Tuesday, November 1, 2011

New! BCS compliant rankings posted

As I've mentioned numerous times, the BCS computers have been required to look at wins/losses only, no margin of victory, for awhile now and are somewhat handicapped in ranking teams accurately.  However, like any computer ratings propeller head, I've toyed with an algorithm that only looks at wins/losses to see how such a system behaves.  It fails, in my opinion, in both its predictive capabilities and in passing the smell test when compared to my regular algorithm, but it is still fun to look at especially at this time of year.

I've hinted at its rankings in a few BCS related posts lately about Boise State's and Stanford's chances of getting to the title game, but haven't published the full list until now.  You can see the list of all the FBS schools here.

You will see that this algorithm ranks teams most similarly to Sagarin and Massey's, for example we all have Oklahoma State #1, LSU/Alabama at #2/#3 or #3/#2, Stanford #8, and South Carolina #10, and Oregon #13, #14, and #15.  What this means is that I can run scenarios like I did for the Stanford analysis and get a reasonable idea of what those computers might do with the same results.

Like I said, it doesn't predict games as well, but it also doesn't pass the smell test for me either.  For example, it has Stanford behind an Arkansas team at #7 that just eked by a 4-4 Vanderbilt and had a close game with 2-6 Ole Miss, while Stanford's hasn't lost and their only close game all year was against a good 6-2 USC.  And does anyone think K-State is really going to stick around #6 the rest of the year?

The BCS computers have similar odd results, a few the same as my BCS algorithm a few different, so this isn't an indictment of any specific computer (I'll save that for another post), but just pointing out the flaw in not allowing the score to be used in rating a team.

Using MOV, my regular algorithm has Stanford #4 well ahead of #11 Arkansas, and K-State at #23, both passing the smell test for me.  Similarly, Sagarin's original predictor algorithm has Arkansas #28(!) and Kansas State #29.  Massey's original algorithm has Arkansas #9 and Kansas State #20 so other computer operators agree, they just can't use those algorithms with the BCS.

I doubt we'll see the BCS allow MOV again, the AQ conferences and particularly the SEC and Big-12 have benefited from not using it, so were stuck with it.  So I'll keep posting my BCS algorithm to give all of you a little more insight.

Let me know if you have any questions or want any scenarios runs.  Comment here or follow @computerratings on Twitter.