Do these FEI rankings look at past standings to compare teams across conferences? Our rankings may be skewed because the B10 is horrific this year.
FEI Rankings: Offense and Defense Edition
[Ed-M: diary bbbbbump]
Back in August, MGoUser "Undefeated dream season of 1992" did this fabulous diary on predicting a team's FEI rankings using previous seasons rankings and combining it with regression to the mean, returning starters (and bonus to an offense for bring a quarterback back), coaching change, and Rivals star rankings over the previous three recruiting cycles. This model predicted that Michigan's offense would settle in at #16, down from #2, and the defense would move up from #108 to #71.
Here we are after Week 7. The first set of FEI rankings specific to offense and defense have been released. Our Wolverines are ranked #15 on offense and surprising #17 on defense. The offense is right where UDSO92 predicted it would be (well, one spot higher). However, how did Michigan so dramatically improve on defense when it was only supposed to jump up to around #71, especially considering that the model has generally found that coordinator changes generally have a negative affect on the defense? Let's look at what UDSO92 himself had to say at the time:
One year is not definitive (except in the case of GERG, natch); in fact, a team that woefully underperformed the previous year could look great just by rebounding the following year.
What exactly were the expectations of the Michigan defense in 2010? The 2010 defense was predicted by the model to finish 46th. GEEEEEEEEERG indeed. Futher, over the period of 3 years, Michigan's defense underperformed by a whooping 37 positions of FEI. Now while this might not be an empirically correct thing to do, let's assume for a moment that we can adjust Michigan's 2011 FEI expectation. Let's say that perhaps Michigan was an exception to the general rule that changing coordinators hurts your defense in the immediate season, just because of how horrendously poorly the defense appeared to have been coached. Using this idea, Michigan's predicted finish of 71st can be improved to 34th (the difference of the 37 positions) by removing a product of the underperformance (GEEEEEEEEERG). Considering the previous seasons projection, the fact that Michigan has a good recruiting profile, along with 9 returning starters, this seems like a reasonable enough projection.
The rest of the increase up to 17th, while it may be an early season abberation, is quite possibly due to the fact that our new Greg, Greg Mattison, is not only a good coordinator, but an excellent one and is not a stuffed animal waving lunatic. Now, had someone tried predicting this at the beginning of the year, they certainly would have been called crazy. However, I figured it would be nice to throw out that little bit of overperformance as some actual evidence that not only was the previous coaching staff bad at coaching defense, but that the current one is good (based on the fact that they are overperforming their expectation).
That being said, the offense is exactly where it was figured to be, so despite our disappointment over the MSU game, let's remember that it was the first real road test (Northwestern was not a hostile crowd), and against FEI's #5 rated defense.
Let's take a look at the 7 opponents Michigan has faced thus far to see how they rank compared to our teams performance against them.
|Team||Michigan Yards||Avg/play||Opp Defense (FEI)||Opp Yards||Avg/play||Opp Offense (FEI)|
And now for what we'll be up against the next five weeks after we return from a bye.
|Team||Offense FEI||Defense FEI|
If you made it this far, thanks for reading. Also, mods, if you want, please move this to diaries if you deem it worthy. I would put it there myself, but I typed the whole thing out on the Board and can't copy for some reason. So I'm just gonna post it here. Move as you deem worthy/not-worthy.
are opponent adjusted
Just when I wanted to lose hope and feel pessimistic about my passion for Michigan football, you've made me feel good again. Curse you
a large part of me is still skeptical, but I think that's due in large part to the emotional damage the football portion of my brain has suffered over the past few years rather than any projectible piece of data on this specific team.
kindly around these parts.
How is the opponent adjustment calculated across different leagues?
defer that question to someone who has a deeper understanding of FEI calculations. Hopefully, when more people have seen this diary by tomorrow, we can get someone to respond to that question.
FEI is a drive based metric. S&P is a play based metric.
Here is FEI: http://www.footballoutsiders.com/stats/feidef2011
Here is S&P: http://www.footballoutsiders.com/stats/ncaadef2011
The opponent adjustments are fairly complex and are (by definition) subjective.
S&P currently ranks M's defense as #39.
Since the basic data is the same, this is primarily due to opponent adjustments. Here are 2 more references at Football Outsiders with lots of gory details.
Nebraska's defense, despite all their preseason hype, sits at 74th while their offense is 20th.
It looks like Michigan plays nothing but teams flawed in one way or another from here on out. If fans of these teams are looking at Michigan's numbers (15,17) they should be worried about facing the Wolverines, not the other way around.
Lucky to beat ND by these numbers. Then again, they are only numbers and that's why they play the games.
I am most worried about Illinois and Akron State looking at these stats. Our performance against MSU was not very good. MSU's D shut us down, both Illinois and Akron State look to have a defense good enough to do the same. I am still concerned with Nebraska, sometimes the stats don't tell it all.
I'm hoping the differences between the games will be these
- Not played in tornado (although November games could bring rain and cold that could affect also) so oh noes passes and back sholder throws ;) are not over thrown
- OSU is played at home (and hopefully Illinois is not as wild an atmosphere as Sparty vs Michigan) so we can vary the snap count better to avoid A-gap timed death blitzes
- Trust in coordinators to adjust. We've seen it done so far, hopefully they recognize mistakes made and lean on the strengths of the offense
Illinois is interesting, particular on the defensive ranking. Considering the basketball scores we had last year, it'll be interesting to see just how much their defense has improved. Or how much our offense has fallen off.
actually had a huge jump last year in FEI. Up until the game against us, they were rated very high. I think our offense was really good last year.
It's my general feeling that people get waaayyyy to up over games like Notre Dame (we really should have lost but for the Jeremy Gallon cloaking device) or NU (where we showed some serious flaws) and waaaayyyy too down on the MSU game (one of the top defenses in CFB, very physical rivalry game, first road game in a trash tornado, more tonado-y than usual typical trash tho).
We went more or less down to the wire with Sparty and could have tied the game if we converted that 4th and 1. The FEI is legit, my only concern would be that the B1G is down in terms of strength with the notable exception of Wisconsin. If Jugaloo State University hangs with Wiscy then we have reason to be very optimistic about our performance this past weekend - ALL THINGS CONSIDERED.
Hoke, Mattison and Borges know what they are doing. We've faced our toughest test of the year and failed, but failed nowhere near as badly as we failed against Wiscy, Spartina, O$U or Miss St last year. We're clearly headed in the right direction due in large part to the defense.
Last year after 7 games the M defense was ranked #83 by FEI and then, well, you know.
We still don't really know how good this defense is. The next 5 games can go either way. I am still very very nervious.
Not sure why people are nervous about this D. They are pretty damn good at what they do. With the offense being more of a ball control O and the defense making teams beat them with long drives, I think what we've seen so far, is a pretty fair assessment of what we will see in the remaining games
Wait, what is a ball control O? Because with Denard leading the nation in interceptions, I'm not sure how good of job we've done at keeping control of the ball. We throw a lot of high-risk, high reward passes, go for it on fourth down frequently, and really couldn't sustain any long drives against Notre Dame and only one long drive against MSU. I'm not sure in what respect our offense is about ball control. That's not necessarily a criticism of our offense (it would be sweet if all of our drives were one play that ended in a touchdown), but I'm not quite sure what you're getting at by talking about a ball control offense.
I think M is capable of putting together long clock eating drives because of Denard's running ability. It is not often you stumble across a QB who is a threat to rush for 200 yards on a given day. Plus M takes their time with the play clock further diminishing the time the opponent has the ball. IDK what M's TOP looks like, but the O has the capability, thru long methodical drives, of complimenting the D nicely.
Of course we'll have to wait for the whole season to pan out. Worsening on offense and improving on defense were almost inevitable given our finishes last year. But such a jump on D is incredible.
And with only one case we can offer tons of explanations, but I'll buy Michfan1997's (where's the fans of 2002 and 2007??). GERG was very likely an exception to the rule -- an exceptionally bad coordinator in an exceptionally bad situation that helped to drag down team performance. And GMAT may be the opposite -- an exceptionally good coordinator in an exceptionally good situation.
But I'll wait to the end of the season to count any chickens.
Took a look at more FEI stuff and found the offenses we face in the future are currently rated worse than the offenses we have faced in the past. That should be good news.
Defensive SOS to date for M: = #33
Defensive SOS for remaining games: = #89
That is based on adjusted data. Here are the current FEI offense rankings
ND(6),MSU(37,NW(42),WMU(48),SDSU(94),EMU(97),Minn(104)which is an average ranking of 61
NEB(20),Iowa(53),Ill(73),osu(76),Purdue(99) which is an average ranking of 64
piggy back off of your comment and add a few more things. Last years defense allowed 6.1 points per game. That number has been reduced to 5.4 this season. A lot of that is due to Notre Dame (the #6 offense) putting up a 7.1. Outside of that, Northwestern put up a 6.3. They, as we know, run an option offense that was able to exploit our weakness on the edges. The only other decent offense we've faced were MSU, who put up 5.3 and Western who put up 5.0. Everyone else is below that.
Looking to the rest of the year, Nebraska seems to be the only offense that is terrifying. Sure anything can happen, but I feel we do have the ability to matchup with all 5 remaining opponents.
That is a good sign. Notre Dame is the only team that really carved us up and they are a top 5-ish FEI. Northwestern had a good half against us but otherwise no one has done all that much against our defense. So Nebraska looks like a good test but the other offenses don't look too difficult. We're in really good shape going forward and I truly believe that last weekend was our most difficult test of the regular season.
I looked at the FEI ranking for offense and it appears that an explanation for the low ranking of some teams is the relative strength of the defenses they have faced FEI's OSOS(A).
has to be that they managed to score 30 points a game despite a horrible Field Position Advantage ranking.
Raw statistics for football are relatively meaningless -- especially before teams have played almost their entire schedule. The strength of opponents is so different that without adjustments, the data is almost always misleading.
In fact, if FEI, S&P and other adjusted data resulted in similar rankings for offense, defense, overall, etc. there would be no reason to adjust the data in the first place.
So, it is a bit ironic that a complaint of adjusted college statistics is that they are not the same as the unadjusted data.