Sunday, November 27, 2016

Michigan Is One Of The Four Best Teams, But They Cannot Be Allowed In The Playoffs. Here Is Why:

Ohio State and Michigan came into yesterday's game with a single loss apiece. Besides the normal interest sports fans have in a Michigan/Ohio State game, it was assumed that a playoff spot was on the line. And it showed up in the tv ratings:
The fact is that, in a loss, Michigan impressed. The game effectively was a draw, on the road at Ohio State. In overtime, the game literally came down to a 50/50 call, the 4th down scramble by JT Barrett that could have gone either way. It was so close that whatever call was made on the field necessarily had to stand on review. Had officials ruled Barrett an inch short, Michigan would have won. But they did not, and Ohio State won on the next play. It's literally as close as a game can get.

And the computer ratings recognize that. Michigan's ratings in metrics like the S&P and Sagarin all rose slightly after yesterday's games. And naturally, this is leading commentators to declare that Michigan should not be punished for a loss, and that as one of the four best teams in the nation they must be allowed into the playoffs. Here's a smattering:
Is Michigan one of the four best teams in the country? Probably.

The most accurate ratings for college football (and the ratings that most closely match the Vegas lines) are the S&P and Sagarin, and they both have Michigan safely in the Top 4. While there might be some rating you can find somewhere with Michigan outside the Top 4, it's safe to say that there's a general consensus in the computer models that Michigan is one of the four best teams:

But even though Michigan is one of the four "best" teams, they cannot be allowed in the playoffs.

Why? Because at the end of a season, teams must be rewarded for wins and losses, not for how good they are.

First of all, we all know who the best team in college football is. It's Alabama, and it's not that close. But if we wanted to give the title to the best team, we'd just halt the season right now and hand the trophy to Alabama. Alabama was the best team in the country two seasons ago, and they got knocked off in the playoffs by Ohio State. On any given day, anybody can beat anybody, and the whole point of having a playoffs is to add excitement by increasing the odds that an unlikely team wins the national title.

More importantly, if we rank teams by how good they are rather than by wins and losses, games would stop mattering. Why on Earth would anybody but Big Ten fans give a shit whether Michigan or Ohio State won that overtime? No matter whether that JT Barrett 4th down run was ruled a first down or not, both Michigan and Ohio State were going to end up in the Top 3 of the Sagarin and S&P ratings. So why would any non-Big Ten fans care about whether it was converted? Michigan fans complain that they had bad luck - and they're right! - but that's the point of playing the games. There is random luck, and it leads to unpredictability, and it forces us to watch and care.

To use a college basketball example: Imagine any game that comes down to a buzzer beating shot. Don't you want to care if that shot goes in? Because the Pomeroy ratings don't - not by any measurable amount. But it matters for their resume, and as long as the Selection Committee judges teams by their resume and not by how good they are, we are all going to care about whether that shot goes on in or not.

"Wait a second!" cries the Michigan fan reading this piece. "We have three wins over Top 10 teams! Our resume is still worthy of Top 4!"

While this argument isn't necessarily false, it's a poor one. Despite how common we hear television commentators (and Selection Committee Chair Kirby Hocutt) talk about metrics like "Top 10 wins" or "Record vs the Top 25", these are lousy metrics for four big reasons:

1) They are wholly arbitrary. Why care about Top 10 wins when we could care about Top 15? Top 25? Top 35? Fans are always going to set the line somewhere convenient for their team.
2) They necessarily only look at a fraction of a team's resume. Why only look at three games when we can look at all twelve?
3) The correlation between where teams are ranked in the Playoff Rankings and how good they are is a bit tenuous.
4) They do not take home/road into account.

As far as Michigan goes this season, #4 is a major factor. All three of their "Top Ten wins" came at home. Take their win over Penn State, for example, According to the Sagarin ratings, a home game against Penn State (AP #8) is equivalent to a road game against North Carolina (AP not receiving votes). Somehow that "Top 10 win" seems a wee bit less impressive when it sounds like a win over 8-4 North Carolina.

Say what you will about ESPN's FPI, but it's a decent rating system, and ESPN is nice enough to put out a "strength of record" metric. It asks a simple question: What are the odds that an average Top 25 team would equal or better your record against your schedule? Here is where that stands at the moment:

In other words, an average Top 25 team would have a 4% chance of going 11-1 or 12-0 against Ohio State's schedule, but a 96% chance of going 10-2 or worse. Clearly, if you look at resumes, Alabama, Ohio State, and Clemson are our three playoffs locks (assuming Clemson avoids an ACC title game loss), with Wisconsin, Washington, and Colorado fighting it out for the fourth spot. Michigan is way back at 8th, and there's no realistic scenario where they can get back to the Top Four.

Sure, we can choose to rank teams by how good they are rather than by how impressive their wins and losses are, but if we do that then we would never have a reason to watch a game like Michigan/Ohio State yesterday. And I enjoyed caring about who won that game.

If you want to give the national title to the best team, just give it to Alabama right now. But if we're going to do a proper playoff system, where the four most deserving teams get in, then we have to leave out at least one of the four "best" teams. We have to leave out Michigan.