It's Memorial Day weekend. As per custom, I hope you're all grilling out and enjoying some All-American burgers, done to varying degrees of quality. Try not to let the weather deter you.
A quick PSA: you may want to add some healthier options to your plate than meat and orange-ish American cheese. Potato salad, corn-on-the-cob, bell peppers, etc. All of these are perfectly valid options. However, asparagus is gross. On this, there is no debate.
Michigan - Tot Rat Trending
Michigan - Win Pct Trending
1930-2014 Record: 632-257-20
2005-2014 Record: 73-53
Michigan - Historical Expectations
50 Yr Avg Record: 9-3, Avg Rank: #17 (out of 125)
20 Yr Avg Record: 8-4, Avg Rank: #23 (out of 125)
10 Yr Avg Record: 7-5, Avg Rank: #34 (out of 125)
Currently, I've got Utah over Michigan by roughly 6 pts (65% W-Lk). Let's assume that Michigan ends up winning by 3 pts. What are the reasons why I could have been wrong?
Crap Happens - This is the easy cop-out, but a model that picks Utah to win 65% of the time inherently assumes that Michigan can and will win the other 35%. Even going by MOV, a 9 pt difference is not all that much. Vegas still can't get within 12 pts on average. A Michigan win in this interpretation is not so much an indictment as a "yes, this is something that could happen and did this time."
Missing Info - injury, weather, personal circumstances, etc. These are all thing that could impact a game, but my model just doesn't take into consideration. Some of these are just difficult to find reliable information on (injury info, for example), some of these just aren't available at all unless you're a psychic (personal circumstances, obvs). For the most part, this is "live with it" territory.
Neglected Info - information that is readily available, but not adequately taken into account. Refusing to apply home/road effects to game predictions would be a good example (something I used to omit, but no longer). This type of error would vary based on interpretation.
Bad Model - I could have all the data I need, but it doesn't matter if I feed it to a junk formula. This is probably where my rating system falls short. Maybe it's the "SEC=+10000" line of code, but probably not. This is the most likely option, obviously.
|Conf||Record||Tot Rat||rk||Adj Rat||rk||Res Rat||rk|
|Gms||Prj Rec||Adj Off||rk||Adj Def||rk||Prj SOS||rk|
Michigan - Proj Ratings & B10 Average
Ret Off: 9 (0 QB), Ret Def: 7
|Gm||Wk||Day||Date||H/V||Opponent||Opp Rk||FCS?||Conf?||Prj W-Lk|
Harbaugh? Harbaugh harbaugh. Harbaugh... harbaugh. And Jim.
The previous head coaching stops on Jim Harbaugh's career resume, and how he fared in his first year:
University of San Diego
2003 Record, Sagarin Ranking
2004 Record, Sagarin Ranking
2006 Record, Sagarin Ranking
2007 Record, Sagarin Ranking
San Francisco 49ers
2010 Record, DVOA Ranking
2011 Record, DVOA Ranking
Standard caveats for drawing conclusions from small sample sizes. And presumably Michigan didn't hire their newest coach solely for his first year performance.