Makes me sad.  Someone needs to find AWESOME-O, or his accountabilibuddy, and cheer him up.

Rather than start with factors, i.e. returning starters and look at some results.  What are the 5 or 10 biggest leaps in overall D from one year to the next?  Then let's breakdown what was going on in all those scenarios to see if Michigan's mirrors that in anyway. 

I don't expect a top 25 D...I"m in the top 60ish or so camp.  The points have been made, but I am currently finding typing very soothing. 

A highly regarded D coord who will get to run the defense he knows how to; Players playing in their natural position; Martin getting help inside on the line (with healthy ankles); A more experienced offense that a) will hopefully improve upon their turnover rate and b) while they no doubt regress in yardage, I'd imagine they will increase their time of possession based on a) and the scheme;

Look no further than Sparty's secondary from '09 to '10.  statistically in '09 they were as bad as Michigan's last year (the same exact ranking in fact, 110th or something), in '10 they improved dramatically.  One unit, yes, but why/how did they make the leap?

How big is the variance in the average improvement data? Since Brian himself said it was based on very little data points extrapolated to make something meaningful out of it, the improvement metric that Connolly uses needs to have standard deviation associated with it. If less than 100 teams have been sampled, the error is greater than 10% assuming Poisson statistics (probably a very generous assumption in the case of F/+ improvement). If the mean change in rank is 33 according to Connolly, that's subject to a 4-rank error, putting Michigan anywhere from 78th to 86th.

Now assume that only 10 teams have been sampled. Then the error is about 32%, which means error bars of 10 ranks, putting Michigan from 72nd to 92nd.

If only 2 teams have been sampled (that is, only the two teams that have ranked 115th in defensive F/+ since the F/+ metric first appeared in the 2009 Football Outsiders Almanac), that is prone to a 70% error, giving a set of error bars of 23 ranks, putting Michigan anywhere from 39th to 115th.

In other words, this statistic tells us nothing unless there is some system of variance associated with it.

The deal is though that ending up at 82nd isn't the end of it.

On the other side of the field we should have a junior QB who turns it over less.  A MANBALL running game to kill the clock when desired.  The ability to kick freaking FGs.  In other words the defense should be put in fewer bad spots by the offense and hopefully we aren't playing from behind (and the negative mentality that comes with it).  

The offense may regress (I'd say likely) but hopefully the offense's ability to play Carrball (run off left tackle, throw short route to TE, eat up 7 minutes on a 40 yard drive) improves and the offense can from time to time help the defense.  In concrete examples during the Illinois game when we're trading scores, perhaps the offense manages to kill a massive chunk of time and thus exposes the defense to two fewer Illini drives.  The defense can improve because they're sitting on the bench drinking gatorade.

I enjoyed the statistical analysis.  The number of returning defensive starters is a contributor to what one could predict as an overall win/loss record, which is all that matters regarding on-field results.

I would like to see a similar statistical analysis, covering only the Big Ten over the last decade, regarding what we should expect as a win/loss record based soley on the average player weight for the whole team, i.e. a single-variable linear regression:

%wins = A * average player weight + B.

You could call it the beefy Wellman manball effect.

For really low-sample fun, though, nothing beats projecting what happens with a 2nd-year-starter dilithium QB, but in a new offensive scheme. Which has been the favorite topic of debate ever since we knew Denard was staying.  From my point of view, as long as #16 is our QB and leader, it is going to be fun to watch, easy to root for, and never hopeless.  Lord how I will miss him when he's gone. 

Blackjack has a finite amount of outcomes and a finite amount of variables.  This is why, if you're a math whiz, a computer, or Rain Main, you can predict with some certainty what will happen in a game.  Football has an infinite amount of variables and, depending on what you're looking at, an infinite amount of outcomes.  Football teams are also only equals in the sense that they are made up of young men who wear football gear and who take the field eleven at a time...All of this makes it border-line meaningless to look at what other teams have done in the past and then apply the results to a prediction of what Michigan will do. 

I expect Michigan to go about 6-6 this year, with the D and O both moving toward mediocrity for the time-being.  I don't think you can learn much of anything about Michigan by looking at other teams' improvements (or lack thereof), though. 

Didn't Syracuse improve drastically on D the year after GERG with Shafer and a new staff? I want to say 100 something to 40 something? I could be wrong and the change I am thinking of took two years.
<br>
<br>I expect realistically somewhere around the 60's to be realistic. Taking into account: Michigan D + 1 year + TWoolf + Mattison + Hoke + Competent Position Coaches + 4-3 Manball - GERG - Offensive Turnovers = significantly much more improved D.

that Brian missed.  I imagine we are being compared to teams in the past with horrible defenses and also low talent. I'm talkinabout the EMU's, Bolling Green's, NMU's, San Jose State's of the world.  Even when those teams return 9, 10, or 11 guys on D I can't imagine they improve all that much because their talent disparity against the top 60 or 70 teams is so big.  Michigan has a lot of talent on the defensive side (not as much as maybe in the future) and thus we have more upside.

I'd be more interested (screw the small sample size) about how top 40 (historically) teams did after having a bottom 20 or 30 defense and returning a lot of starters.  I just don't think it's relavant to compare us to teams with returning starts but very little talent.  Maybe if we went back 20 some years be could get enough data points to make is significant, either way it would be interesting to see.

Brian, or shall I say, Harlod Camping (aka "The Doomsday Preacher")...You aren't a very reliable source for predictions.

Extrapolation doesn't seem to work out so well in something as complicated as football. I'm more inclined to ignorantly bask in my misguided hopes until they're proven wrong.

Nebraska went from being one of the worst defenses, to respectable, to one of the country's best in two seasons once Bo Pelini took over, just by coaching 'em up, scheming to his talent, and moving people into the right positions.

There's some merit in your argument, but returning starters, seniors, and recruiting stars aren't everything.

I predict UM's defense under Mattison will improve just like Nebraska, if it does you'll contend for the Legends title.

...say that they can't wait for the season to start.

I can't wait for the season to start.

Perhaps the reason that there is no NFL talent is because guys haven't developed properly over the past few years in regards to weight gain, speed, not to mention....actually playing defense.  I'd like to see what Mike Martin's stock would look like if he had been in a 4-3 with a blitz minded coach and had racked up 15 sacks over the last two seasons.  After all, many of these guys were 4 stars when they came in, so they obviously had the tools relative to their peers at one time.  Some seem to call this "bad luck," but some of it has to be attributable to under achieving due to bad coaching.