Archive for the ‘Strength of schedule’ Category

BCS: On the Merits?

May 6, 2009

College football’s, shall we say, “unique” method of determining a champion stands out for a number of reasons. Among them, resumé supposedly has a lot to do with who plays for the national championship.

Whether or not the system and the voters who make up two-thirds of its rankings actually do reward teams with stronger bodies of work is another issue entirely.
To test the impact of resumé, Homerism collected BCS data on the final rankings prior to the bowl season for the last four years–2005 through 2008. I compiled figures for strength of schedule (SOS) as calculated by the Anderson & Hester rankings, one of the computer polls that make up the BCS system. I then compared strength of schedule against the BCS standings.
(Bear in mind that record clearly has the greatest impact on the BCS standings. The real issue would be whether or not schedule helps teams distinguish themselves from competitors with similar records.)
The universe of 80 teams demonstrated a correlation of 0.184 between SOS scores and BCS averages. As such, this indicates a fairly weak relationship between SOS and BCS scores.
On the other hand, when the teams considered are limited to those only in the top 10 in the BCS rankings, that correlation increased dramatically, up to 0.408. Likewise, the relationship between SOS rank and BCS rank was nearly 0.5. Conversely, among the teams ranked 11 through 20, the relationships between scores and ranks were nearly negligible–around 0.040.
As always, correlation shouldn’t be mistaken for causation. That said, what do these observations suggest?
*The BCS does tend to reward degree of difficulty for higher-ranked teams.
Otherwise, it would seem that the correlations among the top 10 teams should be more in line with the entire group. Looking at 2008, for example, BCS title game participants Oklahoma and Florida ranked 5th and 20th, respectively, in SOS. Meanwhile, other one-loss teams that finished outside the top two had worse SOS rankings: USC (68), Alabama (43), Texas Tech (23), Penn State (52). Texas, which came in 3rd in the final BCS standings, provided a notable exception, with an SOS ranking of 6th.
Similarly, in 2006, the BCS put a one-loss Florida team into the championship game over Louisville and Michigan. The SOS rankings:
  • Florida (11);
  • Michigan (16);
  • Louisville (34).
Of course, exceptions to this trend have arisen. During the 2007 season, eventual national champion LSU finished 2nd in the BCS rankings. However, two other two-loss teams with appreciably higher SOS, Georgia and Missouri, finished behind the Tigers. Likewise, Ohio State ended the regular season ranked number one, while Kansas finished 8th. Yet, both teams finished with similar records and SOS rankings (Ohio State: 60; Kansas: 67).
If you object to the different ways in which the computer polls determine SOS, any analysis like this is pointless. Same applies if you disagree that record should play such a key role in the rankings. However, the relatively strong relationship between SOS and BCS rankings demonstrated here does seem to imply there is at least a little method to the BCS’s ballyhooed madness.