Monday, December 5, 2011

How close did Oklahoma State come to BCS #2 part 3

This will probably be my last post on this topic, but with more details of the Harris poll now available, it deserved an update.

First, let's look at the situation where just one of the Sagarin or Wolfe computers would have had Oklahoma State #2.  I'd previously written that if just 3 Coaches poll voters, the 1 that voted Oklahoma State #5 and 2 of the 5 that voted them #4, had voted them #2, Oklahoma State would have been #2 in the BCS.  So we were 1 computer and 3 coaches ballots away from having a different BCS title matchup.

Now, we can look at what the minimum changes would have been in the Harris poll that along with 1 computer change would have had the same result.  In this case, if the 3 voters that had Oklahoma State #6 (yes, 3 had them there) had instead voted them #2, and just 1 of the voters that had them #5 and voted them #3, Oklahoma State would have been BCS #2.

We can also look at the minimum combination of computer, Coaches, and Harris poll changes.  If again, just one more computer had Oklahoma State #2, and no voter in either poll had voted Oklahoma State #5 or lower, and Oklahoma State picked up just one additional #2 vote in the Coaches poll, Oklahoma State would have been #2 in the BCS.

Now we can ask, how reasonable would this last scenario have been?

If we had transparency into the computers, perhaps we'd find a flaw in the algorithm or the data used, that wouldn't change the algorithm at all, but just make sure it is accurate, and we'd have the necessary change with the computers.  So one data entry error or typo could do it.

In the Coaches poll, one voter had Oklahoma State #5 and seemed questionable as 2-loss Arkansas was ahead of them.  The scenario above would need just this ballot to move Oklahoma State to #2.

In the Harris poll, simply voting Oregon, Boise State, and Houston behind Oklahoma State would get them to #4 or better on all the ballots.  As I wrote in that analysis, it is hard to explain having those teams ahead of Oklahoma State.

So there you have it, we were 1 computer error and simply fixing the unexplainable voting on 8 voters ballots from Oklahoma State being in the BCS title game.

What about if the computers were left the same?  Then it requires more than just fixing strange ballots.

If only the Coaches poll changed, it would have taken 16 ballots moving Oklahoma State to #2.

If only the Harris poll changed, it would have taken all of the #4, #5, and #6 ballots changing to #2s, plus 8 of the #3 votes swapping to #2.

If both polls had changed, a somewhat balanced change that gets Oklahoma State to #2 would have been the 6 Coaches poll ballots with Oklahoma State #4 or #5 moving them to #2, and all of the #5 and #6 votes plus 2 of the 9 #4 ballots in the Harris swapping Oklahoma State to #2.

So, it was awfully close.  It would have been more exciting if Sagarin or Wolfe really would have had Oklahoma State #2 though!  Maybe next year :)