Mid-Week Metrics: Projecting Michigan

Submitted by The Mathlete on September 1st, 2011 at 4:58 PM

Now that Brian has burned through the position previews and depth charts in beautiful excruciating detail, there is little for me to add to the personnel side, so I wanted to look a little deeper at the schedule portion to see how we got to the 8-4 I projected last week.

I also wanted to add a bit of addendum to the Denard struggles on passing downs meme, so to not clutter the site any further I have dropped that at the end of this column for those interested.

Throughout the season, I will be posting a weekly column on Wed/Thurs. I will try and pick out interesting tidbits and trends from the numbers as the week goes. If you have any questions you would like to see answered in the column or ideas on angles, don’t hesitate to hit me up on the twitters.

As always, your handy reference guide is here.

So which are the 8 wins?

Well, it doesn’t really work that way. Obviously no game is certain and no prediction is either. To get to 8-4 I assign values to each team based on the prior three seasons' performance and returning starters at QB and defense. These are factors that I have found significantly improve a season’s forecast.

Each team is then pitted against their schedule, accounting for home field which is worth about 3 points for the home team each game. Each game then gets a spread and a likelihood of winning. When you play out those probabilities, some seasons ended up with as few as 1 win and some ended up with 12. Nearly three quarters ended up with seven, eight or nine wins. My calculated odds of missing out on a bowl are about 1 in 29, about the same odds of winning 11. Going 12-0 is rated at 1 in 327. This is all assuming that Michigan plays at the projected level. If they play better or worse than I have projected, the numbers can and will change.

All that was to say, the eight wins and the four losses change each scenario. The most likely version has losses to Ohio, Nebraska, Michigan St and Notre Dame, but even that scenario is only a 1 in 60 shot. In fact the most likely specific scenario is 6-6 with losses to Illinois and Iowa added to the mix, but that’s still a 1 in 55 shot.

In summary, here is how the percentages break out:


The Individual Teams

Opponent 2010 PAN 2008-'09 Avg Returning Starters Total PAN Michigan Odds
W Michigan -2.8 -3.2 1.8 -1.2 89%
Notre Dame 6.5 1.2 1.8 5.7 48%
E Michigan -10.4 -9.5 1.0 -9.0 100%
San Diego St 6.3 -6.0 -0.6 -0.5 85%
Minnesota -4.9 -0.6 -0.2 -3.0 100%
@Northwestern -6.4 -2.3 1.0 -3.4 84%
@Michigan St 6.8 0.5 0.2 3.9 41%
Purdue -2.5 0.5 0.6 -0.4 85%
@Iowa 8.1 3.7 -3.4 2.5 49%
@Illinois 4.3 0.4 0.2 2.5 49%
Nebraska 10.6 3.6 1.0 8.1 34%
Ohio St 11.7 10.7 -3.4 7.8 35%

The numbers quickly break out into four groups:


Eastern Michigan and Minnesota coming into the Big House without much hope. Eastern was bad every year considered and only gets a slight uptick from returning starters. No points awarded for hiring Mike Hart.

Minnesota saw last year plummet below already-low-for-a-Big-Ten-team values and returning starters push them down slightly further.

Just Don’t Screw It Up

Western Michigan, San Diego St, Purdue, and at Northwestern all seem pretty safe on their own, but there is only a 55% chance we go 4-0 in these four games. Successfully do that and a nine-win season becomes a more attainable. Dropping one or more will make it tougher to top last season’s win total in the regular season.


Notre Dame, at Iowa and at Illinois all place Michigan a percent or two below 50/50. 5-2 between these last two groups keeps us on pace to 8 wins. Iowa overachieved last year but is brought down to earth thanks to a depleted roster. Illinois is heading in the opposite direction after [NAME REDACTED] made one last run to save his job. Notre Dame is the highest rated of the bunch as Brian Kelly begins to purge the Weis ratings from the books. The Domers get the benefit of a strong returning group but are in the mix with Iowa and Illinois thanks to an under the lights meet-up in Ann Arbor.

There’s a Clock for That

OK, so we don’t have a countdown clock for that school down south and four states over (Nebraska), but Ohio and State form the last group. To hold serve on an 8-win season, expect one win out of this group. Ohio has been the cream of the Big Ten for the last several years, but graduation and Tressel-gate have dropped the Buckeyes into the mix. Michigan State and Nebraska both saw 6+ point improvements last season and have a decent group returning. Nebraska should definitely be the better team, but they won’t have the luxury of home field.

PS: Denard and Passing Downs

In general, my data supports what Burgeoning Wolverine Star found on Denard and passing downs. I was curious about which down and distances that Denard excelled and what was their value. For the season, Denard was a non-opponent-adjusted +70 for the season. This includes rushes, passes, sacks, fumbles, picks, everything but garbage time. This is a huge number.

I broke down where the +70 came from situationally.

Down & Distance PAN
2nd & Long (8-10) 21.5
1st & 10 21.2
2nd & Med (4-7) 18.7
3rd & Short (1-3) 12.0
2nd & Short (1-3) 6.9
3rd & Med (4-7) 1.5
2nd & XL (11+) (0.4)
3rd & Long (8-10) (4.9)
3rd & XL (11+) (6.6)

Denard was light years ahead on 1st and 2nd down but considerably below average on 3rd down with at least 8 yards to go. In fact, he was pretty good at 3rd and short and started quickly falling from there.

Ultimately, as long as the offense didn’t lose ground on first down they were still in good shape. Denard could turn a mediocre 1st down around quickly, but if Michigan wasn’t able to get into a third down distance that was manageable, the offense quickly become below average.



September 1st, 2011 at 5:16 PM ^

You obviously have a formula to back yourself up which I appreciate, HOWEVA,

84% at Northwestern seems wildly optimistic to me. You consider a win at an away game at NW to be:

1. Only 5% less likely than a home game against Western

2. Only 1% percent less likely than a home game against SDSU with a new coach that lost its two top receivers to injury

3. Only 1% percent less likely than a home game over yet another awful awful Purdue team with a coach about 7 tiers down from Fitzgerald.

In summary -- does not compute. I do appreciate the effort  you put in to these posts though, I just think something doesnt add up here.


September 1st, 2011 at 5:34 PM ^

2 questions:

1) How did our breakdown of "likelihood of X wins" look last year?

2) Have you given evidence before that your method is sound?  I'm not bagging, just curious, i.e. if we used your prediction now for the 2010 season, how close was your prediction on average to the actual outcome.


September 1st, 2011 at 5:55 PM ^

Do you have time to calculate situational PAN for last year's B1G and also top 20 overall QBs? It would be nice to compare Denard to Forcier, Luck, Pryor, Persa, Moore, etc.


September 1st, 2011 at 5:56 PM ^

8-4 with losses to ND, Iowa, Neb and Ohio.  I feel good about a win over Sparty.  If they manage a win over Ohio the Hoke karma train would be unstoppable. 


And tremendous.


September 2nd, 2011 at 10:52 AM ^

will get some good momentum if we witness a decisive win over the Domers.  And by decisive, I mean no controversy that gives the ND fan base an excuse to say, "Well, we WOULD have won if XXXXX hadn't happened..."

XXXXX being:  the damn refs had said that guy was in bounds, which video evidence contradicts, or the Red Rider BB gun hadn't shot our QBs eye out, or the dog hadn't peed on my homework, or if our coach wasn't so inept...


September 1st, 2011 at 9:27 PM ^

"if Michigan wasn’t able to get into a third down distance that was manageable, the offense quickly become below average."

I fully expect that this will change this season, convincingly and demonstrably. Denard is hitting 70% of his practice passes, we have veteran offense returns, and a renewed work ethic, vigor, and ever increasing skills at key positions. 


September 2nd, 2011 at 10:05 AM ^

"if Michigan wasn’t able to get into a third down distance that was manageable, the offense quickly become below average."

I fully expect that this will change this season, convincingly and demonstrably.

I wouldn't expect it to change very much; in fact, I wouldn't be surprised to see to see it go down. I mean, the same thing could be said for any offense in the country, it's just harder to convert third downs that aren't within a manageable distance. And even though those factors you mention are true, I don't think Denard's completion percentage in practice means very much, and our teams for the past three years haven't been bad due to lack of work ethic or vigor. I'm not saying our offense will suck, but with the transition into a different offensive mentality, new plays, new techniques required, etc., it's not reasonable to expect that our offense is going to improve over last year's.


September 1st, 2011 at 10:04 PM ^

If the defense improves through the season and the offense can score, I could see Michigan knocking off either NE or OSU at home!  I'm not sure what our final record will be...but my best guess is between 6 and 10 wins.  We may knock off IA on the road but may lose a close one at IL?  Also building momentum this season may go a long way, if we can knock off ND at home and go in to East Lansing undefeated, we may take this one from the Spartans. Our team is more experience, so we should perform much better in B1G play. I just have a hunch that Michigan will win more than 8 games this season. But even 6 wins will be fine with me since Hoke is a first year coach.


September 2nd, 2011 at 11:57 AM ^

Can I ask if there is an average PAN breakdown situationally? That is, I imagine that most quarterbacks are better on 2nd and short than they are on 3rd and long. So what would be most interesting to see where Denard excels relative to other QBs and where he is behind the curve.