Week 7 and Parity in College Football

The talk of the town these days is the many upsets of top-ranked teams by unranked teams (see Fowler's blog). In recent weeks, USC losing to Stanford, Cal losing to Oregon State, USF losing to Rutgers, and Oklahoma losing to Colorado have caused a stir. Pollsters are hassled every week to come up with a new top 10.

So, why are the "better" teams losing so often? There are four possible explanations that I deem worthy of consideration: One, the distance between teams from top to bottom is less than it has been in the past; Two, bad teams this year have been winning the lottery more than usual; Three, something in the game of football has changed that complicates predicting game outcomes and we have not yet adapted to these new realities; Four, we are overreacting and this year isn't actually unusual.

I here propose some answers. I will be focusing on streaks of single teams being ranked in the top 10. The AP poll has ranked the top 20 teams since 1936. The following are mostly my own calculations (and therefore not error proof) on data from http://www.appollarchive.com/.

First, what has happened this season? Six teams started the season with streaks in the AP top 10 that reached at least 9 weeks (a streak of 9 weeks in the top 10 places that streak in the top 300 all time longest). USC had the longest current streak at the beginning of the season, and 4th longest in history, which ended at 79 weeks. Florida (22 weeks), Louisville (16), Michigan (14), LSU (13), and Wisconsin (10) were the other five. Of these, only LSU's streak is still alive. Oklahoma (8), Cal (7), and Ohio State (6) may be on their way to breaking in as well.

But how does this compare historically? Are we really seeing something unusual this year?

First, we'll look at the number of streaks of 9 or more weeks to begin and end by decade (2000-2010 projections based on results from 2000-2006). Theoretically, a period of college football with greater parity would have fewer streaks that reach 9 weeks. Top teams in recent years, therefore, have been relatively dominant.

	40's	50's	60's	70's	80's	90's	00's*
Begin	35	38	40	42	53	47	46.3
End	33	38	36	42	53	49	51.3

It is projected that 5 more streaks will end than will have begun. This figure is a bit misleading, but it does suggest that this decade might be experiencing more turnover (shorter streaks), than most.

This next table allows us to take a closer look at the length of streaks.

	40's	50's	60's	70's	80's	90's	00's*
Begin	15.6	17.9	21.7	25.3	21.1	25.0	19.7
End	15.1	18.6	19.5	25.2	20.7	26.4	21.5

The numbers represent the average length of streaks (of those that reached at least 9 games) that began or ended, respectively, in that decade. First, as with the last chart, the first three decades of record experienced much higher levels of parity than the last 4 decades. Second, although the 80's saw the beginning of many streaks, these streaks were relatively short. On the other hand, streaks in the 70's and 90's were relatively long. This suggests that, although college football was still dominated by a few teams, the 80's saw either more teams competing than other decades or a significant amount of turnover over the course of the decade with new teams emerging.

This next table counts the number of streaks alive in the average year of a decade (so a long streak would be counted once for each year it included). The results are not surprising--the last four decades have seen more streaks which imply less parity between teams up and down D 1A.

	40's	50's	60's	70's	80's	90's	00's*
Str.	7.3	8.9	9.6	10.8	10.9	11.1	10.3

This final table spells out the real issue at hand here. It is the probability that a current streak will end. Over a decade, a streak counts in the denominator every year that it does not end and in the numerator and denominator when it does end. So, for example, the current decade has a probability of ending of .50, which means that, in any year of that decade, half of the current streaks will end. The number for the current decade is much higher than might be expected, because 83% of streaks that included 2007 have ended.

	40's	50's	60's	70's	80's	90's	00's
Prob.	46.2%	42.7%	37.5%	38.9%	48.6%	44.1%	50.0%

How should we interpret all this information? First, real parity is a thing of the past. Since the 60's, college football has averaged more than 10 streaks every year of 9 weeks or more in the top 10. Considering that there are only 10 spots in the top 10 and only a few more than a dozen weeks of rankings, this implies very little turnover.

But this season is different. Currently, it has the highest percent of streaks ended of any year in the history of college football. So far, no long streaks have started this season, and no season has failed to produce a streak of nine or more since 1941.

The big difference this season is that there aren't any dominant teams like the Miami's, Oklahoma's and USC's earlier in the decade. But I do believe we are seeing some real changes in college football. Inclusion in the BCS system turned ACC and Big East teams immediately into competitors. To some degree, it is an illusion, but this illusion has turned into recruiting and the emergence of programs like USF.

Two forces, therefore, are creating more turnover at the top. First, more teams are perceived as having an opportunity for a top 10 ranking. The AP voters are trying to accommodate the best 3 teams from 6 conferences. Second, top 10 spots are given to teams that are perceived to be deserving (through success in a BCS conference) but less stable--e.g. USF. As I said before, some of this new competition is real. The Big East is much better than in 2004 when a bad Pitt team won the conference. But I don't think it is a coincidence that 8 of the 9 streaks that have surpassed 8 weeks or might surpass 8 weeks this season are the properties of members of the Big 10, Pac 10, Big 12 and SEC.

One more quick thought--college offenses are becoming more sophisticated. The days of the basic I formation are done. I think that more complicated and dynamic offenses also rely more on luck (e.g. the quaterback making the correct read in a split second is not all skill) and put more on single plays (e.g., a single open field tackle) that can affect huge tracts of turf. Throughout the course of a game, little things can have more influence on the final score. This makes it appear that teams vary more in their performances--because they really do vary more and their opponents vary more.

(To Most Recent)