In response to Chase Brown not understanding the way the BCS worked in today's Skull Session, I have decided to revive this weekly forum post (I skipped last year).
It's always fun to see what the BCS would have done if their rankings still existed in place of the Selection Committee. For the uninformed, the BCS used an average of the Coaches Poll, the Harris Poll, and a Computer Average ranking to compute what was called the BCS Average. This average was used to rank teams, and the top 2 at the end of the season would play in the BCS National Championship game.
To arrive at the computer average, the following procedure would take place. Each of the 6 computers would rank the top 25 (those computers were Anderson & Hester, Billingsley, Colley Matrix, Massey, Sagarin, and Wolfe). You would receive 25 points for a ranking of 1, 24 for a ranking of 2, all the way down to 1 point for a ranking of 25. Anything below a 25 received 0 points. Then the top and bottom computer score for each team would be thrown out, so a maximum of 100 points were possible. Divide by 100 to arrive at your computer average. The Coaches Poll, Harris Poll, and Computer Average had equal weight and each provided 1/3 of your total BCS Average.
As a bit of note - the Harris Poll was created for the BCS when the AP Poll opted out, and I have substituted the AP Poll back in it's place due the the Harris Poll being disbanded after the 2013 season. In addition, only 4 of the 6 computers are currently providing rankings (Anderson & Hester and Wolfe have not yet provided rankings for 2023), so I am taking those 4 rankings as is without throwing any of them out. So it's not exactly the same formula but pretty close.
On to the rankings...
| Team | Record | BCS Rank | BCS Average | AP Poll | AP Avg | Coaches | Coaches Avg | Computer | Computer Avg |
|---|---|---|---|---|---|---|---|---|---|
| Georgia | 5-0 | 1 | 0.9172 | 1 | 0.9684 | 1 | 0.9931 | 5 | 0.7900 |
| Texas | 5-0 | 2 | 0.9044 | 3 | 0.9200 | 4 | 0.8631 | 1 | 0.9300 |
| Michigan | 5-0 | 3 | 0.9020 | 2 | 0.9265 | 2 | 0.9394 | 4 | 0.8400 |
| Ohio St | 4-0 | 4 | 0.8946 | 4 | 0.8755 | 3 | 0.8781 | 1 | 0.9300 |
| Penn St | 5-0 | 5 | 0.8224 | 6 | 0.7917 | 6 | 0.7756 | 3 | 0.9000 |
| Florida St | 4-0 | 6 | 0.7802 | 5 | 0.8620 | 5 | 0.8588 | 8 | 0.6200 |
| Washington | 5-0 | 7 | 0.7640 | 7 | 0.7826 | 8 | 0.7494 | 7 | 0.7600 |
| Oregon | 5-0 | 8 | 0.6790 | 8 | 0.7181 | 9 | 0.6988 | 8 | 0.6200 |
| Alabama | 4-1 | 9 | 0.6656 | 11 | 0.5942 | 10 | 0.6225 | 6 | 0.7800 |
| USC | 5-0 | 10 | 0.6639 | 9 | 0.6949 | 7 | 0.7669 | 11 | 0.5300 |
| Notre Dame | 5-1 | 11 | 0.5791 | 10 | 0.6291 | 11 | 0.5781 | 11 | 0.5300 |
| Oklahoma | 5-0 | 12 | 0.5657 | 12 | 0.5420 | 12 | 0.5550 | 10 | 0.6000 |
| North Carolina | 4-0 | 13 | 0.4618 | 14 | 0.4684 | 13 | 0.4969 | 14 | 0.4200 |
| Washington St | 4-0 | 14 | 0.4060 | 13 | 0.4936 | 14 | 0.4444 | 17 | 0.2800 |
| Mississippi | 4-1 | 15 | 0.4044 | 16 | 0.3974 | 15 | 0.3856 | 13 | 0.4300 |
| Oregon St | 4-1 | 16 | 0.3524 | 15 | 0.4084 | 16 | 0.3788 | 18 | 0.2700 |
| Miami FL | 4-0 | 17 | 0.3146 | 17 | 0.3800 | 17 | 0.3538 | 23 | 0.2100 |
| Utah | 4-1 | 18 | 0.2835 | 18 | 0.2929 | 19 | 0.2675 | 15 | 0.2900 |
| Tennessee | 4-1 | 19 | 0.2512 | 22 | 0.1974 | 18 | 0.3063 | 19 | 0.2500 |
| Missouri | 5-0 | 20 | 0.2264 | 21 | 0.2123 | 22 | 0.1769 | 15 | 0.2900 |
| Kentucky | 5-0 | 21 | 0.1748 | 20 | 0.2219 | 20 | 0.2225 | NR | 0.0800 |
| Duke | 4-1 | 22 | 0.1557 | 19 | 0.2478 | 21 | 0.1994 | NR | 0.0200 |
| LSU | 3-2 | 23 | 0.1306 | 23 | 0.0961 | 23 | 0.1056 | 24 | 0.1900 |
| Louisville | 5-0 | 24 | 0.1233 | 25 | 0.0581 | 25 | 0.0619 | 19 | 0.2500 |
| Texas A&M | 4-1 | 25 | 0.0990 | NR | 0.0200 | NR | 0.0369 | 21 | 0.2400 |
It's fun to make observations about this data. At first, it looks like Ohio State is undervalued by the human polls since they are tied with Texas at 1 in the Computer Average. Then I see Alabama at 6 in the computers and I realize that 5 weeks isn't enough for anything meaningful. Also, some teams have played more games which introduces bias, particularly in Colley's formula (his formula starts every team basically at 0.500 and add/subtracts for each win/loss, so early in the year when teams have an unequal number of games played can skew the rankings).
This will get better closer to the end of the season, but to be honest, it's really not that far off of what my gut says things should be. You actually have to play games for computers to rank well, they aren't impacted by projections or poll inertia, so they are going to take time to adjust. But not too bad after 5 weeks.
Curious as to everyone else's thoughts.
