How Much Of A Problem Is Michigan's Defense?

There is an interesting post on thebiglead about how Michigan is actually better on defense than the defensive efficiency stat indicates:

http://www.thebiglead.com/index.php/2013/01/23/michigan-plays-better-def...

Crux of the argument is that they are up among the nation's leaders in "shutdown stretches" defined as a 4 min period in which the opponent does not score. The article doesn't speak to when these stretches occurred, so some of these could be biased to lower ranked opponents or less critical game situations, but it's an interesting stat...

Thanks for taking a look at Winn's column. I usually like his articles but was annoyed by how he wrote off Michigan without talking about how good the offense has been. His argument was way too simple and I am either too lazy or dumb to fisk it.

Instead, I settled for the reductio ad absurdum argument of a theoretical team with a bottom-50 defense that always scores. They can't lose but would be eliminated using his criteria.

Florida's going to be an interesting case for advanced stats. On the one hand, their efficiency stats are remarkable, and they're housing nearly everyone they play. Their closest win in 13 points, and their next closest is 17. On the other hand, the two times they've been in competitive games, they've collapsed down the stretch, including an epic collapse at Arizona when they led by, IIRC, 3 scores with under a minute left. This pattern mirrors their struggles in close games for the last few seasons.

The weakness of the SEC means that their efficiency numbers are likely going to remain off the charts. They only have 2 RPI top 50 and no top 25 games left. Their remaining conference schedule includes 7 games against RPI 100+ teams. The SEC is bad, man. It also means they're not likely to have a bunch of experience in one of the key "intangibles" that gets thrown out by skeptics of advanced stats, ie, "handling pressure,' "clutchness," whatever you want to call experience down the stretch in close games. 

And a one-off tournament is obviously not the best measure of the predictive success of statistical measures. Florida's combination of blowout wins and losses in their only close games does make me skeptical of how well they'll react when they play tournament opponents who aren't  overmatched and have a lot more experience in late game scenarios.

Would be interesting to revisit these margins right before march madness. Please do this. Might be worth filling out a bracket using this methodology to see how it shakes out.

I have seen them go on crazy runs, but the thing that stands out to me, and something that isnt measured in these other ranking and stats is this teams crazy ability to not allow the other teams to sustain runs. outside of the ohio state game nobody has really bee able to sustain runs against us. a big part of that, I feel is our #1 offense that just doesnt go cold for long stretches because somsone always come up with that dagger shot to kill a run.

That is a HUGE advantage in college because its based so much on runs and momentum, that and we are still a young team playing WAY more mature than our age. grinding out B1G games is going to go a long way come tourney time.

I also have seen that some of the easy second chance points teams are putting up on us are partially because we are small underneath and are are forcing so many outside shots that since we already out on the perimeter our wings are just conceding that if our bigs dont get the board they are on thier own selling out completly that if our bigs do get the boards the race is on. which also contributes to our #1 offense.

It will be interesting to see if we move to a moar balaced approach as the B1G season goes on or if we keep pushing the way we are.

Either way, this team is freakishly athletic and VERY careful with the ball (sans Timmy), unlike a lot of the other teams of the elite eight of recent years.

For instance, that 2009 NC NC team finished 16th in ADE but it's number 89.6 is very close to Michigan's and would place them currently 38th -- would they somehow be ranked higher this year?  Would their defense not come out to 89.6 this year?

Could you plausibly set up the limit as being achieving a certain number on defense vs rank?  Maybe there are more offensive and defensive oriented years?

Sometimes your offense leaves you (Nebraska, OSU first half), but defense rarely does. Now sometimes opposing offenses go off regardless of defense (Ohio last year), but the point is still relatively solid. Not having a great defense doesn't necessarily mean you won't go deep into the tournament, you just can't have an offensive let down, making it much harder over a stretch of games. If you're the other way around (top notch defense, mediocre offense) you are a more consistent team, but maybe not the upside from one game to the next.

Brian, can you post the raw data (google docs?) where you got the R values from?

...MGoStatistic'ed.

I think it's interesting to note just how balanced most of those NCAA champions are.  All those NCAA champs rank in the Top 20 both offensively and defensively and 6 0f the 10 were ranked in the Top 10 in both offensive and defense. 

"You could hack off 3.5 points of that, toss it on Michigan's defensive numbers, and come up with a pairing of the country's #1 offense with the 16th-best defense"

I took this a step further: Keeping their AdjO in the top ten. #10 ND is 8.3 points behind UM. Lop that off their AdjD and you see UM's offense at #10 with their defense becoming #3 in the country. It seems their offense is truly making up for their defense on an epic scale.

Another aspect of this that leaves me somewhat optimistic is that we are still so young. People forget that we are 338th out of 347 teams in experience per Kenpom. Hopefully some of these young guys like Stauskas, GR3 and McGary will start to see the intensity and physicality of defense now that we are playing OSU and MN etc and try to match it in their own games. Not sure that intensity is something you can re-create in practice against guys you play everyday. You would have to think that such a young team would be more likely to improve over the course of the season than would the avg team; although it would be interesting to see if any data backs up that theory.  

Also, Burke was clearly more active defensively off the ball against Minnesota than he normally is. My hope is that after struggling against Craft again that perhaps he will try to take a page from his game and re-dedicate himself to improving his defensive game just by being more active on D. 

What is factored into those "adjusted" offensive and defensive efficiencies? Can't you make the argument that Florida's efficiencies are artificially high given how horrible their conference is?

I don't really think that a regression analysis is appropriate here.  It seems like you are "plotting" maximum NCAA Tourney Round (assuming the elite 8 is round "zero"...all that means is that the y-intercept changes, nothing meaningful would be added by counting the round of 64, 32, 16...) vs. AOE, DOE, and Delta OE and comparing the correlation statatistics for each group.

I'm thinking that what you really want to look at is _effect_size_.  http://en.wikipedia.org/wiki/Effect_size

You would then compare the AOE, DOE, and Delta OE for the elite group in Winn's sample to the entirety of the NCAA (no need to add this year's teams).  Determining which statistic has the largest effect size seems would more strongly test your hypothesis. 

You could also see how effect size changes when you restrict the meaning of "elite".  That is, what happens when you look at teams who made the FF (instead of just the EE)?  What about just the RUs and NCs?

This does, however, address a bigger problem with Winn's analysis.  Winn used ranks, and that was the wrong choice, especially when the difference between 1 and 10 might be much larger than the difference between 51 and 60.  Being an outlier could be very helpful.

With every game ever invented, the person who wins is the one who scores the most.  Being more offensive always is what the final result is judged on.  Saying "defense wins championships" is just the equal statement to the double negative "lack of offense loses championships".  The cliche has been crap since the first time it was uttered.  It gains momentum because sports writers have to write about something, and interest is generated by taking opposing view points whether they are justified or not.

The most memorable anecdotal situation that demonstrates this in my memory is the Patriots versus Rams super bowl.  This super bowl put a team with a stronger defense than offense (Tom Brady ended up the starter when Bledsoe was injured and out for the year) against a team with a stronger offense than defense.  The game started out strong for the Patriots are they defense was able to blank the Rams offense, and even generate points (at least one pick six in my memory without googling it.)  The Patriots always had a excellent field position in the beginning of the game, and the Patriots offense conservatively managed the ball, and put up more points.  BUT, eventually the defense was wearing down, and the Rams offense got on a hot streak.  The Patriots offense had now response, and with less than two minutes (maybe even less than one minute) left the Patriots got the ball with the game tied.  John Madden made the cliche statement that every announcer makes when a team has the ball deep in their end.  As he was saying, "the Patriots will clearly take a knee here and go into overtime" Brady drops back to pass, and through a series of quick short passes to the sidelines, he was able to get Vinatieri into range (which of course is like 55 yards even in snow) and the Patriots Offense won the game.  Belichek knew that this was his best shot.  Going to overtime was a 50/50 chance of getting the ball first in the exact same position.  His defense was not going to hold up, because it hadn't held up for the remainder of the second half, and certainly not at all in the 4th quarter.

Another anecdote very similar is of course the 2011 Michigan - ND game.  And in that game the programmed announcer response was that "clearly the coach is going to take the safe play".  And yet in a decision of offensive bravado, Hoke took his shot, and the rest is Denard to Roundtree history.  Offensive game winning history.

I think the analysis shows that Michigan is a young team with a lot of talent.  That youth has a red flag on it's defensive skill, which is something that gets better with hard work and time together as a team.  I don't think this article is pointing out anything John Beilein didn't already now when he recruited these players.  But every game they play is another game they get better at Defense, and everyone except Trey Burke gets better at offense. ( I mean really, how much better can Burke get?  Nebraska had to triple team him to take him out of the game, and OSU is the only team with the defensive skill to take him out with double teams).

Yes statistics are great for predicting what is likely to happen to a population in general and with enough data points and over a long period of time.  But each event is it's own unique probability.

And that's why they play the game.

It was briefly mentioned above, but is anybody else seeing dollar signs flashing right now when it comes to this study and filling out a tournament pool. I mean, the correlation between advancement in the tournament and the delta in AdjO and AdjD seems like a gold mine. I'm going to use this as a baseline this year when making my selections.

If I win, I'll gladly donate 10% of my winnings to MGoBlog or something else of Brian's choosing.

in the sucky Sec and not the mighty Big 10 - gotta be something to that I would think.

Besides it often comes down to matchups - some teams match up better with other teams - obviously if you can disrupt Trey Burke as Craft did then Michigan becomes very pedestrian offensively.