OysterMonkey

[Ed-M: Bumped because this totally punctuated my equilibrium. The best indicator yet of year-to-year defensive evolution. And great news: the mean has magnetism!]

Richard Goldschmidt hypothesized that the incremental changes to organismal phenotypes over the course of even thousands of generations was insufficient to explain the change from one species to another. He posited that evolutionary change is powered by great leaps forward, instances of saltatory mutation that generate a new species from the old. Goldschmidt’s ideas were ridiculed, mostly, and with good reason. The overwhelming evidence of population genetics and the theoretical triumph of the Neo-Darwinian synthesis seem to indicate that evolutionary change is effected gradually over time by the additive effects of allele substitutions in the genetic makeup of a given population; population change happens slowly, if at all.

But there are situations in which sudden changes to an organism’s ecological niche—a new predator or prey introduced, migration or population bottlenecks, climate change, a massive meteor falling from the sky and killing all the dinosaurs—opens up the opportunity for rapid (on the geological time scale) evolutionary change.

The defense was bad last year. And bad the year before. And the year before that. A number of reasons have been put forward for the awfulness. The defense was decimated. Really decimated. Seriously, it was decimated. GERG is a force of nature complete with his own effect. The coaches thought making in-game adjustments was tantamount to cheating. And so on. At the risk of overstraining the metaphor, it certainly felt as if we were watching the extinction of that species of animal previously known as the Wolverine defense. It’s at the very least an endangered species. But if the combination of the addition of Hoke and Mattison, Nebraska joining the BIG, and the tattoo-laden implosion of the 614 area code don’t count as a change in the environment that opens the possibility of rapid change, then my metaphor has no validity at all.*

Folks have tried to take a stab at what might happen this year, based on small sample sized studies of returning starters, even smaller sample sized bits of anecdotal evidence, and a healthy dose of Hoke-A-Mania! I collected data from http://www.cfbstats.com/ on total defense numbers from 2006 through 2010 and analyzed year to year changes for every team, based on total defense rankings. Even though I’ve got five years of data, I’m going to talk in terms of “Base year” and “Year 2;” since I wasn’t looking to find multi-year trends in defensive performance all I care about is the movement from one year to the next. So with five years of data I have four years (2006-2009) worth of data in my “Base year” set and four years (2007-2010) in my “Year 2” set

This diary doesn’t propose to do anything other than aggregate a little bit of data about what we can expect based on very recent history and to show how many teams over the last few years have been outliers. From there we can start to see what Michigan’s chances are of bucking the odds of Darwinian uniformitarianism.

Natura non facit saltum: The Case For Phyletic Gradualism

My first task was to look at the aggregated data on a very coarse grain. I wondered how much movement there was in rank from year to year, so I grouped teams into sets of ten based on their base year finish (top ten teams, teams 11-20, etc.) and then tracked where those clusters of teams finished on average in year 2.

The result:

So the 40 teams in the data set that finished in the top ten in the base year averaged a finish at around 20 year 2. If a team finished in the 111-120 rank range, they could expect to be at around 95 in year 2. The obvious thing that jumps out is regression at the two ends of the line. This suggests what should be obvious: it is difficult to sustain excellence or ineptitude. So, by staying terrible last year, Michigan is already an outlier. Yay? But as you move away from the ends of the line, the movement away from the base year gets less and less, so that teams that are average appear to stay average.

Then, since I care mostly about one of the teams at the gruesome end of the line, I looked more closely at teams that finished the base year in the 90-120 range, and got this for my troubles:

This looks at every spot in the ranking from 90 to 120 and plots the year 2 average for the teams that finished at each of those spots. There is a lot of noise here, because for each ranking spot there are only four data points, but the trend line is pretty much what we’d expect. The worse you are in the base year, the worse you can expect to be in year 2.

So the numbers look gloomy, suggesting that expecting much movement in one year is a recipe for disappointment. These numbers provide the baseline for the geological timescale. The pace of change appears to be slow.

Hopeful Monsters: The Case for Saltationism

Despite this evidence of evolutionary stasis there have been a number of teams who’ve managed macromutation from one year to the next, both up and down. Since 2006, 37 teams out of a possible 278 (obviously only teams ranked 51 or worse could possibly make a 50 spot leap) have managed a leap of 50 or more spots in the ranking from one year to the next, and 107 out of 378 possible have made jumps of 25 or more spots.

For what it's worth, these percentages are higher than I expected prior to compiling the numbers. It's not worth anything, by the way.

My original goal was to analyze the factors that these saltatory leaps might have in common, but finding reliable data on returning starters, experience, changes to coaches or defensive co-ordinators, etc. has proven difficult. I might try to look in detail at a few case studies to see if there are any similarities between Michigan 2011 and the hopeful monsters who point to the possibility of rapid change, but provide a link to my table so that anyone else who may want to can do the same.

Viva la evolucion.

*Yes, I’m aware my metaphor already has no validity at all.

Edit: I think this is what the first commenter is asking for.

[Ed-M: Bumped anyway!]

I was going to put all this into a diary and make it totally clever and informative and interesting before the tournament really kicks off, but I'm not going to have the time to do that so, in lieu of that, some unanalyzed charts for your pleasure.

I got all this information by compiling data from running searches at this database: http://projects.washingtonpost.com/ncaa/mens-basketball/search/.

First I recorded the winning percentage of all 16 seeds in each round of the tournament:

Then using this I calculated the percentage of chance a given seed had to get to each level of the tournament:

Using this I calculated the expected wins for a team at each seed:

And using this I calculated the expected wins that each conference should get in this year's tournament:

So, the B1G's 7 teams should total 9.28 wins based on seeding to meet historical expectations. Since Michigan is going to win six, I don't see this as being a problem at all.

There is some discussion on the internets about the UofM match-up against UConn, with the near unanimous thought seeming to be that Connecticut has a decent to good defense against the run (ranked 45^th last year in opponents' rushing yards per game) but is susceptible to offenses that pass well (ranked 88^th in pass yards per game against).

Here are some UConn defensive yardage stats to back that up (FCS tomato can free):

Now, I watched UConn in two games last year (vs. Cincinnati and WVU) and they seemed exceedingly terrible to me against the run. This made little sense to me, given their fair to middling numbers. Then I engaged my similarity noticer and noticed a similarity between WVU and UC. They both run versions of the spread offense. So I decided to contrast UConn's defensive stats against spread and non-spread teams.

According to the Worldwide Leader 48 teams run spread offense sets at least 75% of the time. UConn played four of these. Here's how their defense did:

For comparison, here's how they did on average against the 8 other FBS opponents they faced:

Passing numbers are similar, but the rushing numbers are definitely not.

Lest we think that the spread offenses they faced were simply supergreatrushingmachines, here's how these teams did on average for the year against everyone:

I didn't take the time to remove FCS opponents or UConn stats from these numbers for two reasons. One, I am very, very lazy. Two, removing those numbers would make these averages lower still, rendering my point even more forceful. I don't like to be forceful.

The upshot of all this is that UConn actually did a little better against the pass (in yds/att) than these teams' other opponents (except Cincinnati, who just went batshit crazy all over UConn's defense). But three of the four teams found running against UConn a less stressful affair than they encountered the rest of the year (by more than a full yard per carry) and USF was basically level with their average.

From this I conclude: yeah, UConn kind of stinks against the pass no matter what offensive scheme they're up against. But they really struggle to stop the run against teams that run from spread formations. Spread teams averaged 65 more yards per game on six fewer carries against UConn than did their old timey non-spread counterparts. That turns out to be over 2 and ½ more yards per carry on average. And I know that just as there isn't one “non-spread” offense there isn't just one “spread” offense but it's not like they were seeing teams that chuck the ball 70 times a game and run once a quarter to keep the defense honest. These teams ran the ball almost 32 times per game against UConn. Even the typically pass heavy offenses like UC and Baylor ran a lot. I conjecture that this is because they could.

If there's someone out there that has watched a significant amount of tape of UConn on defense, I would like some insight as to what might be the cause of this large discrepancy between spread and non -spread teams against the run. The little bit that I looked at on YouTube basically just looked like UConn was in a pretty standard 4-2-5 nickel package against WVU and UC, but I don't know too much about scheme and formation and so forth. A lot of what I've read about UConn indicates that their linebackers are a strength of their defense and their secondary a weakness, so could this simply be a matter of taking a starter-quality player from a good unit off the field and replacing him with a guy that can't crack the starting lineup of a lousy unit?

This makes me feel pretty good about the offense's ability to put up some good numbers Sept 4^th,, at least until someone explains to me why all of this is misguided, and how UConn's front seven will stonewall us. Then I can go back to worrying about things I can't control.