Sunday, December 4, 2011

How close did Oklahoma State come to BCS #2? Did anything fishy take place?

So what we all thought would happen was confirmed.  LSU and Alabama finished #1 and #2 in the BCS and will face each other in a rematch for the BCS championship.  But how close did Oklahoma State get to passing Alabama?

First, lets take a look at the final BCS numbers for the two teams,

Oklahoma State0.93331,3670.92682,6540.92310.950

From this, we can see that the margin was a mere 0.0086.  Let's take a look at how close that was in each of the components.

First, in the computers, Oklahoma State was #2 in 4 and #3 in 2 of them.  When you throw out the high and low it was a 3 to 1 advantage resulting in the 0.950 to 0.930 advantage.  But what if one of the two computers that had the Cowboys #3 had them #2 instead?  Sagarin had them close, 99.49 to 97.81, and Wolfe had it 8.969 to 8.700.  If either one of them swaps the two teams, the computer margin changes to 0.960 to 0.920.  That would make the totals like this.

Oklahoma State0.936613670.926826540.92310.960

Now the margin has dropped to 0.0019, much closer.

Let's take a look at the polls now.  In both, it is safe to assume that Alabama was either #2 or #3 on each ballot.

That means in the Coaches poll they received 42 second place votes and 17 third place votes resulting in the total of 1,399.  Then lets also assume that Oklahoma State received the remaining 17 second place votes.  In order to arrive at their 1367 total, they could not have received all of the remaining third place votes, and instead received votes at #4 or lower on some ballots.  The best case would be that they received 35 third place votes and 7 fourth place votes.

Fishy point 1 - Some coaches, as many as 7 of them, voted Oklahoma State #4 or lower on their ballots.

In the Harris poll, using the same logic we see that Alabama would have received 78 second place votes and 37 third place votes.  Oklahoma State would have then received 37 second place votes and best case 50 third place votes and 28(!) fourth place votes.

Fishy point 2 - Some Harris poll voters, as many as 28 of them, voted Oklahoma State #4 or lower on their ballots.

So what would it have taken for the polls to move Oklahoma State close enough in the polls to move ahead of Alabama in the BCS?

There are a number of combinations, but lets look first at what would happen if Oklahoma State had garnered all of the second place votes from the Coaches.  Here is what the totals would look like.

Oklahoma State0.938213740.931526540.92310.960

The margin has now dropped to just 0.0003!  If just 3 of the Harris voters that had Oklahoma State fourth on their ballots moved them to third, Oklahoma State would be in the lead.  In this scenario, one computer, and 10 voters is all the difference was.

If we go from the other end, leaving the 7 fourth place votes from the coaches and instead change 17 of the 28 Harris poll fourth place votes to thirds, Oklahoma State would be in the lead.  This scenario is one computer and 17 voters.

If instead of moving fourth place to third place votes, we look at a few voters swapping Oklahoma State to #2 and Alabama to #3.  If just 5 coaches that votes Alabama #2 and Oklahoma State #3 swapped their votes, Oklahoma State would be in the lead!

So in the end, Oklahoma State was 1 computer and 5 voters away from being #2.  It doesn't get much closer than that.

Now, we can debate whether the two computers were correct and we won't get far as the algorithms are not public (except for Colley's which had Oklahoma State #2), but do we really trust computers that have little to no transparency into how they work or what goes into their inner workings?  We should get to see how all the Harris and Coaches poll voters voted soon and it will be interesting to see what if any injustices there were in the voting.  Hopefully, the best case I noted above is the case, it would be terrible to see any voter having Oklahoma State any lower than #4 on their ballot being the cause for Oklahoma State not being in the BCS title game.