Any yokel with a computer can have a college football ranking system

December 7, 2012 12:13 PM

He is not the only one. Computer rankings are proliferating, said Kenneth Massey, a professor of math at Carson-Newman in Jefferson City, Tenn., who has been ranking teams since 1995. “When I started, there were six or seven,” he said. “But every year, it gets bigger and bigger.” Massey currently tracks more than 100 college football rankings.

With so many competitors, what is the appeal of creating one’s own rankings?

“It’s kind of a nerdy hobby,” Massey said. “It combines sports with math and computers, three things that don’t ordinarily go together.”

In light of this, David Austin decided to reveal the math behind making one's own ranking and comparing it to the BCS rankings (in this case, from 2010), based on using Hodge theory to rank a large number of alternatives in which the Condorcet paradox may appear.

posted by DynamiteToast (20 comments total)

“It combines sports with math and computers, three things that don’t ordinarily go together.”

Ahem

I think these three things pretty ordinarily go together, and have for quite some time.
posted by Brak at 12:19 PM on December 7, 2012 [1 favorite]

One thing jumped out at me: In the conference rankings, the SEC was ranked behind the Big 12. That is the definition of bugfuck batshit gibbering poop-throwing insanity. 5 of the top 10 teams are SEC teams (compared to two Big 12 teams) not to mention the 6 straight national championships (with another one an extremely strong possibility.)

It's tough to plug into a computer model, but last weekend's Alabama-Georgia game could tell you everything you need to know about the level of football being played in the SEC.
posted by deadmessenger at 12:51 PM on December 7, 2012 [2 favorites]

It's possible the model includes an Insufferability Matrix.
posted by Holy Zarquon's Singing Fish at 12:58 PM on December 7, 2012 [1 favorite]

Hodge theory? That's just showing off.

I'm actually reading Who's #1: The Science of Rating and Ranking right now, which is pretty good. (But pretty math-heavy. Let's say I'm reading it for work.)
posted by madcaptenor at 12:58 PM on December 7, 2012

One thing jumped out at me: In the conference rankings, the SEC was ranked behind the Big 12. That is the definition of bugfuck batshit gibbering poop-throwing insanity. 5 of the top 10 teams are SEC teams (compared to two Big 12 teams) not to mention the 6 straight national championships (with another one an extremely strong possibility.)

Well the bad SEC teams are pretty damn bad. The Big 12 only has one team that's under .500 so that's helping their average.
posted by DynamiteToast at 1:00 PM on December 7, 2012 [1 favorite]

last weekend's Alabama-Georgia game could tell you everything you need to know about the level of football being played in the SEC.

Especially the clock management at the end.
posted by The Hamms Bear at 1:05 PM on December 7, 2012 [2 favorites]

This is as good a place as any to drop my prediction that ESPN-and-Twitter-outrager Northern Illinois University will do better against Florida State in the Orange Bowl than "#1" Notre Dame will do against Alabama in the BCS Championship Game.

(Betting is closed on this because I've already managed to hopefully rankle enough ND fans elsewhere to start 2013 with some extra funds.)
posted by MCMikeNamara at 1:35 PM on December 7, 2012

Well the bad SEC teams are pretty damn bad. The Big 12 only has one team that's under .500 so that's helping their average.

Yeah, but the computer rankings usually take into account what teams you beat to get your wins, and since most of the teams you play are in your conference, a winning record isn't necessarily a strong data point in favor of conference strength -- you might be getting your wins against terrible teams and losses against average/good teams. Of course, deadmessenger's invocation of the last six national championships is even more irrelevant, because the rankings are trying to assess how the teams/conferences are playing this year, not six years ago.

You can't just look at overall winning percentage of each conference (SEC 62% to Big XII 59%) because who knows how many cupcakes each team padded their out-of-conference schedules with. You can't just look at the polls, either. I don't think you can really say which conference is empirically "better" without getting into really the sophisticated analysis these computer rankings are trying to do, so simply saying it doesn't pass the smell test because of your arbitrary criteria isn't a good rejoinder.
posted by tonycpsu at 1:58 PM on December 7, 2012

BTW, I discovered StatMilk the other day, and it's a pretty nice way to get quick access to NCAA football data with a decent interface. I hope they expand it to other sports in the future.
posted by tonycpsu at 1:59 PM on December 7, 2012

deadmessenger, 6 of the top 10 teams in the BCS are from the SEC:

2. Alabama
3. Florida
7. Georgia
8. LSU
9. Texas A&M
10. South Carolina

The AP or Coaches may be slightly different, but the BCS rankings are the ones that matter since they determine who plays for the national championship in the FBS division.
posted by wintermind at 2:03 PM on December 7, 2012

Conventional wisdom is that the SEC is dominant.
Therefore they get inflated support in the polls.
Which is used to buttress conventional wisdom that the SEC is dominant.
posted by fleacircus at 2:08 PM on December 7, 2012 [2 favorites]

Also the SEC enjoys a structural advantage by having two divisions vs. the Big 12's single division.

If you have a single division, you can only have 3 teams with 0-2 losses (conventional wisdom can barely swallow the idea of a 'good loss', and hates 3+ total losses). If you have two divisions, the top teams can be in separate divisions and be protected from playing each other, avoiding those pesky losses.
posted by fleacircus at 2:12 PM on December 7, 2012 [1 favorite]

tonycpsu, one of the huge problems is that strength of schedule was removed from the BCS calculations in 2004. This was allegedly because they're included in the computer rankings that receive 1/3 of the weight in the calculations, but a lot of people believe that it was really a response to LSU jumping USC in 2003 based on strength of schedule to finish #2 over a perennial media darling. I don't have a good link to substantiate the conspiracy theory, but it's considered "common knowledge" among many of us LSU fans, a team that has benefited tremendously from the BCS, flaws and all.
posted by wintermind at 2:13 PM on December 7, 2012 [1 favorite]

fleacircus: " If you have a single division, you can only have 3 teams with 0-2 losses (conventional wisdom can barely swallow the idea of a 'good loss', and hates 3+ total losses). If you have two divisions, the top teams can be in separate divisions and be protected from playing each other, avoiding those pesky losses."

You can still get this with monolithic conferences because you have more opponents than conference games. It's just a bit rarer.
posted by tonycpsu at 2:17 PM on December 7, 2012

That StatMilk site is incredible. Also, I hadn't realized that Baylor is 1st in total offense, and 123rd (out of 124) in total defense. Awesome.
posted by DynamiteToast at 2:20 PM on December 7, 2012

You can still get this with monolithic conferences...

True, though the Big 12 is not big enough for that, and if you a big monolithic conference you should probably split into divisions (or just schedule that way and pretend to be monolithic).
posted by fleacircus at 2:23 PM on December 7, 2012

Well, having nine games on the conference schedule is a relatively new thing. When the Big "Ten" was a monolithic 11-team conference (after adding Penn State, before adding Nebraska) the conference schedule was eight games, you had two teams each year you didn't play.
posted by tonycpsu at 2:30 PM on December 7, 2012

Failure to properly consider strength of schedule leads to such ridiculous things as LSU playing Idaho, North Texas, and Towson as part of their non-conference slate. I'm in favor of just about any change that will penalize good teams who play against very poor out-of-conference competitors.
posted by wintermind at 2:46 PM on December 7, 2012

I think the cupcakes are as much about the goofy economics of college football as they are about getting easy wins. You want as many home games as you can get, so you end up with teams that won't demand a game at their place in return, or who will be happy to give you two home games if you travel to their place once.
posted by tonycpsu at 3:22 PM on December 7, 2012

I know that setting up a good home-and-away with non-conference teams is very hard, and sometimes are good when the schedule is made and not good when the games come around. But it just rubs me the wrong way to watch LSU play an FBS team. And we won't talk about the time that my Tigers lost to UAB in our home opener one year. Stupid Josh Booty. I know you can't expect to play only top 25 teams, but if you're trying to stake a claim that you're an elite program then I think that you need to play other elite teams.
posted by wintermind at 3:39 PM on December 7, 2012