Football Facts are trying to make is that the Passer Efficiency Differential is the best indicator to tell you whether or not a team will be successful.  In fact it might be the most relaible predictor.  What the stat tells you is how well they are taking advantage of their passing opportunites and how well they defend the pass.  Do both of those well, or one well enough to make up for a weakness in the other, and you are almost guarenteed to win, and not just win, but win it all.