I’ve been following ESPN’s college football conference rankings all year. I like them because they try to balance the impact of the top teams in the polls against the computers. When looking at the final results, I started thinking.
Is their logic sound?
When they started talking about the number of teams ranked in the top 25, I realized that there is an unfair comparison taking place. The SEC has 14 teams while the Big 12 only has 10 teams.
Of COURSE the SEC has more ranked teams.
When I look at the final AP poll of the regular season, there is a 6-2 count of teams in favor of the SEC. However, if you weight the count to normalize 14 teams down to 10 teams, you get 4 teams. (technically 4.3 teams but you can’t have a fraction of a team)
Since the Big 12 only has two ranked teams, there would likely be very little impact to the conference rankings. When I look at Sagarin’s conference rankings, the Big 12 and the SEC are the closest together so it wouldn’t take much to give the SEC the overall lead.
As for the Pac 12, it trails the SEC in both the computer and the polls and since they have only four teams ranked, discounting the SEC isn’t likely to make the difference.
Even though there currently would be no difference doesn’t mean that the method isn’t flawed.
Factoring in Elite
The reason that ESPN factors in the human polls is that they are trying to reward conferences for having Elite teams.Using multiple teams from the whole poll allows for a conference from being over-rewarded for one really strong team. The idea is sound even if it falls apart in execution.
In a twist, let’s limit the benefit. Let’s give credit for only the top 30% of a conference’s teams. For a ten team conference like the Big 12, that would be 3 teams. For the Pac 12 and SEC it would be four teams.
Three teams in a ten team conference should be enough to measure the cream of the crop. For larger conferences, adding the fourth team should bring their average down; helping to compensate for their increased size.
It also insures that some of those “elite” teams will have played each other.
That should be enough to measure “eliteness”. Just take the average vote totals of those teams. If necessary, dig into the Others receiving votes. This gives us:
- SEC: 1241.75
- Pac 12: 855.50
- Big 12: 665.00
- Big 10: 525.25
The Big 12 is hurt in the “elite” category by only having two teams ranked but the SEC’s lead is massive.
Conclusions and Deeper Consideration
It should be noted that this only represents HALF of the equation. In itself, it cannot represent the total quality of a conference.
It should also be noted that the bowl games will shake the rankings up quite a bit though performances in bowl games can be quite uneven.
Finally, there is logic in leveraging computer rankings of individual teams instead of the human polls. The top three conferences using Sagarin’s ratings instead of the AP poll:
- SEC: 92.67
- Big 12: 89.44
- Pac 12: 89.15
While it is a different scale, it is still clear that the SEC has the lead here. No other conference has a top team with a higher rating than any of these averages.
This approach is more impartial than the voters. For example, Oklahoma State’s computer ranking of 16 greatly helped the Big 12 in this method. As a reulst, it better reflects the fact that the Big12 has a deep conference more than the AP rankings.
Of course, there is still the final question, How much influence should “elite” teams play in ranking the Conferences?