Why the collapse? A tougher second half of the season?

Submitted by michelin on
There’s been a debate on this board whether UM had a tougher schedule during the second part of the season. No, based on the Sagarin PREDICTOR ratings for B10 plus ND games only. Our strength of schedule (SOS) has NOT been tougher in the second part of the season, even when the games are corrected for 3-point home-road advantage-disadvantate.* Our schedule during the second half of the season was about 2 points easier on average. However. If we include ALL of the games, the second half of the season WAS tougher overall. They were only mildly tougher. However, statisticians often seek to improve the reliability of the data by throwing out the highest and lowest ratings (here, IA and DSU). If you do that, there is a significant trend toward tougher games throughout the season (r-.47) —making each successive game almost 2 points tougher than the next, on average. See link http://cid-4bf9d75c782b05b1.skydrive.live.com/self.aspx/UM%20opponent%2… When we add OSU (87.1) next week, our second-half season ratings will seem even tougher. , *Summary (for comparison purposes, note that UM’s current ratings is 69.0). IA 80.8 PSU 83.2 Actually 83.8>80.2 with Home-Road (+3,-3) ND 79.8 > WI 77.7 actually 76.9 80.7 with H-R (-3,+3) MSU 76.6> Purdue 69.0 actually even more so 79.6>66.0 (+3,-3) Ind 64.9 > IL 63.9 actually 61.966.9 (-3,+3) Total ratings for 1st vs 2nd half of B10 and ND games only: 302.1>293.8 actually with the same corrected ratings (H-R balances out) NonB10 WMU 59.3 actually 56.3 EMU 56.5 actually 53.5 DSU 38.6 actually 35.6 **** Based on the current rating, UM should have lost to ND, beaten Purdue and Il. We beat ND , lost to Purdue and Il. The other five major games (Ind, IA, PSU, WI,MSU) went as expected, when you make adjustments for H-A. Note that to really make a fair test, however, we should recomputed the sagarin rating for all the games except the one we are considering (ie to determine whether we should have won or lost it). I don’t have the software to make those corrections, however.
MGoBoard
Michigan football
collapse season
sagarin ratings
strength of schedule

oakapple

November 16th, 2009 at 4:00 PM ^

I don't get the sense that these stats are statistically significant. For me, the most significant reasons for Michigan's drop-off are: 1) Losing David Molk 2) Games in which Brown or Minor (or both) were unavailable or playing hurt 3) Games in which Forcier was playing hurt 4) Teams started to figure out Michigan's weaknesses

Joined: 02/06/2009

MGoPoints: 2130

CriticalFan

November 16th, 2009 at 4:35 PM ^

I guess things like Penn State > ND, Wisconsin > Michigan St. and Iowa > Indiana aren't worth considering? Teams are ranked for some reasons, reasons like ability to kick the hell out of conference cellar-dwellers like Michigan (which is where we are). Also, all your bullets are happening to other teams too. ND losing Floyd and Clausen, Iowa losing Stanzi, etc. (4) is laughable. Is it impossible for Michigan to figure out other teams weaknesses? We have as much film on them as they do on us.

Joined: 10/14/2009

MGoPoints: 1807

oakapple

November 16th, 2009 at 5:37 PM ^

Iowa lost Stanzi, and sure enough they started losing. ND was a mirage team that finally ran out of miracles. As far as (4) goes, isn't it rather apparent that opponents started hammering Michigan in the middle of the field, where it is weakest? Once it became clear that Michigan couldn't defend the post pattern to the tight end, then every opponent started throwing it. Of course, if Michigan had had an answer for that, we wouldn't be having this conversation.

Joined: 02/06/2009

MGoPoints: 2130

michelin

November 16th, 2009 at 8:48 PM ^

I agree about many of your reasons. But I am only looking at the narrow question of the strength of competition. You are correct that the data would not reach usual scientific standards for significance based on these data points alone (ie p less than .05). They only justify about 68% confidence that the rating changes are not random. Of course the sagarin ratings are based on voluminous data, so it is hard to regard their values as fluctuating very widely. I am not sure that a conventional statistical analysis really does justice to these data.

Joined: 09/22/2009

MGoPoints: 2541