Monday, October 31, 2011

Breaking News! Stanford could get a raw deal from the BCS too!

I written some BCS observations as well as a detailed write-up on why Boise State could be left out of the championship game even if 3 teams ahead of them lose in the past few day, but after some comments on the latter I did more analysis and believe that Stanford could suffer the same fate as Boise State.

A lot of the analysis and scenario is the same as for Boise State.  If Stanford wins out, it would include wins over Oregon, Notre Dame, and likely Arizona State, so it seems logical that with one of LSU/Alabama losing by definition, if Oklahoma State were also to lose that Stanford would move to #2 in the BCS and play the LSU/Alabama winner in the championship game.  But will that actually happen?

Let's take a look at how things stand today.  Stanford is #4 in the BCS but well back of the top-3 and barely ahead of Boise State.  This #4 is made up of a #3 and #4 ranking in the polls, and an average of #6 in the computers.  They are lower in the computers because their schedule strength to date is weaker than the competition, but with Oregon, Notre Dame, and a potential Pac-12 championship against Arizona State to come, that should be at least somewhat improved.  But they should certainly be #2 in both polls should they win out, even if a few voters have Boise State ahead or even a one-loss LSU/Alabama loser or Oklahoma ahead.

So if they are #2 in both polls, you'd think they would safely be the #2 team in the BCS.  But what could thwart that would be a too low a ranking by the computers coupled with the #3 team in the polls being very close.  How low could they be in the computers?

Today they are a 4, 5, 5, 6, 8, and 8 and #6 overall.  With their tougher schedule at the end of the season and Boise's easy schedule, they'll likely get to #5 ahead of Boise.  But they are already behind Oklahoma overall who is #5 and are behind them in 3 of the 6 computers but also behind Arkansas and K-State in 2.  If OU wins out, they are likely to stay ahead of Stanford in those 3 and could potentially move ahead in another because their remaining schedule is stronger than Stanford's.  And just like they could be behind OU, they could be behind the LSU/Alabama loser as well as potentially Oklahoma State if they lost.  That means they could be #4 or #5 in the computers.

To test this out, I ran a few scenarios.

First, the Colley Matrix, which has a published algorithm, kudos to it, has a feature on the web-site to add games to do what-if analysis.  Adding Stanford over Oregon, OU over Oklahoma, and Alabama over LSU results in Stanford being #4 behind Alabama, OU, and Oklahoma State, BSU at #5 and LSU #6.  And adding Boise State over TCU drops Stanford to #5.  Now each team has other games to play, but with those key game results Stanford is #4 or #5.

Second, I have a BCS "compatible" algorithm (doesn't use MOV) that has Stanford #8 just like 2 of the computers.  So it is similar to those algorithms, and if I run the algorithm with Stanford winning out including beating ASU in the Pac-12 championship, OU winning out, Alabama winning out, and Boise State winning out.  In this scenario, the top-5 are Alabama, OU, Oklahoma State, LSU, and Stanford.

So, Stanford is conceivably #5 in 3 of the computers.  Even if they get to #2 in the other 3, their average could be 4th with a computer percentage around 0.870.  A #2 OU could have a percentage around 0.940 if they are #2 in 3 of the computers.  That would leave Stanford a 0.07 gap to make up in the polls.

If Alabama is the unanimous #1 and Stanford a unanimous #2, Stanford's poll percentage is at best 0.960 in each poll.  An OU with a unanimous #3 would be at 0.920.  But we know it won't be that clean, especially with Boise State taking a few first, second, and third place votes.  If the gap between Stanford and OU gets down to 0.035, say Stanford at 0.950 and OU at 0.915, we get the following equations.

Stanford BCS rating = (0.870 + 0.950 + 0.950) / 3 = 0.923
OU BCS rating = (0.940 + 0.915 + 0.915) / 3 = 0.923

A dead heat.  A vote here and there or a few bits in a computer could sway it either way and Stanford could find themselves losing out to a one-loss OU.  It could be Oregon and Nebraska in 2002 all over again.

Will it happen?  Only time will tell, but it sure is fascinating to think about!