Writers of Pro Football Prospectus 2008

30 Oct 2016

ESPN: NCAA Five Undefeated Teams Left

Seven undefeated teams went on the road this weekend and four failed to come away with a win, narrowing the list of the teams most likely to run the table to five. Who has the best chance to finish the regular season undefeated? It's not as cut-and-dried as you might think.

Posted by: Brian Fremeau on 30 Oct 2016

8 comments, Last at 03 Nov 2016, 1:59pm by ekshi


by pm :: Mon, 10/31/2016 - 10:43am

"Michigan Wolverines
58 percent likelihood to win out (up from 57.8 percent last week)

our model still gives Ohio State an advantage in the matchup with Michigan because the game will be played in Columbus."

How do both sentences reconcile? Michigan has a better than 50/50 shot at being undefeated but they are underdog against Ohio State? That makes no sense. Their odds of going undefeated should be under 50% if they are underdogs to Ohio State.

by Chappy :: Mon, 10/31/2016 - 4:13pm

I can't view the article, but it's got to be wrong. 58% implies they are 87.3% likely to win each of their next for games (even higher if the Big Ten championship is included). I'm a Michigan fan, but there is no way they should be that level of favorite--except for maybe the Maryland or Indiana games which are at home. According to the website below, this implies they'd be roughly an 18 point favorite for each of their next four (or five) games. Plus, you're correct that, even if they are 100% favorites for all but Ohio St. game, they'd still, by definition, need to be 50% chance of going undefeated.


by ChrisS :: Mon, 10/31/2016 - 4:33pm

Perhaps the following: since OSU already has one loss they are less likely to be good as we "think" at this time so in a significant number of projections they lose more games and then are not projected to beat U of M. Whereas U of M currently with no losses is projected in the vast majority of scenarios to be undefeated going into Columbus.

by Eddo :: Mon, 10/31/2016 - 4:45pm

Yes, but even so, the two sentences contradict each other.

Even if Michigan were projected at 100% likely to win every game not against Ohio State, that means they would be 58% likely to beat Ohio State.

But the second sentence says that the model gives Ohio State the advantage in their upcoming matchup.

So either (a) the first sentence's "58%" figure is wrong, (b) the second sentence is wrong and Michigan would actually have the advantage over Ohio State, or (c) the second sentence is using an confusing definition of "advantage" that doesn't mean "likely to win".

by ChrisS :: Mon, 10/31/2016 - 10:06pm

In the projection new information accrues and the probabilty of U of M beating OSU must change. Whereas right now OSU would be favored based on current info. Assume there are 4 states in the projection UofM dominant/weak, OSU dominant/weak, perhaps UofM is dominant in 80% of the scenarios and beats weak OSU 75% of the time and losses to dominant OSU 55% of the time whereas OSU is dominant only 20% of the time with slightly better win odds.Then both statements can be true.

by Chappy :: Tue, 11/01/2016 - 11:48am

I don't see how this could possibly be the case. The projection model is almost certainly just running off their current FEI projection, and is Bayesian, so I really doubt what you're saying is possible. Even if I assume you are right, I don't see how that translates to saying OSU is favored. If you take the expected value of either state, Ohio State wins 39% of scenarios while Michigan wins 61%. I don't see how the author could possibly consider this to be projection where Ohio State is favored.

by ChrisS :: Mon, 10/31/2016 - 4:33pm


by ekshi :: Thu, 11/03/2016 - 1:59pm

nice on this and may be it would be 58% likely to beat Ohio State. http://www.gudtechtricks.com/2014/09/android-4.4.4-kitkat-updates-free-d... I don't see how the author could possibly consider this to be projection where Ohio State is favored.