dnak438

Thanks to MGoUser Enjoy Life, my attention was drawn to the statistics available on mgoblue.com. This has allowed me to increase the historical depth of my analysis. (As usual, click on each graph to embiggen).

We'll begin with scoring defense, the strength of this year's defense. (I've removed the numbers from the WMU game):

This is incredibly impressive, especially given that this defense isn't nearly as good when it comes to yards per play:

Although the average yards per play have improved (note especially how much better we are against the pass), they are still not as good as the averages from the Carr years, which basically were 5.0 yars per play or better (excepting 2000).

Finally, here are the statistics for total yards per game. In order to make the curves comparable, I've expressed each as a percentage of the 2010 stastics. The drop-off is pretty intense. (Yes, those numbers from 2006 are right. Michigan allowed 43 rushing yards per game that year.)

The overall result is that this defense is (overall) about average for a Michigan defense during this span:

Here is a link to my spreadsheet, based on the statistics from mgoblue.com.

Since the defensive improvement has been far and away better than anything any of us could have expected, I thought it would be interesting to see the extent of the improvement in historical terms. I collected data on scoring defense and yardage defense from Rivals as far back as 2003 to the present. Here are the results (click on the graphs to embiggen):

And, to compare the two data sets, I've superimposed them by representing each as a percentage of the worst historical performance (i.e., 2010):

The result is striking. Thus far, Hoke, Mattison and the rest of the defensive staff have turned this squad into one of the best 3 defensive teams in the past 9 years.

UPDATE: MGoBlog user Mat suggested that I look at our defense's yards allowed per play. I did some googling and found stats dating back to 2003 foryards per play. The following graph is based on statistics that only take into account performances against other FBS teams:

Which confirms Mat's impression (and ours) that while this defense is excellent we are not yet elite. That's not surprising given that it's year 1 of yet another defensive system and that we are starting two freshmen.

When we weren't getting bubble-screened by Northwestern, we were defending the option with mixed success. I was curious about what was happening and I decided to picture-page some plays to satisfy my curiosity. I'll be interested to see how Brian treats it in his UFR. At any rate, here is a play from Northwestern's first drive (click on images to embiggen):

Colter has just gone in for Persa. NW is going to run a veer (I think) option, as follows:

As the interview with Mattison suggests, the ends have the quarterback, Kovacs has the pitch man. But the linebackers need to flow to the ball, and all too often -- to anticipate my conclusion a bit -- both of the linebackers would jump the first option, allowing the QB to keep and get to the outside. Here are the next couple of shots:

and

You can see that both Hawthorne and Demens have been sucked inside, and Roh too has lost contain on Persa. I've never played organized football, but I think that Demens needs to respect the quarterback keep here. At this point, Roh, Demens and Kovacs have all realized that the QB still has the ball and are flying to Kolter.

The image above is the moment of the pitch. The Michigan defenders have made up ground, but they are going to give up five yards on this play.

The second play is on NW's first scoring drive (second drive overall). It is second down and 6 on the 15 yard line, Colter is again in for Persa. [UPDATE: this has also now been picture paged by Burgeoning Wolverine Star].

NW will basically run the same play:

The right side of the NW line lets Will Campbell and Jake Ryan through and blocks/seals Demens and Hawthorne. Here we are at the mesh point:

Will Campbell and Jake Ryan are through, and have a free shot. You can see that both linebackers (Demens and Hawthorne) have taken a step or two forward and towards the hashmarks. Hawthorne is about to get sealed by #70.

The problem: four Michigan defenders tackle (or head towards) the first option, the dive:

This is NOT good. Hawthorne and Demens are engaged and can't get out to the edge. Kovacs is streaming to the QB now, but a split second later, you see the result:

Kovacs is in a real bind: one defender on the QB in space with a pitch man. But Kovacs misses the tackle anyway:

The result: TOUCHDOWN.

My diagnosis is that the defenders were too eager to get to the first option, and forced Colter to keep. This was bad insofar as the DEs didn't keep contain. On the second play, Ryan has to let Will Campbell take the first option and hit the quarterback, forcing a pitch. But Ryan and BWC tackle the same guy, giving Colter a free release.

Next, Hawthorne and Demens can't (I think) both step towards the center. This gets Hawthorne (the playside LB) sealed, leaving Kovacs one-on-two. I noticed another big option play where Hawthorne got sucked in, and I wonder if this is why he was pulled for Morgan (until the 3rd quarter, when Hawthorne went back in).

We must have corrected this in the second half--but that's the topic of another diary (or more probably, Brian's defensive UFR).

Again, let me plead ignorance; if the more football-savvy among you can tell me if I'm wrong (or right), and why, I'd appreciate it.

MGoBlog user the fume suggested in the comments of my last diary that Kovacs' interception in the 2nd quarter came from the same defensive play call that led to Notre Dame's final score, so I thought that I'd satisfy my own curiosity and look at that play too.

It's 3rd and 9 at the 36 yard line, ND 14-UM 0. Michigan brings all of its defensive personnel to within 3 yards of the line of scrimmage, with three down linemen (Black, MM, RVB), three linebackers (Hawthorne, Demens, Ryan) and five defensive backs (Avery, Gordon, Robinson?, Kovacs, Floyd). Here's the look:

Rees checks into a new play. This is the play he checks into:

He's focused in on Floyd at the bottom of the screen. Michigan, however, is going to rush 3 and drop 8 into coverage. It's a 3-deep zone coverage with five players in the short zones:

The pressure on the defense will come at the bottom of the screen, since the slot receiver will run a seam route straight up the hash marks, and Floyd is running a 12 yard curl route. Although the seam route is open, Rees apparently expects man coverage, as he seems to have decided already to throw it to Floyd.

The seam route is open. If Rees sees it, it's probably a touchdown. Note that the safety playing the center of the field is at the 44 yard, on the other hash marks, running like crazy up field. He's the only defender who has a chance at preventing a touchdown if Rees sees the seam. 

But Rees is staring at Floyd, and Kovacs is watching Rees' eyes. Here's what it looks like when Rees starts to throw, with Black breathing down his neck. Note how open the slot receiver still is.

Rees compounds his error by making a poor throw too far inside. I suppose he thinks that Kovacs is running with the slot receiver and that Floyd will be wide open.

The result: Kovacs makes the pick.

It's a great play by Kovacs. This defense clearly has a problem, however. The player who plays the center deep third of the field needs to turn at the snap and get deep. He doesn't have time to survey the field and adjust his position. Here it's even worse than on the touchdown in the fourth quarter, as the deep center defender is on the opposite side of the formation from the receiver he needs to cover.

As several people commented, Mattison is not adverse to taking risks with his defensive play-calling to generate turnovers and uncertainty. Sometimes, as here, it works. Sometimes it doesn't. But I think all of us are happier with this approach than the very passive approach taken by the past coaching regime.

Here's the video (1:00 mark):

[Ed-M: Gord morning. No it wasn't a dream. Read this. Also: AIIIIIIIIIEEE!!!]

I was curious to see what Mattison dialed up on Notre Dame's last score, to see what he was trying to accomplish and what went wrong. Here is what it looked like:

If you count Michigan and ND's players, you get to 10: there must be another WR at the bottom of the screen, covered by Troy Woolfolk. Michigan has everyone near the line of scrimmage, but the call is actually a Cover 3 and they will rush three defensive linemen, leaving 5 players to play the short zones:

I think that Woolfolk's assignment is the deep third at the bottom of the screen, but thanks to ESPN we can't see him. Here is what the defense looks like right after the snap:

You can see the three rushers, four of the five short defenders, and two of the three guys trying to get deep.

Notre Dame is going to run the following play:

Floyd is in the slot, and is presumably Rees's main target since it is third down and they need to convert (although it is obviously four-down territory).

The result of the play we all know.

(The play starts at 2:24)

I don't know anything about football beyond watching and reading mgoblog and smart football, but I think the idea of the call is this: by putting all our defenders close to the line of scrimmage, to bully ND into checking into a play that involves a quick pass (remember it's 3rd and 5). Then you rush 3, flood five players into the short zones, hopefully allowing you to break up the pass or make a tackle before the first down markers. The problem was that Rees didn't force it to Floyd, who was covered by Jake Ryan; instead he threw long, and Woolfolk and Marvin Robinson don't cover Theo Riddick.

I think that Mattison's call was sound; either Woolfolk or Robinson should have had Riddick (although it's hard to be sure since we can't see the whole field on ESPN's feed). The problem is, as Dr Saturday and Chris Brown of Smart Football pointed out,

In my last diary I suggested that there might be some room for extra defensive improvement due to the upgrade at DC from Greg Robinson to Greg Mattison. Although I am in general agreement with Brian that massive improvements in the defense should not be expected, I began to wonder what was possible--that is, in the past 5-6 years, has a team improved its defense by leaps and bounds? To that end, I looked at scoring defense ranks of all 120 FBS teams from 2003-2010 to see how teams improved from year to year. Based on the numbers at Rivals, here is how the data shake out:

Note: the x-axis represents changes in rank (negative is good), the y-axis number of examples (out of 840 [120 teams * 7 years]). So the distribution is more or less normal, with a change of 80 rank positions (in either direction) being the maximum, more or less. The largest improvement in our dataset is 94 positions, so if that is the maximum possible then Michigan in 2011 could move up from the 102nd scoring defense (in 2010) to 8th (in 2011). HOORAY!

I had originally suggested that this level of improvement was unlikely, but turd ferguson pointed out that my percentages were misleading, because middling- to highly-ranked defenses simply cannot improve by a large margin. Looking at teams ranked 91st or worse in scoring defense, then, we get the following chart:

You can see that teams with bad defenses improve 20 ranks on average, in part because they have more room to improve than they do to regress. 31% of the time teams ranked 91st or worse improve 30 ranks or more; and 17% of the time they improve 50 ranks or more. To get into the top quartile of defenses, a team ranked 102nd (like Michigan) needs a 70 rank (or more) improvement, which has happened 5% of the time. Looking at the teams with huge improvements, it is difficult to generalize about how they did it. Here are the most improved teams in each year for which we have data (bolded numbers represent the year in which the big improvement was made):

In some cases they seem based on the emergence of a superstar player on defense. For instance, Suh for Nebraska in 2009, which jumped from the 84th scoring defense to 2nd, or Von Miller for Texas A&M, which jumped from the 104th scoring defense in 2009 to the 27th in 2010.

In other cases you have teams that are consistently fairly good who for some reason have a collapse but then recover to their old form. UConn, for instance, usually has a pretty good scoring defense. In 2005, they were 21st, and in 2007 they were 11th in the country, but in 2006 they were 94th. Likewise, TCU has a pretty amazing scoring defense but in 2004 they were 106th in the country. The year before they were 27th, the year after they were 12th. They had some NFL talent, but all 2nd day draft picks or free agents.

Michigan is obviously not in the second type of team. Our defense hasn't been top 20 since 2006. It seems likely that for Michigan to have a good-to-great defense next year, something unexpected will have to happen. The most probable in my opinion is that one or two of our defensive players becomes dominant. Note: my excel spreadsheet is available for download here.