Monday, October 31, 2011

Why Boise State won't get to the BCS championship game

I wrote up some observations of the week 9 BCS ratings yesterday including some scenarios on who might make it to the championship game.  The latter garnered some good discussion on Twitter centered around what scenario would be required, or if there was one, that would get Boise State in.  I thought it deserved its own analysis so here goes.

The simple view on who gets to the championship game looks at the order of the BCS rankings and handicaps the teams in order.  Using this, assuming no unexpected losses, the LSU/Alabama winner is in the championship game, Oklahoma State is the first in line to meet them, Stanford is next should the Cowboys slip up, and Boise State would be next if both have a loss.  Even though my computer projects all three to finish undefeated, with Oklahoma State still having to play Kansas State, Texas Tech, and OU and Stanford hosting Oregon and going to ASU, the two losses could reasonably happen.

So, is Boise State "in" in that case?  One can make that case, there'd be only two undefeated teams (ok, Houston too) so they should be in right?  That might be the "right" result, but right and wrong doesn't always happen with the BCS.  We have to look at how the BCS, polls, and computers work to get a better idea.

First, it isn't safe to assume that Boise State simply not losing will preserve their position in the BCS.  We already saw Stanford pass them this week, going from 0.02 back to 0.01 ahead, even though it took them three OTs.  This is because the BCS computers, not being able to factor MOV in, can move a team up with a narrow win over a good team or can move a team down even in a win, if the opponent is weak.

Stanford is a perfect example of this as their early schedule wasn't that strong (66.7 by my computer before playing USC) and the computers had them averaging a #9 ranking behind one loss Oklahoma and Arkansas last week before playing USC.  BSU's schedule on the other hand is front loaded, strong enough to date (71.7, on par with Alabama's 71.0, better than Stanford's) but they finish with TCU and four weak teams meaning their schedule could end up around 68.5.  It is having to play these weak teams that could hurt them in the computers.  See Houston with a weak schedule (61.9) being undefeated but an average of 13 in the computers, and not even top-25 in one of them.  BSU's schedule won't get this low, but you get the idea, they could give up a spot or two in the computers.

So what one-loss team might move ahead in the computers?  Oklahoma is already ahead in one computer and no more than 2 spots behind in any other, and if they win out beating Oklahoma State I'd expect they'd be ahead in most if not all of them.  Then the LSU/Alabama loser, particularly if it is Alabama given they'll get to recover playing tough opponents South Carolina and Auburn plus decent opponent Mississippi State, could very well move ahead.  And if Stanford loses it is likely to Oregon so they'd have a chance, but they are a ways back in the computers.

Second, there are pollsters like Craig James that simply refuse to rank BSU high because of their "body of work", and their finishing schedule is not going to convince them otherwise.  It is entirely likely that a one-loss LSU, Alabama, or Oklahoma could at least be close to BSU in the human polls if not ahead because of this.  And if they are even just close, the computers giving the nod to one of them over BSU could move them ahead in the BCS.

For example, BSU is 4th in the computers this week with a 0.870 percentage ahead of OU at #5 and 0.820.  If Oklahoma State and Stanford lose, BSU and OU would move up naturally to #2 and #3, but for the reasons listed above I'd guess OU moves ahead to #2 in most computers leaving OU with a likely 0.960 percentage to BSU at 0.920.  And a one-loss Alabama/LSU could also shake things up so the gap could perhaps be more than 0.04, say at least 0.05.

Given that, BSU would need to make that up in the polls.  Today's polls have OU two spots back, 280 and 120 points and 0.09 and 0.08 percentage respectively, but they have more room to move up and fill the Stanford/Oklahoma State void, and beating Oklahoma State would impress the voters meaning I expect them to be no more than 80 and 40 points back at the worst, or about a 0.02 to 0.03 percentage.  That small gap plus the advantage in the computers puts them in a dead heat for the #2 BCS spot when you do the math.

It would ultimately come down to the poll voters deciding if they are going to vote honestly or if they are going to begin to manipulate the system based on whether they think BSU should be given a chance or not.  You can easily see some voters keeping them behind not just OU but a one-loss LSU/Alabama and perhaps even Oregon as they are ahead of OU today and in this scenario they would have beat Stanford. Some voters could decide to vote them #1 to manipulate things that way, but that is harder to do and less ability to impact as you can only go so high but you can drop them a ways.

In any case, this exact scenario will probably not happen, but it is still interesting to think what will happen if it does.

5 comments:

  1. I can't argue against the brilliance of (fire) Craig James, but let's examine your scenario: Oregon defeats Stanford (possible), Oklahoma wins out, including the Big 12 championship game (possible), and Alabama defeats LSU. LSU has obvious dibs over Oregon. Boise St. wins out. So, the options are:

    Alabama-LSU
    Alabama-Oklahoma
    Alabama-Boise St.

    The first is the highest quality game, but it is a rematch and that's obviously unacceptable.

    The second is still a very good game, but it gives lie to the sanctity of "the greatest regular season in sports". Voters would rebel, and computers can be hacked. Perhaps Anonymous would step in.

    The last option is the obvious people's choice. Butler vs. Duke. In such a scenario the Lords of the Realm would be forced to let Boise St. in. Remember, the computer formulas aren't public so even if it is fishy, Boise St. would be protected in such a scenario. In truth I don't see Stanford losing, and if the Big 12 falls to form Oklahoma will lay down for Oklahoma St. in Bedlam.

    Seth Burn

    ReplyDelete
  2. Seth, you are right that the voters will do everything they can to avoid an Alabama-LSU rematch. Well, all the voters outside the SEC. But that doesn't mean those won't be the best 2 teams and they shouldn't meet again.

    And you are right that there is a segment of voters that will vote for Boise given what they've done the past few years and the purity of having available unbeatens play, but the same argument would say Houston should be there if they finish the year unbeaten and that clearly wouldn't be just. The question is when is a one-loss team more deserving and "better" than a no-loss team?

    I also don't believe the computers will be hacked or any shenanigans will take place with them. They'll just give us the odd results they do. Boise's only hope in this case is that the voters decide this is the year to let them in, and I don't see that happening since OU is already so close and should they win out folks will be forced to move them up so they are close enough for the computers to make the difference.

    ReplyDelete
  3. Do you remember Bowling Green in 1985? They may have been the least respected undefeated team in our lifetimes. They ended up getting pantsed by Fresno St. in the California Bowl 51-7. Houston gets a similar level of respect and wouldn't get the nod over an 11-1 Oklahoma.

    Now, as for more deserving, can we consider past seasons? Boise St. has earned it, particularly if the alternative is Oklahoma. However, it is reasonable to not consider past seasons, even as a voter. Given that, I'd suggest that an 11-1 Oklahoma squad would go undefeated against Boise St.'s schedule more often than not. However, Boise St. would also go 11-1 or better against Oklahoma's schedule more often than not. It is the "or better" that would convince me to place Boise St. ahead of Oklahoma.

    Let's talk about "Deserve"

    http://www.youtube.com/watch?v=5SO5VO2ixWY

    You can skip to 6:35 if you like.

    There is a difference between the best team and the most deserving team. The best team is the one that would be favored on a neutral field. The most deserving team is the one that...?

    Has the best record? Has the best ELO rating? Had the "best season"?

    Boise St. has become a well known team. If they go undefeated you cannot credibly put a 1-loss team ahead of them. That means that the voters must overrule the computers (and again, shenanigans).

    As for LSU, yes, I'd take them over any non-Alabama option, but rematches are awful. Remember, "greatest regular season in sports" doesn't look so hot if the game of the century of the year is meaningless:

    http://www.onionsportsnetwork.com/articles/ohio-state-defeats-michigan-4239-in-ultimately-mea,2094/

    ReplyDelete
  4. There you go trying to insert logic into the BCS. :)

    Without a rule against rematches, the computers can't do anything to preclude it and the LSU/Alabama loser could still be highly rated. The pollsters can vote to avoid it but may have to go to great lengths, i.e. voting the LSU/Alabama loser abnormally low, and go against the idea of voting the "better" team higher.

    Regarding Boise vs OU and schedules, my computer has BSU higher right now by nearly 3 ratings points so it would agree that they should get the nod. The problem is A) the BCS doesn't care about my computer despite it being more accurate that the BCS algorithms and B) OU still has tough opponents to play while BSU has weaker opponents and in the BCS computer's eyes they likely swap places.

    To test this, we can use my "BCS compatible" algorithm that doesn't use MOV. It is not exactly any of the BCS algorithms but behaves similar to some so can be educational to look at. It presently has OkSt, LSU, Alabama, OU, and BSU the top-5, OU being 0.5 ratings points ahead of BSU. It also has Stanford #8 and Oregon #15. If I plug in OU, BSU, UofO (beating ASU in championship game), and Alabama winning out (beating South Carolina in championship game), the rankings would be:

    1 - Alabama
    2 - OU
    3 - OkSt
    4 - LSU
    5 - UofO
    6 - BSU

    So in this scenario Boise could be a ways back behind 4 1-loss teams having an even larger deficit to make up via the polls.

    If LSU beats Alabama, just swap them in the above list.

    If Stanford beats Oregon, interestingly Stanford would only be #5, moving ahead of BSU but not getting past OU, OkSt, and LSU/Alabama loser. My BCS algorithm's current ranking for Stanford and UofO is similar to two of the BCS computers so that may be some insight into how Stanford might get the raw end of the deal.

    But the summary is that my earlier analysis may have been optimistic about how high BSU might be in the computers and it could be worse.

    ReplyDelete
  5. You have convinced me that Oklahoma St. will go undefeated or Stanford won't. Stanford being left out for a 1-loss Oklahoma team is less acceptable than screwing Boise St. This is actually fascinating :)

    ReplyDelete