Sunday, December 04, 2016

Overton Windows And The College Football Playoffs

Joseph P. Overton was a long time public policy expert at the Mackinac Center for Public Policy, and he's the namesake for a popular concept in policy wonkery known as the "Overton Window".

In short, the premise of the Overton Window is that on any given issue debate, humans narrow down the set of choices into reasonable options and unreasonable options. We create a range of plausible opinions within which we choose our own personal favorite. To take a political example, imagine the issue of abortion. There are a number of positions that we accept are reasonably held by people, and thus inside the Overton Window - say, that abortion should be legal in the first two trimesters but not the third, or that abortion should be banned except within the first trimester. In contrast, there are positions on abortion that we consider so ridiculous that we'd never spend a moment thinking about and thus are outside the Overton Window - say, that we should be able to kill 6 month old babies, or that any woman who gets an abortion for any reason whatsoever should be thrown in jail for life.

Over time, Overton Windows can shift, and the shifting of Overton Windows in public policy is an area of intense study. But whatever the Overton Window is at any given moment in time is the range of opinions that any normal person will be willing to consider thinking about. Anything outside that window will be dismissed without a moment of rational thought. We don't spend any time thinking about positions outside the Overton Window because they are so preposterous as to be laughable to even consider. 

College Football Rankings Are Dominated By Overton Windows
Imagine if you attempted to honestly try what a football pollster claims to do. Imagine if you tried to analyze dozens of different teams, then take into account all 12 or 13 games that each of them have played, then take into account the hundreds of games played by the teams on their schedules, then take into account home/road, then take into account the scores, then take into account how they looked playing, then take into account injuries, then take into account who finished stronger, then take into account....

It's exhausting even thinking about it. It's impossible. So instead what we do is make it simpler. We start with our generic AP Poll style rankings, and then consider tiebreakers. We all know how the AP Poll rankings work: Teams within each conference are ranked by number of losses, and teams with the same number of losses are broken up by head-to-head and, if not that, by strength of schedule. Then we mash the conferences together by how we view their strengths, so a 1-loss SEC team will be ahead of a 1-loss Big Ten team, and a 1-loss Big 12 team will be ahead of a 2-loss Big 12 team. And so on.

This season, we all considered the debate between 2-loss Penn State and 1-loss Washington as an argument within the Overton Window. But 2-loss Penn State vs 4-loss LSU? No, that is not allowed. Even though plenty of reputable computer rating systems have LSU ahead of Penn State, and there's a good chance that LSU would be favored in Vegas over Penn State on a neutral field, if you were to suggest to somebody arguing that we want to rank "The Four Best Teams" that LSU be considered ahead of Penn State, you would break their brain. The suggestion is outside the Overton Window, and thus beyond the pale to even discuss.

How Well Did The Rankings Correlate With The Computers?
The answer to this is the same as it was last season and will continue to be for the foreseeable future thanks to our tight Overton Windows. The five computer ratings are as follows:

1) AP Poll
2) FPI Strength of Record (measure of resume strength)
3) Massey Ratings (measure of resume strength)
4) Sagarin PREDICTOR (measure of team strength)
5) S&P (advanced measure of team strength)

In general, one would expect the Playoff Committee to basically track the AP Poll, for measures of resume strength to be closer than measures of team strength, and for the S&P to be the furthest off. The reason the S&P will be further off from the Playoff Rankings than the Sagarin PREDICTOR is because the PREDICTOR just looks at scores of games while the S&P looks to control for randomness within those scores, such as teams that have had a lot of turnover luck.

The Playoff Rankings line up with the AP Poll even more than it appears from that graph since the nature of a mean-squared calculation is that a few teams at the bottom of the Top 25 being four or five spots off throws the average. The Top 12 teams are all within one spot of the AP Poll, and the only difference in the Top 9 was Clemson and Ohio State being swapped, a distinction without a difference since they're going to play each other anyway.

In contrast, measures of resume strength (Massey and FPI Strength of Record) were off by an average of around five slots. The Sagarin PREDICTOR was off by an average of eight slots. And the S&P was off by a whopping 14 slots. And that's not just getting dragged down by the bottom of the Top 25. A full three teams in the Top 10 (Penn State, Wisconsin, and Colorado) are at least seven spots off their S&P ranking. That's huge.

So let's say that we were truly ranking teams by how good they were? If we really wanted the "Four Best Teams"? The S&P and Sagarin both say that USC and LSU had as good of a case at Penn State. If we care about best resumes? Well, the Massey ratings have USC in and both Penn State and Clemson out.

Why Wasn't Oklahoma A Playoff Contender, Exactly?

Most media coverage is focusing on Penn State getting left out for Washington, and rightly so. Penn State is out of the playoffs because they played a couple of competitive non-conference games and lost one (on the road), while Washington played three cupcakes and got to the finish line with only one loss. If Penn State had played three cupcakes in non-conference play then they'd have been in over Washington.

But what about Oklahoma? The Sooners are outside the Overton Window why, exactly? The answer is: Two losses. But they have two losses because they lost to Ohio State in non-conference play. Had that Ohio State game been replaced with Rutgers, Oklahoma would have gone 11-1, won the Big 12 title, and been basically a lock for the Playoff Top Four. Even with Ohio State being replaced by Rutgers, Oklahoma's strength of schedule still would have been ahead of 1-loss Washington.

Let's go even further: Why wasn't USC a contender? Three losses, of course. But one of those losses was to Alabama. Take that out and suddenly they're 10-2 with wins over Washington and Colorado and two very tough road losses (Stanford and Utah).

And why wasn't LSU a contender? Four losses, of course. But, again, one was to Wisconsin in non-conference play and another was by 5 points at Auburn. One bounce of the football against Auburn and replace Wisconsin with a cupacke, and LSU is 10-2 out of the SEC West and would have had a hell of a case to pass Washington for that fourth spot.

But if you sat through ESPN's four hour selection show (and God help you if you did), you wouldn't have heard a single person suggest Oklahoma or USC or LSU, because "number of losses" trumps all and pushed those teams out of the Overton Window of acceptable opinions.

How Do You Design A Schedule To Make The Playoffs?

1) Be in a Power 5 league
2) Schedule only cupcakes in non-conference play
3) Lose at most one game in conference play

If you do those three things then you're a near certainty to get in. Nobody is ever going to rank a 1-loss team behind a 2-loss team from the same conference. In fact, it has literally never happened in any of the three College Football Playoff seasons that a team was ranked behind a team from the same conference with more losses. Never. Once.

Ohio State didn't get in because they beat Oklahoma - they got in because they lost one fewer game than Penn State. One of the biggest breaks the Buckeyes got was not having to play in the Big Ten title game and having to risk a loss to Wisconsin that would have knocked them out of the playoffs. They had their one loss, and so they were locked into the playoffs by sitting home on the couch.

There might not be another area of sports commentary with an Overton Window as tight as college football polling. "Number of losses" has such a complete chokehold over the sport that it suffocates all other debate. And until that changes it harms the sport, since any team with national title hopes has a strong incentive to schedule nothing but cupcakes, and invitations to conference title games are punishments rather than rewards. Don't we want more competitive games to watch?


BW said...

Hey Jeff, obviously it doesn't change the crux of your argument but in the interest of accuracy Oklahoma State (3 losses) was actually ranked ahead of West VA (2 losses) in today's CFP. However, I think this is an outstanding piece that mainstream media members should be required to read and agree with it all.

Keep up the good work.

Jeff said...

Thank you.