There has been talk for a long time about a single elimination tournament for the top tier of college football teams to replace the current BCS system. Usually it is thought of as an 8 team playoff.
"A committee of university presidents approved a plan for a four-team playoff put forward by commissioners of the top football conferences. ... No. 1 will play No. 4, and No. 2 will play No. 3 on Dec. 31 and/or Jan. 1. The sites of those games will rotate among the four current BCS bowls – Rose, Orange, Fiesta and Sugar – and two more to be determined. "

Note that in the illustration above the matchings have been chosen so that the first two bowls on the left have been chosen so that if what is planned for next year were applied this year, that is, 1 against 4 and 2 versus 3, with the winner playing for the championship in the following bowl. (All outlined in a red square.) In this case the strongest team wins, the "championship." But, note, it is not the final winner in the 8 team field.
On the right the results of ~100 retrials of this identical set up are shown. The two highest ranked teams, Notre Dame and Alabama, win about 1/3 of the time, but notice how well number 5, Kansas State, does.
The results for another 100 runs would differ, and be significantly differ if the original matchups were changed.
The problem with one loss and out tournaments is chance can play a big role, and in football, unlike baseball, multiple games between the same two teams to determine the "true" best team is not practicable.
| TEAM | BCS Average | Assumed* Strength |
|---|---|---|
| Notre Dame | 0.9978 | 10 |
| Alabama | 0.9441 | 9 |
| Florida | 0.8984 | 9 |
| Oregon | 0.8621 | 9 |
| Kansas State | 0.8226 | 8 |
| Stanford | 0.7683 | 8 |
| Georgia | 0.7583 | 8 |
| LSU | 0.7511 | 8 |
| Texas A&M | 0.6756 | 7 |
| South Carolina | 0.6604 | 7 |
The user can choose the teams and change the strength of each to any number from 1 to 10.
Playoff2012 is setup so that, if a team of strength 8 plays a team of strength 2
a great number of times it will win 80% of the games.
In any given 100 such games this number may very from around 75% to 85%,
but usually will be close to 80%.
Finally the user picks the initial pairings.