Updated Massey win probabilities for Michigan and Nebraska

Submitted by Yeoman on October 29th, 2012 at 11:55 AM

I posted this on one of the weekend threads but the Massey site has now been updated with Saturday's results so here's a new version. Massey win probabilities for the remainder of the season:

Michigan:

  • Minnesota 78%
  • Northwestern 68%
  • Iowa 81%
  • Ohio State 37%

Nebraska:

  • Michigan State 51%
  • Penn State 72%
  • Minnesota 89%
  • Iowa 71%

That works out to:

Michigan:

  • 4-0 16%
  • 3-1 43%
  • 2-2 31%
  • 1-3 9%
  • 0-4 1%

Nebraska

  • 4-0 23%
  • 3-1 44%
  • 2-2 27%
  • 1-3 6%
  • 0-4 0%

Chances of winning the division are now 71% Nebraska, 29% Michigan, ignoring the  unlikely possiblity of someone else stepping in (I'm not going to run the numbers but eyeballing it I'd say Northwestern's chances are about 2-3%).

Comments

HermosaBlue

October 29th, 2012 at 12:31 PM ^

Not my normal line of commentary, but the negging here makes no sense.  Denard is wonderful.  Our overreliance on him, not so much.  Doesn't deny that Denard is amazing (he is), and Saturday night in Lincoln kinda capsulizes why people who say this sort of thing say this sort of thing.

Painful truth.

Bodogblog

October 29th, 2012 at 12:35 PM ^

Most teams are reliant on their best player.  Nebraska is reliant on Martinez.  Without Bell, Sparty loses to Boise and probably EMU.  Miller for Ohio.  Almost every team is going to be signficantly worse without their best player.

It's being negged because overly relying on Denard isn't the problem.  The problem is a lack of depth at QB, and the lack of depth at WR that necessitated the Gardner move.

HermosaBlue

October 29th, 2012 at 7:09 PM ^

I don't think anyone's suggesting that we have better options.  One can be unhappy about overreliance on Denard without being delusional that there's a better alternative, all of which is born of the fact that we have incurred substantial transition costs.

It's more of the shaking angry fist at cruel twist of fate, rather than shaking angry fist at Borges et al.  

Perhaps I'm overly forgiving, but I get where the poster's coming from, even if it comes off as a bit of a whine given that there's no alternative.

ChalmersE

October 29th, 2012 at 12:11 PM ^

Given how (in)effective Gardner has been at wide receiver, the coaches have to take some blame for moving him out of the back-up qb position. One virtue of Gardner as the back-up is that the offensive scheme stays similar. Bellomy is a more traditional qb and making that type of shift mid quarter is not going to have a happy ending very often.

JamieH

October 29th, 2012 at 2:09 PM ^

Anytime you lose a senior QB your season is in danger of being derailed.  That's kinda just how it is.  Sure, it would be great to always have an awesome backup QB, but you can't count on that every season, not these days.  Awesome backup QBs now go to mid-major schools and start for them instead of sitting around on your bench waiting for an injury.

turd ferguson

October 29th, 2012 at 12:01 PM ^

Interesting. Thanks for posting. These seem much more plausible to me than the probabilities we've been getting from The Mathlete's model over the last few weeks.

I assume these are blind to Denard's injury (except insofar that the effects of his injury were baked into the Nebraska score)?

Yeoman

October 29th, 2012 at 12:11 PM ^

That's correct--Massey's model is completely blind to anything except the final scores.

When I ran these before the Nebraska score (and thus any effect of Denard's injury) was baked into the probabilities, the odds came out to 50/50. If he's healthy the rest of the way that seems like a reasonable estimate.

mGrowOld

October 29th, 2012 at 12:18 PM ^

Feel like last year much?  Road game against division rival for inside track to B1G Championship game and offense lays egg.  Lose tie-breaker because of head to head loss.

Sure hope the outcome is different at the end of but just like last year we didnt help ourselves any in achieving the goal of winning the B1G.

Yeoman

October 29th, 2012 at 2:59 PM ^

That's a good assumption, if Michigan wins out (as you proposed in your post). It's also a good assumption if Northwestern wins out, except that then the two horses are Northwestern and Nebraska.

Northwestern's 7-2 and one of the losses was by one point to Nebraska in a game they led by two scores late. If they hold onto that lead, or make the field goal that was barely wide at the end, they're a game ahead in the standings holding the tiebreaker on all the teams one behind them.

Massey doesn't think there's much chance they'll win two games in Michigan, and that's probably right. But they're far from out of it.

LSAClassOf2000

October 29th, 2012 at 3:05 PM ^

I agreed with Massey, which had Northwestern with 32% and 37% for Michigan and MSU respectively, but the simulated season results were interesting:

0-3 came to 2.4%, 1-2 came to 43.1%, 2-1 was 42.4% and 3-0 was 12.1%. In essence, there is one game in there that may be a good each way bet. They'll easily beat Illinois, or what is left of Illinois, so I would think that at least 1 win is a given for them.

justingoblue

October 29th, 2012 at 2:01 PM ^

and I partially said that for effect. I'd bet a lot that either Michigan or Nebraska goes to Indianapolis. Maybe typing that as a reply to a question actually seeking information was the wrong way to go, but I don't think the chances of Northwestern getting there are high at all, and Iowa would require a miracle.

Sideline

October 29th, 2012 at 12:45 PM ^

Depth is our biggest problem... Especially at a position like QB... I really want the coaching staff to re-evaluate the Gardner situation, he's not as threatening as most people assume, and if Denard is going to miss significant, meaningful game time, like what we experienced in Lincoln, I'd have to hope that Gardner gets some looks... I like Bellomy, but I don't think he has enough supporting cast to come into the offense yet... At least not when we NEED to move the ball. I feel that Gardner could move it better than Bellomy with Gallon and Dileo proving to be more than serviceable and Roundtree had some breakout moments in Lincoln...

Hopefully Denard will be okay in the long-term though! Hate to NOT see him play...

michelin

October 29th, 2012 at 4:19 PM ^

The final calculations of total record probabilitiles (eg 4-0, 3-1 etc)  appear to assume that the game outcomes are all independent of one another.  That assumption is likely to be false.  For example, IF denard is healthy, then all of the win probabilities will increase for UM.  IF he is not, then they will all decrease.  The net effect of such dependency (on the health of denard) will be to increase the effect of extreme outcomes (eg going 4-0, which we are likely to need to win the Legends division).  There are likely to be other types of dependency as well, which may afffect both UM and NEB.

Yeoman

October 29th, 2012 at 4:58 PM ^

Yes, I suppose I need to say that every time I post one of these but I tend to assume it's obvious.

You can build in some autocorrelation if you like--if it's reasonable and applied equally to all teams it probably wouldn't be material (these are obviously very rough estimates anyway) to the main N vs. M question except for the particular special case of Denard. A similar consideration for Nebraska, say the health of Martinez, would increase their chances of going 4-0 (which would lock Michigan out entirely) or going 1-3 or 0-4 (which would open the door for everyone).

In general, the net effect of increasing the likelihood of extreme outcomes is a small increase in the chances for the team behind (Michigan) and a significant increase in the chances for the outliers (Northwestern, Iowa).

 

michelin

October 29th, 2012 at 8:53 PM ^

To give a purely hypothetical and oversimplified example:

Suppose that having denard increases doubles the odds of UM winning--with a corresponding decrease by a factor of two if denard is out.*

Suppose that Neb has a much better backup than Bellomy, so that having martinez (or not) only increases (or decreases) the odds of Neb winning each game by 25%.

Then, even if Martinez had the same prior chance of being out of all four games as Denard (say 10%), the chance of Neb going 4-0 increases only modestly (from 23% to 28%).  But the chance of UM going 4-0 nearly doubles (from 16 to 31%).  That is, UM now has a greater chance of going undefeated than NEB; and if UM and not NEB went undefeated, then  UM would win the Legends division. 

I don't have time to calculate or sum all the other probabilities, also one can question my assumptions, and I am sure one can create other scenarios that yield different results.  However, I only seek to make one point:

Dependencies on the presence of the starting QB can be large and counterintuitive.

 

 

*note that I am taling about doubling the Odds, not the probabilites.

Yeoman

October 29th, 2012 at 9:15 PM ^

The original post in the weekend thread was done before the results of the Nebraska game were incorporated into the Massey probabilities and thus were determined on the basis of a healthy Denard. As I've mentioned a couple of times above, the un-auto-correlated probabilities were evenly split 50/50 between Michigan and Nebraska at that point. So, yes, a QB injury obviously has a big impact, once you know an injury has happened. We went from 50/50 to 30/70, and that's before we know whether Denard's injury will linger. (To be precise, we went from 50/50 with a healthy Denard to 30/70 with a 1/8 chance that Denard will miss half of any particular game. And with a healthy or at least playable-if-slightly-dinged Martinez in both cases.)

But the odds of a season-changing multigame QB injury are a lot less than 100%, before it happens. With what seems like a reasonable estimate of the odds of a major incident like that--maybe 5% over a 4-game period? or 10%?--and applying it to all teams, you don't get a big shift in the two main probabilities. When you start trying to differentiate the reliance on key individuals among the teams you get into an area that probably neither of us has enough information to get anywhere with. What do you really know about the depth of Nebraska's squad or the quality of Martinez's backup? And they don't exactly have a defense that gives them a chance if their offense isn't moving the ball.

Of course you can probably (or maybe not, but that's another post I'm saving for the off-season) do better than these simple models if you want to pour a lot of time and analysis into it and make a lot of ad-hoc assumptions.

My assumption has the merit of being really simple, which makes it easier to see how it applies and where it could go wrong and also means I have time to actually set up the calculation in 20 minutes instead of wasting all my spare time on it.

rob f

October 30th, 2012 at 12:31 PM ^

that might be factored into next week's calculations of the odds, is the health of not only Denard, but of Nebraska's Martinez.  Keeping in mind that UNL plays State this coming weekend, an already banged-up Martinez is probably at a somewhat higher risk of further injury Saturday than he will be in other remaining games, given the "style" of play of State's defense.  Wisconsin's loss to MSU this past Saturday probably wouldn't have happened had Stave played the entire game; State's defense has a history of damage to opposing QB's.

MaizeRage89

October 29th, 2012 at 10:11 PM ^

That we lose 1 fucking division game and we are potentially out of the running. These divisions suck and shouldn't come down to just one game winner takes all. Same bullshit happened last year too