Doug Nussmeier, Advanced Stats Take

Submitted by The Mathlete on January 9th, 2014 at 5:52 PM

There are two main metrics by which I look at an offense, with different philosophies emphasizing different elements. I look at how well an offense does at converting first downs ([# of plays gaining a first down]/[# of first downs started]) and how good an offense is at stretching the field with explosive plays (Any yards gained beyond the first down line).

Below are the season numbers for each of Doug Nussmeier’s seasons alongside of the last 11 Michigan seasons for reference:


Blue=Michigan Red=Alabama Purple=Washington Gray=Fresno St

The top right quadrant is the Oregon zone. Offense that are really good at both. They consistently generate first downs but also produce big plays. The lower right quadrant is feast or famine. Lots of big plays, but can’t consistently convert first downs. The top left is probably where Brady Hoke wants to be, not consistently pushing the tempo or the big plays, but able to grind out first down after first down. The bottom left is for offenses that can’t do either well.

The Washington Years

As noted by Brian, in 2009 Nussmeier took over a tire fire of an offense. If there was a dot for 2008 Washington, you wouldn’t see it because it would be even lower and left of 2008 Michigan! His first year the offense improved along both dimensions and moved to bad but not awful. 2010 saw a bit more explosiveness but in year three the offense took a major step forward along both metrics.

Consistent improvement over three years is a very good sign. In fact, if you compare 2008-2011 Washington and Michigan, every year but 2010 is very similar and demonstrate a lot of positive improvement.

The Alabama Years

For a reference starting point, 2011 Alabama was most similar to 2004 Michigan. That was the team that beat LSU in the national championship. You can have an offense like that when you have a defense like that allows 37 bonus yards/game and an absurd 42% first down conversion (MSU was 59% this year).

In his first year turned the mediocre 2011 offense into a very good chain moving offense in 2012. For 2013 the moved further in that direction. The 82.7% first down conversion in 2013 was the third highest number since 2013. Some of that was due to the overall regression of defenses in the SEC in 2013. Texas A&M actually set the record this year with 82.9% conversions.


This does seem to be the coordinator who can do the things that Borges can’t while still fitting into Hoke’s desire for what his team’s offense looks like. Where Michigan has spent the last three years moving backwards, every single Nussmeier coordinated offense has shown year on year improvement. There aren’t going to be fireworks or a spread offense, most likely, but there should be a lot of first downs and hopefully consistent improvement.

From a watchability standpoint, this won’t be the fun offense many of where hoping for. It is a system that in the presence of elite talent and great defense can do everything you need it to. I have a working hypothesis that if your goal is national championships this is the way to go. Great defenses seem to have lower variance than great offenses. Put a team together around an elite offense and you get 10 amazing games and 2 games where the wheels fall off. Build it around a great defense and you are probably in all 12 games. Elite offense is great for making the leap from bad to good but if you want to get good to great, it has to start on defense. I’ll be pulling some more data this offseason to test this out.


Everyone Murders

January 9th, 2014 at 9:29 AM ^

Nice analysis here.  Is it feasible to examine the explosiveness / efficiency variables depending on field position too?  One advantage of an offense that excels on converting first downs would seem to be that you can limit the downside of bad field position.  It may just be anecdotal observation, but it seems that teams that do well churning out first downs are also the ones that make chicken salad out of chicken-shit field position. 

Conversely, it seems that churning out first downs in the red zone is more valuable than explosiveness, since there are only 20 yards to get in the first place.

I also wanted to be one of the first to thank you for your hard work by asking for more hard work!


January 9th, 2014 at 10:08 AM ^

Fremeau has a stat that maybe is what you're looking for here--it's simply the yards gained on a drive divided by the available yards at the start of the drive. What percentage of the way to the other goal line did you get?

Alabama was #10, about what you'd expect from their general level of offense. They were #15 in avoiding three-and-outs.

Everyone Murders

January 9th, 2014 at 1:39 PM ^

Not quite what I'm looking for (b/c it does not address the "getting out of trouble" element and focuses on the arithmetic mean rather than the result), but that's a really interesting stat that covers some of the territory.  Thanks!

I'd be interested in knowing how field position impacts 1st down conversion rate, and vice versa.  (I hope that makes sense.)


January 9th, 2014 at 9:32 AM ^

The colors are confusing, though.  How would a regression of points scored against these two variables work?  I would think explosiveness will help with scoring and first downs really help your defense by keeping the opponent's offense on the sidelines.


January 9th, 2014 at 9:59 AM ^

Right before I came to this post I happened to be looking at "expolosive drives" at footballoutsiders. Fremeau's definition is "any drive averaging more than ten yards per play."

Alabama was #1. 29% of their drives averaged more than 10 YPP.

I don't know how much of the difference between his work and yours is the different definition (a drive where you hit the defense for 10-15 yards on every play is explosive for him but not at all for you) and how much is the fact that he's removed all the garbage-time drives, and with a pretty liberal definition of "garbage time" that probably takes out at least a quarter of Alabama's season if not more.


Another interesting thing I noticed is confirmation of your comment about the generally poor state of defense in the SEC. Alabama was #9 in the country in offensive FEI but they were sixth in their conference. SEC teams were 1, 2, 5, 7, 8 and 9. On defense the top SEC teams were ranked 7, 8, 10, 11, 20, 23. Bit of a difference there.

Last year, and every prior year I've checked, that was the other way around. What happened?


January 9th, 2014 at 10:34 AM ^

Keep in mind that 10 YPP does not mean 10-15 yards on every play. If you go 20 yards on 7 plays (two third-down conversions and a TFL) and then get a 60-yard TD pass on play 8, that's an explosive drive.


January 9th, 2014 at 10:40 AM ^

But that's an explosive drive under both definitions. Fremeau has Alabama as an explosive offense, the Mathlete doesn't, so  I'm looking for a type of drive that's explosive in one and not the other. Moving the sticks on first down, but not by much, fills the bill.

Blue in Seattle

January 9th, 2014 at 11:46 AM ^

FEI is a drive based analysis, so having at least one play in a drive as "explosive" will label that drive explosive. When I read the Mathletes description my understanding is that he is rating each first down achieved by how many yards beyond the goal of a first down were achieved. In other words how many explosive first downs does a team achive, rather than explosive drives.

As a quantitative analysis I think this shows that Nussmeier achieves improvement that is independent of the state of the program. Granted it is only two cases on opposite ends of the spectrum, but how much data does any single coach generate? The fact is, in his last two jobs he created improvement on these metrics, with first unproven or limited talent, and then again with proven talent.

The question I ask myself is "why did Saban tire of him in two short years?" I assume Saban knows more than what we can analyze, and there is the fact that he hired him from Washington and must have done so expecting improvement, which he got. So why the change? My conclusion is that as soon as the master recruiter, Lane Kiffin, became available room was going to be made for him, and OC was the only position that wouldn't be an insult.


January 9th, 2014 at 1:44 PM ^

I had similar thoughts regarding why Saban would part ways with him.  Not that we would hear it, but Saban had nothing bad to say about Nuss and even acknowledged the production and explosiveness of the offense the past two years.  I think Saban knows that between he and Kiffin and the talent level on hand, they should be able to replace that schematic effectiveness and most of it's production.  Chances are with the graduation of McCarron they want to go back to a more run heavy and short/intermediate passing approach.




January 9th, 2014 at 5:02 PM ^

From what I've read, it wasn't so much an issue with his abilities or anything like that. Supposedly, Saban has a very abrasive personality in tense situations and Nussmeier couldn't deal with it like the other OCs could. If that's really the case, then Nussmeier should really have a good time here because Hoke rarely gets in the way of play-calling.


January 9th, 2014 at 6:55 PM ^

statiscally "explosiveness" is however you define it. neverheless, it would be interesting to see the YPP number paired with a measure of variance. maybe the standard deviation of the YPP or even an average of the median play for each drive.   you could get a better comparison of teams that have the same high YPP and understand if they are an efficient balanced offensive machine that just tends to get relatively consisent chunks, or if it's a big play generator. 


January 9th, 2014 at 7:02 PM ^

is get a picture of Nussheimer offense at Alabama. That it's explosive by Fremeau's definition but isn't explosive by Mathlete's tells me a lot more than either fact does on its own. The only scenario I can think of that fits is an offense that's spitting out a whole lot of 8-15 yard plays. If they were hitting home runs it would show up on the Mathlete's chart; if they were grinding it out like Wisconsin it wouldn't show up at FEI.

Ron Utah

January 10th, 2014 at 1:15 AM ^

Watching Alabama against Oklahoma and Auburn, you see a lot of drives where 'Bama is consistently getting 11-13 yards on 1st down, or 6-9 yards on 2nd and 5.

They move the chains like crazy and pop a couple of big plays every game, but they're not Baylor.


January 12th, 2014 at 12:43 AM ^

but not so much against Oklahoma. The drive list:

  1. hit a 53-yard pass, TD
  2. interception on first play
  3. hit a 63-yard pass, TD
  4. hit a 67-yard pass, TD
  5. long 11-play drive ended in a lost fumble
  6. 8 plays, 35 yards, punt
  7. 2-minute drill ended in missed FG
  8. 3 and out
  9. 43-yard run for TD
  10. 7 plays, 31 yards, punt
  11. 3 and out
  12. 3 and out
  13. 61-yard pass for TD
  14. first-play sack, fumble, returned by OU fir defensive TD

Every drive failed unless they hit a big play, but they hit a lot of them. They finished with 516 yards; more than half of those yards, 287, came on five of their 65 plays.

Baylor ground it out against OU by comparison--only two plays over 30 yards, none over 40, several successful drives w/o any big plays at all.



January 9th, 2014 at 10:06 AM ^

I agree that Hoke probably would rather be in the top left than the bottom right, but top right doesn't hurt either.

And if this is how you should build to win National Titles, how should you build to win the B1G? because that's our only goal /Drew Sharp'd


January 9th, 2014 at 10:24 AM ^

Given the history of improvement, would a conservative estimate for next year's offense be in the neighborhood of 2011 Washington/Michigan?


January 9th, 2014 at 10:53 AM ^

"Put a team together around an elite offense and you get 10 amazing games and 2 games where the wheels fall off. Build it around a great defense and you are probably in all 12 games."

I think of it a little differently. If you build your team around one elite unit, whether it is offense or defense, your other unit is going to have to win you a game or two each year. 

Take Michigan State this year: Their defense didn't play poorly against Notre Dame, by all standards they were great. But they lost because their offense couldn't do anything. But at the end of the year, when they needed points after their defense gave up 24 to Ohio State, the offense came through and scored 34 to win the game. 

Alabama's defense was fantastic last year. The offense still had to come through against Georgia. I'm sure we could go back and find similar examples from every season. 


Moonlight Graham

January 9th, 2014 at 10:53 AM ^

and the last half of your last paragraph was stellar ... It's why we haven't seen a full-out blazing basketball-on-grass offense like Kelly's Oregon, Leach's TTU, Briles' Baylor, or RichRod's WVa (among others) win the MNC. Oregon always seems to falter once or twice each year against a Stanford, and was held in check in the '10 NCG ... WVa steamrolled through the Big East in '07 only to play the clunker against Pitt ... so on, so forth. 


January 9th, 2014 at 11:13 AM ^

It's not like Alabama is immune to this sort of stuff - even with all their talent advantages they've only gone one year without losing a clunker, 2009. In 2011 and 2012 they were given the benefit of the doubt by the polls, but they still dropped clunkers. And I think Auburn 2010 and FSU 2013 belong in the basketball-on-grass offense category, but you left them out for some reason. I think you're cherry-picking your data a bit.


January 9th, 2014 at 11:28 AM ^

  • 2013 Florida State, #12 offense, #5 defense
  • 2012 Alabama, #5 offense, #4 defense
  • 2011 Alabama, #11 offense, #1 defense
  • 2010 Auburn, #1 offense, #8 defense
  • 2009 Alabama, #5 offense, #1 defense
  • 2008 Florida, #3 offense, #1 defense
  • 2007 LSU, #3 offense, #6 defense

It doesn't quite rise to the level of statistical significance but there's an edge to the defenses on that list.


January 9th, 2014 at 6:25 PM ^

I agree with Moonlight.  This is why Stanford has won the Pac-12 three years in a row now, even though it always looks like Oregon would run away with it.  The Stanford model (more similar to ours) is less explosive but varies less from game to game.  Stanford has fewer 40-point victories, but they also have fewer losses.  


January 9th, 2014 at 9:10 PM ^

I'm not totally disagreeing with you, but Oregon does have a misperception issue.

2010, Auburn #1 Offensive FEI

2010, Oregon #15 Offensive FEI

2010, Oregon #3 Defensive FEI

2010, Auburn #8 Defensive FEI


Oregon's defense was rated higher than Auburn's defense that year. It was bad fortune that Auburn had the most dominant short yardage DL, and most dominant short-yardage QB both on their team.


January 9th, 2014 at 10:58 AM ^

I had of Alabama's offense over the past two years. It was neither exciting or explosive, but it was very efficient. How much of that was the Saban influence is hard to tell.


January 9th, 2014 at 11:50 AM ^

You should change your name to Clichelete.

Just teasing.  Awesome work as always.  Looking forward to the elite D vs. elite O analysis.


January 9th, 2014 at 12:52 PM ^

It's more about variance than it is about averages.

Offenses like Oregon have gaudy numbers (averages), but it will take a good amount of luck to have everything fall into place to get 13 games of excellence (because of high variance).  Usually, you will get a bad game (or two).

Defense does seem to have much less variance, and thus you can rely on it more consistently.  As a result, you should put more of your chips into defense if you want a consistent winner.

But I teach intro statistics, so variance is always an issue in my mind. 

There's a lot of random variability going on in any football result that we like to attribute to our favorite scapegoat (OC, QB, headset-wearing, sitting in a booth, etc.).  I think that's why most coaches come around to the execution issue.  Improving execution improves reliability, reduces variance, and limits the effects of the random variables.

If you want a consistent winner, don't forget variance.

steve sharik

January 9th, 2014 at 2:30 PM ^

A lot of variables, each with a lot of variance = hard to predict.  Vegas sports books love it.

Any coach/football mind that can figure out how to reduce variance across all phases will win a lot of games.  The key is to figure out the true variables.  A lot of analysts think stats are the independent, input variables whereas they are truly dependent variables.  Winning depends on yards, first downs, etc., but those are dependent results on decisions and abilities of the players.

Baseball works better b/c the stats are mostly individually based, especially over a long season.  Football works just the opposite.  For example, what would Barry Sanders and Emmitt Smith's career rushing totals be if they were traded for each other in 1990?


January 9th, 2014 at 8:58 PM ^

And baseball also works better because there are so many more "observations."  Sample size is the issue.  Even with random variance, it's easier to predict season outcomes than individual game outcomes.

Thus, advanced stats are useful in predicting the likelihood of your baseball team making the playoffs, but once the playoffs start, it's a crap shoot (too few games, and thus random chance has a strong pull).

Football seems like baseball playoffs.  So few games.

And then consider the number of variables in any given play from scrimmage.  You have 11 players on either side trying to execute their assignments, play call, weather, down and distance, etc.

Eye of the Tiger

January 9th, 2014 at 4:53 PM ^

"Offense wins games and defense wins championships" (for unexplained, largely ethereal FEELINGSBALL reasons).

Rather, it's defense puts you in games more consistently than offense does, because of lower variance.


January 9th, 2014 at 6:07 PM ^

in general it is easier to consistently play good defense that it is to consistently play good offense.  I think that is true in many sports, including football and basketball.   Maybe because you can get good defense out of pure effort sometimes rather than needing great execution like you need on offense.

Of course, that ignores games like Indiana and Ohio State this year, where it looked like we didn't even bother to play defense. 

Brewers Yost

January 10th, 2014 at 9:08 AM ^

Defense atleast in football relies more on the unit as a whole. Offense on the other hand can depend on the performance of 1 or 2 great players.

I know I might be cherry picking but when Dennis Dixon went down for Oregon in 2007 the team totally collapsed; even getting shut out by a 5-5 UCLA team. Also, see Michigan-Nebraska after Denard went out. On the other hand defenses seem to be less effected by losing their best player, MSU without Bullough. Importantly, this holds true not just for injury/suspension but also having a bad game. Great offenses can look average when their star player is held in check.


January 9th, 2014 at 6:13 PM ^

You stated:

"In fact, if you compare 2008-2011 Washington and Michigan, every year but 2010 is very similar and demonstrate a lot of positive improvement."

2011 was the regression, not 2010. 2010 "demonstrate[d] a lot of positive improvement.


January 9th, 2014 at 6:49 PM ^

I think the biggest variable (unmeasurable) is the excitment this MAY bring to an otherwise downtrodden unit.  Can the guy coach, sure Saban trusted him. The question is do we have the talent on the roster that the recruiting guru's think we do. If that answer is yes then his play calling will fit just right.  He likes to get a lot of guys to the point of attack and out number you in the run game then sting you deep. So the issue is can we keep defenses honest next year w/ anything even remotely resembling a push up front.

03 Blue 07

January 9th, 2014 at 7:41 PM ^

I'm a moron, but I can't understand what this means:

The 82.7% first down conversion in 2013 was the third highest number since 2013. Some of that was due to the overall regression of defenses in the SEC in 2013. 

Does Mathlete mean the third highest in 2013? Or the third highest since some date other than 2013? I have no idea what the sample size is.