Freakonomics For College Football
11 comments so far
July 17, 2006 by
Freakonomics by Steven Levitt and Stephen Dubner has been a runaway bestseller for some time. It’s a book where an economist uses data to answer a bunch of interesting questions like why drug dealers live with their Moms. Being a research analyst by trade myself, I thought about an interesting college football question that there just happen to be some great data for (also, since my name is also Stephen, I figured I was obliged to try).
The question is, can we find evidence of an East Coast Media Bias?
Well it just so happens that our friends at the Bowl Championship Series have been collecting some interesting data for the past 8 years. You see, in addition to polls taken from the Associated Press, the BCS compiles a number of computer polls used to rank teams. Now the method for each computer poll is different but it always comes down to a select few criteria such as:
1) Win-Loss records of teams, their opponents, and their opponents’ opponents
2) Whether a game was at home, on the road, or at a neutral site
3) The final score (including possibly some accounting for overtime games)
4) Recent performances versus past performances
And that’s pretty much it – nothing that would convey an advantage toward a particular team or conference (except perhaps teams from smaller conferences who play good teams on the road and only get weaker teams at home). There may be other factors one thinks should be considered when ranking teams (e.g. weather creates sloppy game conditions leading to a close game or upset, officials make a lot of questionable calls affecting the outcome of a game, garbage time scoring, key players suspended or injured, etc.), computers can’t consider them and AP voters can. But when you look across eight seasons and take the top 15 BCS teams each year (only 15 teams were released in the early years), in expectation those things would even out and not benefit one team or conference. Any apparent differences where the AP rated one team or conference higher should reflect partiality by AP voters toward (or away) from that team or conference.
Expecting to find a bias, I thought that the ACC, Big East, and SEC would be consistently overrated. I was one third right. On average, the 24 SEC teams appearing in the Final BCS standings were rated more than one full spot higher by AP voters than by the computers. The 10 Big East teams were actually ranked a half point lower on average by AP voters than the BCS and the thirteen ACC teams were also ranked a half point lower by the AP than computers on average.
So the theory that teams from the East Coast are getting a boost did not hold up.
What about those poor small schools that “aren’t getting their due”? Well schools like Louisville (prior to joining the Big East), Marshall, TCU, Tulane, and Utah have nothing to complain about. AP voters liked them as well or better than the computers. Miami of Ohio and Boise State were dinged a lot more. The computers ranked Miami of Ohio sixth in 2003 but voters said they were only the fourteenth best team. Boise State ranked seventh by the computers in 2004, but tenth by the Associated Press. Taken together, all these teams netted out a half point higher ranking by the press than the computers. Not a huge case to be made that the press is slighting them (unless they think the computers are too).
How did the rest of the schools and conferences do?
Notre Dame appeared in the BCS three times in eight seasons and on average got boosted one full spot by AP voters. Cheer, cheer for old Notre Dame, I guess.
The sixteen Pac-10 teams appearing in the final BCS standings over the past eight seasons netted out just an eighth of a point lower with AP voters than computers - a trivial difference. So ignore those who complain that people back East are asleep when the games are played. Or perhaps TiVo has rescued the Pac-10.
No surprise the most consistently overrated conference in the analysis was the Big 10. The twenty Big 10 teams appearing in the final BCS standings in the last eight years were also a full point higher on average in the AP Poll than in the computer polls. Who’d have guessed that having all your games on ABC and ESPN would lead to higher rankings?
Get ready to put your angry hat on, Big XII fans. The Big XII was the most consistently underrated conference by AP voters when compared to the computer averages.
The twenty-six Big XII teams appearing in the Final BCS rankings 1998-2005 were rated an average of nearly one full spot lower by the AP than the computer averages. Nebraska was dinged an average of nearly two spots each time it appeared in the final BCS rankings (remember 2000, when Nebraska beat Notre Dame in South Bend only to watch the Irish get crushed by Oregon State in the Fiesta Bowl while the Huskers embarrassed Northwestern in the Alamo Bowl?) Other teams had it worse though.
Texas A&M was downgraded three spots by the AP both times it climbed into the final BCS rankings. Last season, Texas Tech was dinged five spots by the Associated Press (maybe they saw the game in Lincoln). Even the hype of the Red River shootout didn’t spare Texas and OU from being underrated by the AP about a point each. The only Big XII BCS teams that were not underrated were Kansas State (netting out just two-tenths higher among AP voters than computers) and Colorado (a spot and a half higher in the AP than computer polls in two appearances).
It makes you a bit queasy as a Husker fan, doesn’t it? The Buffs and Wildcats get more love than the Huskers. Freaky.
» Enjoy this article?
Send it to your friend or get Husker news by email!
Related Stories: BCS
Post a comment
Ron O'Connor Jul 18 06
There are not 16 teams in the PAC-10. There are only 10. Thats why they call it the PAC-10. Arizona, Arizona State, California, Southern California, UCLA, Oregon, Oregon State, Washington, Washington State, and Stanford. Ten teams, Pac-10.
Steve Jul 18 06
That was never in question. In the past 8 seasons, 16 teams from the Pac-10 have made it into the Top 15 of the final BCS standings (e.g. 2004 USC and 2005 USC both count toward the total)
Quick - how many teams in the Big Ten?
Mick Jul 21 06
Steve,
Excellent article. You did a bang up job. The book might have opened up some interesting ways to look at data. I always thought the Big 12 got over looked but I wouldnt have guessed it was by that much. Now just get ESPN to comment on your article.
Nicco Jul 26 06
TiVO cannot really rescue the PAC-10 because I believe the AP votes shortly after the last game is completed on late Saturday night / very early Sunday morning, so there's not allowable time to watch the games. However, if the vote could be submitted on Tuesday night and made public on Wednesday morning (instead of Sunday at noon), this would provide the voters with the necessary time to view tape and vote accordingly. And then the BCS could be released on Wednesday evening, because those prima donna BCS fellas really don't need an extra day to figure out simplistic spreadsheet data that could be crunched out by grandpa's Commodore 64 in less 10 seconds. Today's multi-Gigahertz systems can spit out a thousand weeks of game data, polls, and rankings in milliseconds. Give the voters their ample time and then quit grand standing and pipeout the info asap.
Nicco Jul 27 06
Steve,
Please, am I to understand that this is what you are saying?
HOW does your college football conference fair in AP favoritism according to FREAKoNOMICS:
+1.5 points
SEC – “more than one point higher” with 24 teams appearing
+1.0 points
BIG 10 - “a full point higher” with 20 teams appearing
NOTRE DAME – “boosted one spot” higher with 3 appearances
+0.5 points
even
PAC10 – (one-eighth) -0.125 lower with 16 teams appearing
BIG EAST – “a half point lower” with 10 teams appearing
ACC – “a half point lower” with 13 teams appearing
-0.5 points
BIG 12 – “nearly one full spot lower” with 26 teams appearing
-1.0 points
It’s safe to assume that 24-year old cub reporters are not invited to be AP voters, that honor is reserved for the old dog sports editors, many of which are contemporaries of Keith Jackson. Isn’t Bear Bryant’s older brother Keith Jackson? :-)
Okay, so maybe there’s not an East Coast “positive” bias at work here, but I’m a born and raised Texan who is aware of Bud Wilkinson’s Sooners dominating the ‘50s, and in the ‘60s it was President Nixon giving the trophy to Texas (and Darrell Royal) instead of Penn State (and Joe Paterno) on national tv. That’s gotta sting to this day. I’ve seen Texas dispose of many good east coast teams over the years. The ‘70’s was a decade of Texas, Nebraska, and Oklahoma having their way with teams from the east. In the ’83 Kickoff Classic, the Cornhuskers murdered Paterno’s boys, 44-6, then in the ’86 Orange Bowl, Switzer, the Boz, and the wishbone dominated another undefeated Penn State squad. Tom Osborne and Nebraska simply dominated the mid ‘90s, and now in this decade (over the six seasons from the 2000 to 2005), a BIG 12 team has played in the BCS Championship Game in five of six years. No other conference has more than 2 appearances:
Pac10 (USC) 2,
Big East (when Miami was a member) 2,
Big 10 (Ohio St) 1,
ACC (Fla St) 1,
SEC (LSU) 1.
You would think that the Big 12 would get some respect, but after reading your article, I have concluded that maybe there’s not an East Coast “positive” bias at work here…..BUT Maybe there is a BIG 12 “negative” bias at work. As a Texan, I might construe your data and my research to equal an East Coast, Mid-East, AND Pacific Coast bias.
Hanway, when you look at the three conferences with the most BCS appearances, the Big 12 (26), SEC (24), and Big10 (20), and the two conferences east of the Mississippi are near or above a full point advantage, and the conference west of the Mississippi sits at a full point disadvantage if you cannot conclude a level field of voting for all, and the East Coast
Steve, do you have an opinion?
Maybe there is a BIG 12 “negative” bias at work. Steve, do you have an opinion?
Steve Jul 27 06
I consider it a television bias. The Big 10 has a TV deal that means anywhere in the country you can watch their games just by turning on the TV (you don't need pay per view, etc.). If I'm not mistaken, the SEC also has a nice deal with CBS.
I live in ACC country and if I want to catch most Big XII games, I need to pay a premium to do so. I think most of these voters won't do that.
Any game that's on TV is going to be hyped by the announcers. They're going to make each team/player/coach sound like something special (even if they're not). When you listen to these teams get praised for 3 hours every week, it's bound to leak into your judgements. If you're only seeing box scores or highlights of Big XII games, it won't have the same impact. Also, most of the voters live outside the Big XII footprint and so they aren't getting the direct coverage by AP voters either.
The Pac 10 has the benefit of being on at night, so any night owls will catch those games (and the three hours of praise) going with it. Not enough to be a net benefit but perhaps it offsets some of the downside of not getting the sweet deals that the Big 10 and SEC have. The Pac 10 also has a decent population base to support it (California, Arizona, Seattle) so they probably have more direct coverage from media people.
Nicco Jul 27 06
HANWAY,
a "television bias"? The last I checked, TELEVISION is a major MEDIA format, as well as being the preferrred format of college football fans.
IF A = B, and B = C, then A = C ....
Unlike 30 years ago, the TV media is intertwined with the PRINT media in more cases than not (ex: ABC, Espn, Espn the Magazine, ESPNs Sunday Morning Sports Reporters = one big company)--- obviously, the lines dividing media sectors do get blurry.
I was once told that two-thirds of the AP voters were from east of the Mississippi...can you confirm?
Other than a rare appearance by a California reporter, all of the reporters who frequent ESPNs The Sports Reporters hail from Big East, ACC, or Big 10 country. They preach Penn State, USC and the Florida 3 on a weekly basis. Oh, forgive me, I forgot about the coronation: that USC was crown the ALL-TIME KING of College Football HISTORY in December 2005 by ESPN (and the ABC family.) How did the Trojans fair in the January 2006 contest against a Big 12 team?
Now, do I believe that respect has been achieved? Do I expect that this victory will change the negative bias away from the Big 12? Actually, I believe JUST THE OPPOSITE WILL OCCUR....that 41-38 win is just another salty twist of the dagger to the biased media's dreams and desires, to further create even more disparity between the Big 12 and the rest of the college football universe. So the haters can keep hating, but the Big 12 is not going anywhere. Your beloved teams will still have to face us on the field of battle because the computers make it so, they provide the boost that keeps the Big 12 near their rightful place. The computers are more reflective of an even playing field. Remember, the Longhorns were #1 in the BCS computers for the majority of last season.
a "television bias"? (Or in essence a MEDIA bias where the large majority of media resides on the East Coast.) Steve, I think you proved your point.
Steve Jul 27 06
Let me put it another way. I don't see anything deliberate on the part of the voters to favor one conference or another. Rather, I think they tend to be influenced disproportionately more by games they see than games they don't. They see more games played by the Big Ten, Notre Dame, and the SEC and it influences their vote.
So where it sounds like you're arguing for an active hatred and deliberate slights, I'd argue the nature of the bias is far more passive. In fact, it might best be described as "lazy".
Zeb Aug 02 06
I agree with your last comment - the bias is not deliberate, and it's due to the factors you mention - media coverage, availability of game broadcasts, etc. - and it is inaccurate to portray any bias as being confined excusively to the East Coast.
What I want to know is how (or if) you factored out the AP influence on the BCS rankings in your comparison. I haven't checked recently, but poll rankings accounted for more than 40% of the BCS score in the early years of the BCS. If you don't remove that from the comparison, you're essentially comparing a number to itself - which doesn't invalidate your results, but it probably does water them down.
Steve Aug 02 06
Hi Zeb,
The aggregated computer component was available separately from the overall BCS tally for each year so my comparison was strictly AP to computers. Otherwise, as you suggest the biases would actually be bigger.
AustinHusker May 02 07
Steve,
When are you going to do another one of these? Here is one I suggest, go back and look at the 'ranking' of each kid that came out of HS with a high ranking and see where they ended up, All American, All Conference, NFO draft or something along those lines and see if the HS ranking means anything.
This is still the best article I think you have ever done. I just quoted it to Stewart over on SI.com.
Good luck
Mick