Writers of Pro Football Prospectus 2008

Infographic: Declining Frequency of Top 25 Games in College Football

This infographic is the first in a series I'll be producing throughout the off-season at my blog, BCF Toys. Comments and suggestions for future infographics are welcomed.

In 1989, there were 106 Division 1A (now called FBS) teams. A total of 582 games were played between those teams, including 18 bowl games. 52 of the total games (8.9 percent, or about 1 in 11) were played between teams ranked in the Associated Press final top-25.

In 2009, 120 FBS teams played a total of 714 games against one another, including 34 bowl games. Only 38 of the total games (5.3 percent, or about 1 in 20) were played between Associated Press final top-25 teams.

The AP final top-25 was significantly more connected in 1989 than 2009. Only nine ranked teams played at least four games against other ranked teams last season; in 1989, 18 ranked teams did so. Twenty years ago, the AP top-10 either played or shared a common opponent with an average of 17 other ranked teams. In 2009, the AP top-10 either played or shared a common opponent with an average of only 12.6 other ranked teams.

In twenty years, the frequency of games played between top-25 teams has been nearly cut in half. The primary reason for the decline has been conference expansion. In 1989, 25 teams were independent, including AP final Nos. 1, 2 and 3 and six of the top-25 overall. In 2009, only 3 FBS teams were independent, none of which were ranked. Additionally, there were 94 FBS vs. FCS games played last year, 17 involving AP top-25 teams. Only 50 such games were played in 1989, two by AP top-25 teams.

1989 2009 APTop25 Connectivity

Comments

1
by Dean :: Tue, 03/02/2010 - 5:35pm

What of those games are conference games which cannot be controlled, and what of those games are non-conference?

2
by JasonK :: Tue, 03/02/2010 - 5:51pm

My knowledge of college football is nearly nil, but I wonder how much, if any, of this can be explained by an increase in the propensity of AP voters to put teams outside the big-name conferences in the final top-25.

Also, who is the author of this post?

5
by masoch (not verified) :: Tue, 03/02/2010 - 6:08pm

Brian Fremeau

15
by mm (not verified) :: Tue, 03/02/2010 - 10:04pm

I think this is part of it, as reporters seem to take pleasure in giving non-BCS teams high rankings.

I'd like to see this done with a computer ranking system. It may have a bias, but it would not change through time.

23
by Pat (filler) (not verified) :: Wed, 03/03/2010 - 3:18pm

Except in 1989, the top 3 teams were independents. No big name conference to put there. They were putting big names, but, well, Boise State, Utah, etc. are 'big names' now anyway.

The problem isn't a media bias, it's exactly what Brian said: conference expansion. This is why anything which encourages teams to spread out more among the conferences (i.e. more autoberths, increased revenue sharing to minor conferences) is bad. There are way too many teams, and forced games against crappy teams just means more crappy games.

3
by 2468ben :: Tue, 03/02/2010 - 5:52pm

So 1989 had 52 of those matchups and 2009 had 38, but can you post the totals for the 18 years in between? I want to learn more about this, but two samples isn't enough to relieve the swelling in my statisticles.

13
by Crushinator :: Tue, 03/02/2010 - 9:31pm

I agree with most of the above post.

Maybe its just me, but it seemed like 2007 and 2008 had a very high percentage of games played between top 25 teams. I kind of wonder if last year was just an outlier.

4
by Danish Denver-Fan :: Tue, 03/02/2010 - 5:59pm

Fancy graph!

I know next to nothing about college football, but why this trend? I mean, if anything top-games should generate more attention and money? Is it because all the teams have, suddenly, become afraid of an early loss or two thus falling out of the BCS championship?

9
by Kibbles :: Tue, 03/02/2010 - 7:45pm

A lot of it is just the death of the independent. For teams in a conference, maybe 75% of their schedule is played against other teams within their own conference. That results in FANTASTIC connectivity within the conference, but very little opportunity for connectivity to other conferences. Back in the day, there were a lot of independents that served as a go-between, though. Since they didn't have a conference, 100% of their games were "out of conference" games. They wound up becoming great connectors, providing links between all the different conferences. Teams in the Pac 10 might not play teams in the Big 10 all that much... but teams in the Pac 10 played Notre Dame, who then played teams in the Big 10, which then provided a 2nd-order connection between the Pac 10 and the Big 10. Now, Notre Dame is still an independent and they still serve as a great connector, but where Notre Dame was once joined by dozens and dozens of other independents (including several very good teams like FSU and Miami), now the only other independents are Army and Navy. Heck, even Air Force has joined a conference.

10
by Kevin from Philly :: Tue, 03/02/2010 - 8:36pm

More likely, it's because the top conferences don't have the monopoly on the rankings that they used to have. Blame that on NCAA scholorship limitations and smaller rosters. Back when leagues like the Big-10, PAC-10 or Big-12 might have 4-5 ranked teams, of course there'd be interconference games between those ranked teams. Now, when you're likely to see a non-BCS team like Boise State or Utah or even a Ball State in the 20's for a part of the season, it's got to be less likely that interconference games will be between ranked teams.

24
by Pat (filler) (not verified) :: Wed, 03/03/2010 - 3:51pm

Nope.

1989: Independent: 5 (MIA, ND, FSU, PSU, WVU) Big 8: 2 (Nebraska, CU). SEC: 3 (Tennessee, Alabama, Auburn). Big 10: 4 (Michigan, Illinois, Michigan State, Ohio State). ACC: 2 (Clemson, Virginia). SWC: 4 (Arkansas, Houston, Texas Tech, Texas A&M). Big East: 1 (Pitt). PAC-10: 3 (USC, Washington, Arizona). WAC: 1 (BYU).

That's 6 guys not from 'major' conferences.

2009: Independent: 0. Big 12: 3 (Texas, Nebraska, Texas Tech). SEC: 4 (Alabama, Florida, LSU, Mississippi). Big Ten: 4 (OSU, Iowa, PSU, Wisconsin). PAC-10: 2 (USC, Oregon). ACC: 4 (Va. Tech, Ga. Tech, Miami, Clemson). Big East: 3 (Cincinnati, Pitt, WVU). WAC: 1 (Boise State). MWC: 3 (TCU, BYU, Utah). MAC: 1 (Central Michigan).

That's 5 guys not from major conferences, and in the next few years the MWC will probably become a 'major conference.'

The reason for the dropoff is the conference growth, loss of independents and the increasing number of FCS games, although primarily it's the loss of the independents and conference growth. The 'top conferences' never had a monopoly on rankings, and just last year there were 3 conferences with 4 teams in the Top 25.

31
by Travis :: Thu, 03/04/2010 - 10:31am

Pitt was still a football independent in 1989. The Big East football conference didn't start until 1991.

14
by mm (not verified) :: Tue, 03/02/2010 - 10:00pm

In addition, conference expansion has led to an increase in the number of conference games each year, leaving fewer games for non-conference opponents.

Before the SEC added 2 schools in 1992, the SEC schools played 6 conference games; now they play 8. Even though there is 1 more game played now, that is 1 less opportunity to schedule against a non-conference opponent.

Of course 14 schools joining division 1 also means its less likely (just by chance) that 2 teams meeting will be in the top 25.

22
by GoVikes (not verified) :: Wed, 03/03/2010 - 3:05pm

Much of it also has to do with the BCS. There was another stat released not too long ago (and I can't seem to find it now) that indicated that this past year was a record in the number of 1-A teams taking on 1-AA squads (sorry I think the new names the NCAA uses for the divisions are bogus), whereas in 1989 it was unheard of for teams to do that. The point is, going undefeated in a BCS conference is your only guaranteed ticket to the title game. Strength of schedule is not considered in the computer rankings, and the polls rarely reflect it either. So an 11-1 squad in a BCS conference whose only loss was on the road to a ranked opponent will still get ranked behind the 12-0 squad who loaded their non-conference schedule with 1-AA opponents and rollovers from the Sun Belt Conference (easily the worst in 1-A). Even if a team has more wins over ranked opponents than an undefeated squad the lone loss will essentially keep them out of the title game.

As a result, you get the opposite of college basketball. Since everyone plays in the same tournament come March, it makes sense for high-powered schools to take on other ranked opponents from other conferences - it not only improves your chances of getting into the tourney and/or getting a higer seed (as strength of schedule is weighed heavily into the equation), but you can scope out the big competition, hence why small schools like Gonzaga travel alot and take on big-ranked opponents - both schools win. In football, the incentive is to get away from high-ranked non-conference games. And as smaller-conference schools like Boise State, Utah, BYU, and TCU get ranked this disparity only increases, because the BCS formula gives little reward for beating the mid-majors, but punishes greatly for losing to them.

There are alot of other factors into play, but if 1-A had a playoffs format I'd bet my next paycheck that we'd see a drastic spike in non-conference games between ranked foes.

25
by Pat (filler) (not verified) :: Wed, 03/03/2010 - 3:57pm

The point is, going undefeated in a BCS conference is your only guaranteed ticket to the title game

This is not the reason why they're scheduling FCS opponents. It's economics. ADs don't schedule teams to try to get to the title game. They don't care. They schedule teams to get home games and make money.

Strength of schedule is not considered in the computer rankings, and the polls rarely reflect it either.

The computer rankings are, in fact, nothing but strength of schedule rankings weighted properly by victories/losses. The reason why going undefeated in a BCS conference is typically a good way to get to the title game is because it's really, really hard to go undefeated in a BCS conference.

In years where a BCS conference is relatively weak compared to others - say, the Big Ten in 2008 - going undefeated probably wouldn't be enough.

27
by peachy :: Wed, 03/03/2010 - 4:17pm

Agreed twice over - for the vast majority of schools, getting to the title game isn't a consideration. They schedule I-AA teams for a (cheap) revenue-enhancing home game (and - in some cases - a high-probability win to pad the numbers for a bowl bid.) For the legitimate heavyweights, who do care about getting into the title picture, there's effectively no difference on the field between a bottom-feeding I-A weenie and a I-AA team (if anything, the top AA teams are better - still a distinction without a difference if the heavyweight is legit); again it's a matter of cash and availability.

And, as you say, the computer algorithms have basically nothing to go on but record and SOS due to the official exclusion of MOV; I'm not sure why football needs six formulas, when basketball can get by with one.

30
by Pat (filler) (not verified) :: Wed, 03/03/2010 - 7:52pm

Limited number of games means that undefeated teams show up quite often - exactly how you deal with an undefeated team is ill-defined in a maximum-likelihood ranking, so yeah, you're best off using the average of several. Also, the ridiculously poor connectivity means that how you handle huge disconnects (FCS!) varies, as well.

Still, though, several of the football rankings aren't worth anything (Billingsley) some are unjustifiable (Massey's time-weighting) and some aren't even described anywhere (Anderson-Hester).

Colley, Wolfe, and Sagarin are the three that are most solid from a theoretical point of view. There are merits to all three, so I don't have a problem with having all three of them.

28
by GoVikes (not verified) :: Wed, 03/03/2010 - 5:27pm

Ok, I agree that economics plays a role in it, but we're talking about the dropoff in ranked matchups. The powerhouse conference teams (the ones ususally ranked) go out of their way to schedule creampuff opponents - whether they be 1-A bottomfeeders or 1-AA pipsqueaks. You can't divorce a title run from the economics - being a "potential contender for the BCS title" means greater interest in your school, a more active booster/alumni base, more fan attendance, and more sales in merchandise as your team inevitably gets more media coverage. That's way more valuable than the additional home game, both in revenue and in the additional free advertising your school recieves from the publicity. Even if you stumble, a 1-loss school in a BCS conference will get huge considerations for the other BCS bowls or the prime-time New Year's games that have big paydays (see Penn State this year or Kansas in years past). The top schools in the conferences would make the most difference in the ranked-opponent dropoff. There is certainly a financial incentive to play other ranked opponents - most of the time it's a 2-year "home and home" deal where the teams swap home games so there's not much of a loss there, but playing big opponents in the regular season boosts your school's profile.

And no, the BCS doesn't really do much to differentiate strength of schedule. There has never been a time in BCS history where a 1-loss program from a BCS conference was selected over an undefeated school in another BCS conference. That's why Ohio State got its sorry ass to all those title games. The BCS can claim it's part of the formula, but mostly it judges you based on how many other opponents from BCS schools you've played, thereby making it a self-perpetuating process. The committee has even said they gave more weight to the polls specifically because they believed the coaches would consider SOS better than they would, but they haven't. You're right, it is all about money. And if you're a top school in a BCS conference, your relevancy is based on your perceived ability to get to a BCS bowl or the title game, which is why you play weak non-conference opponents.

29
by Pat (filler) (not verified) :: Wed, 03/03/2010 - 7:47pm

The powerhouse conference teams (the ones ususally ranked) go out of their way to schedule creampuff opponents

Because those guys they can get cheaply. It has nothing to do with the quality of the team. Michigan's loss to Appalachian State didn't make teams suddenly go out and try to sign the worst 1AA teams possible. They don't care about whether or not their team is likely to win or lose. They care about the money - they make more money from 2 1AA home games than 1 game vs Texas at home and 1 game vs Texas away. That's the economics.

There has never been a time in BCS history where a 1-loss program from a BCS conference was selected over an undefeated school in another BCS conference

Right, because it's really hard to go undefeated in a BCS conference! It has nothing to do with the fact that they schedule creampuffs in OOC play. If you go undefeated in a BCS conference, you have a really good shot of winning all of your OOC games anyway, regardless of who they're against.

Find me a time in the past 5 years where the BCS got it wrong - where an undefeated team deserved to be ranked higher than a 1-loss team, based on a statistical ranking.

You can say "you shouldn't be deciding things by statistical ranking! won-loss means all!" if you want, but if that's your argument, you're complaining about scheduling, not the BCS.

Even if you stumble, a 1-loss school in a BCS conference will get huge considerations for the other BCS bowls or the prime-time New Year's games that have big paydays

Yes, that's because other than the National Championship game, the other BCS bowls are only concerned with getting attendance, and gasp several teams in autoberth conferences have huge fanbases.

It's not the fact that they're "1-loss in an autoberth conference." It's the fact that they're a huge school. No one would be jumping up and down happy to get, say, Northwestern or Illinois.

That's why Ohio State got its sorry ass to all those title games.

End-of-season Colley rankings, 2002: Miami #1, Ohio State #2.
End-of-season Colley rankings, 2006: Florida #1, Ohio State #2.

Ohio State was not undefeated in 2007. They didn't go "just because" they were undefeated in an autoberth conference. They went because their schedule - the whole thing, out of conference and in conference - validated it.

The BCS doesn't "claim" anything about the formulas. The statistical rankings are just math, sometimes from the 1700s based on the very simple assumption that "if A beats B, A is likely better than B." That's it. Then you add the human polls, and there you go.

The committee has even said they gave more weight to the polls specifically because they believed the coaches would consider SOS better than they would, but they haven't.

That's completely untrue, actually. They gave more weight to the polls because they differentiate teams with more clarity than the statistical rankings do because they can fairly deal with margin of victory.

Seriously, if you don't think the statistical rankings consider strength of schedule, you don't understand how they work. They removed the additional "SOS" component because it was double counting since it's already the foundation of the statistical rankings anyway.

And if you're a top school in a BCS conference, your relevancy is based on your perceived ability to get to a BCS bowl or the title game, which is why you play weak non-conference opponents.

Weak non-conference opponents do not help you get to a BCS bowl. I don't get why this is so hard to understand. If Penn State schedules USC, Texas, and Alabama, and loses all three, and sweeps the rest of the Big Ten, they're almost without a doubt going to a BCS bowl, since they probably will have just won the Big Ten.

With regard to the title game, if you don't play difficult opponents, those opponents will face someone weaker - and they're your prime competition for getting into the game in the first case, and going undefeated in a BCS conference is already harder than beating one team.

6
by drobviousso :: Tue, 03/02/2010 - 6:46pm

Comments and suggestions for future infographics are welcomed.
Don't bury the lede. Put the pretty picture first, and the talky stuff after.
Otherwise, infographics are cool, and you should have more of them.

7
by Chris M (not verified) :: Tue, 03/02/2010 - 6:55pm

Yeah, I'd be curious to see when this drop happened. Was it gradual? Or did it fall off a cliff somewhere? This would help us tell whether the drop came from teams like FSU and Miami joining conferences, or from adding non-BCS conference teams to the top 25 (or both).

8
by Ryan D. (not verified) :: Tue, 03/02/2010 - 7:02pm

How many of the pairings took place in bowl games? These are essentially "forced pairings." Removing the forced pairings of bowl games and conference games might tell us more about any trends in top teams not scheduling contests against one another, presumably out of national title contention hopes.

11
by Eddo :: Tue, 03/02/2010 - 9:16pm

Good point, especially considering how many more bowls exist today than twenty years ago.

As I type that, I realize that the extra bowls, for the most part, allow non-top-25 teams to play against each other, but it's still worth exploring.

12
by Still Alive (not verified) :: Tue, 03/02/2010 - 9:22pm

I agree this means little without all the years, and I liked some of the other people's suggestions as well.

Also, I know 1/20 vs 1/11 looks a lot more impressive than 1/19 vs 1/11, but 5.3 basically IS 1/19. Fudging numbers is a habit you expect from Peter King or Bill Simmons, not a statistics site.

That is the road which many of us fled to get away from when coming here.

16
by mm (not verified) :: Tue, 03/02/2010 - 10:11pm

I mentioned I'd like to see it done with a computer ranking system, but it'd also be interesting to see using of the year rankings, since they generally reflect the quality of the teams during the season better. It's always annoying when you hear how a team defeated 3 'top ten' teams, but none of those teams finished the year anywhere near the top 10.

17
by Alexander :: Wed, 03/03/2010 - 12:16am

You (almost) all seem to be ignoring the fact that there is no incentive for good teams to play other top 25 teams. If you are Texas, Florida, Alabama (Hell they even played VA Tech and didn't have to at all) there is no reason to play each other.

Games like USC-Ohio State are the exception, because that game basically ruins one of those teams' shot at a national title.

18
by Tim Gerheim :: Wed, 03/03/2010 - 12:34am

I think that latter point may suggest another answer. I'd say there's probably more depth in the FBS now than in 1989, so it may take fewer losses to drive a team out of the top 25. Then the team(s) that beat them don't get credit in this graph. The number of games between teams that are ranked at the time they play (which would be MUCH harder data to mine) might be more useful, at least if the object is to see if teams are facing genuinely easier schedules. (The FBS-FCS numbers are a pretty good indication of that, of course.)

19
by Tarrant :: Wed, 03/03/2010 - 10:37am

The more striking thing, IMO, is the striking reduction in non-conference games between two big teams, regardless of ranking.

The problem is that, given the current BCS structure, there is no strength of schedule that can make up for a loss. The reason for getting rid of strength of schedule in the BCS formula was that both the polls and the computer rankings can take strength of schedule into account. But the reality is the polls simply don't.

A team from a BCS conference that goes 12-0 with a Division I-AA win and 3 other nonconference games against Tungsten Tech/non-BCS schools will almost always be ranked ahead of an 11-1 team in the BCS rankings, even if that 11-1 team had all its nonconference games against other BCS schools, and even if all 4 were against ranked teams and the loss was to the #1 team. The computers would likely rank that 11-1 team #1, but the polls overwhelm the computers in the formula.

Every year football writers, fans, everything, complain that it's rare to see marquee nonconference games anymore - USC still schedules them, Texas has, Ohio State does - they do exist - but the reality is that as long as only two teams at the end of the year have a shot at the title, there is no incentive to do anything that could possibly knock you out of one of those slots. Going undefeated and being in a major conference usually gets you a spot (unless you're Auburn). Having the toughest schedule in the nation - even if every single person admits it was the toughest schedule - and being 11-1, doesn't.

21
by Eddo :: Wed, 03/03/2010 - 1:45pm

"Going undefeated and being in a major conference usually gets you a spot (unless you're Auburn)."

Or Cincinnati.

26
by Pat (filler) (not verified) :: Wed, 03/03/2010 - 4:07pm

A team from a BCS conference that goes 12-0 with a Division I-AA win and 3 other nonconference games against Tungsten Tech/non-BCS schools will almost always be ranked ahead of an 11-1 team in the BCS rankings

Penn State wouldn't've in 2008. Not to mention the Cininnati or Auburn example. People constantly put this idea out there, but it's just not true. Playing a relatively weak schedule would've killed Penn State's shot in 2008.

But it wouldn't've been from their OOC schedule - it would've been due to the in-conference weakness. But that's a 'duh' - you play 2-3 times more games in conference. A weak or strong out-of-conference schedule just doesn't have nearly as much effect. Why would it?

The computers would likely rank that 11-1 team #1, but the polls overwhelm the computers in the formula.

Not really. The Harris/Coaches' Poll ranks are weighted by actual votes, so if #1/#2 are close, the fact that one is ranked #1 and the other is ranked #2 is pointless. Statistical ranks are not: the #1 ranked team gets all the weight, regardless of how close.

That alone is enough to make it so that the statistical rankings can easily overwhelm the human voters. It's just 2/3, 1/3, after all.

This is all academic anyway, as teams don't schedule based on whether or not they think their team will win. It's economic, completely and totally. After all, if you're worried you're going to lose to Oregon State, in what bizarre world are you thinking your team can make the National Championship?

20
by Brian Fremeau :: Wed, 03/03/2010 - 10:42am

Thanks for the feedback, everyone. I'm going to put together a more comprehensive table with as much of your suggested data as possible, probably posted at the end of the week.

32
by Brian Fremeau :: Fri, 03/05/2010 - 5:24pm

See this post at BCF Toys for a table filling in the gaps between 1989 and 2009.
http://bcftoys.blogspot.com/2010/03/20-year-ap-top-25-trends.html

I have also edited the original post as my 1-in-11 vs. 1-in-20 comment was misleading and inaccurate. 39 games were played in 2009 (not 38; Clemson and Georgia Tech played twice).