I don't have the spreadsheet on me but I can easily put that together in the morning.  However 1 thing to note is that the MSU game wasn't that much worse than the Notre Dame game - we just had a ton more possessions to get those points.  I did take a quick look at what it would have been like had I left out the MSU game entirely and it only accounted for a few points - most of that is due to the fact that IA's defense is only giving up .5 pt/poss this year and still less than 1 last year so even a .5 pts/poss increase to the whole is only going to see a 6 point swing.

Either way, I'll plug the numbers in in the morning and post an update w/ those.  I probably won't include the missed FG at the EOH because, as mentioned, we probably won't usually make that anyway.

As promised - I put in 3 TD's instead of 2 INT's and 1 FG and it really doesn't end up doing too much to the predicted score since it's just 1/6th of the input.  The projected offensive points go from 20 and 22 to 22 and 24 respectively and bump the total average to UM 20 and IA 32.  So, still basically the same thing.  If there's anything to take from this is that if we're having an off day having the other team moving the ball quickly helps since ND was our second worst performance to date but we had many, many more posessions in that game than any other game.

For example Iowa has played virtually the same slate as last year (2 cupcakes and exactly the same 3 big opponents; Iowa State, Penn State, Arizona) but those teams aren't the same as last year and neither is Iowa (or maybe they are?) so there is in fact a common dataset of the first x games of last year vs the results of this year. Using this metric rather than the average of the entire season last year removes the bias of playing an entire Big Ten schedule. 

What this prediction does is compare where we were last year with where we are this year and compare that to where we were last year and how we did against the same opponent.  So if we outperformed our average O against IA last year I'll use that % and apply it to what we're looking to do this year.  If you go back and compare the first few weeks it's not really going to tell you anything because the averages and numbers are moving.  This just simply answers the question of "If we perform at the same efficiency vs our average as we did last year this will be the result" and I've combined that from both the Iowa perspective and our perspective. 

You could simply just say IA is giving up 1 PPP and UM is scoring 2 PPP so it'll be 1.5 PPP but that doesn't take into account how Michigan performs, relative to itself, against Iowa.  What it does take into account is that idea that you perform better against some schools (compared to moving average) than you do against others (i.e. OSU always seems to have our number and that'll show up in the %'s when we play them).  It is by no means a sure thing or anything as in depth as what Mathlete does but it does still have both a common data set and a scientific, neutral basis.