Diving Into the BCS Results with Descriptive Statistics
The calendars have just turned to December, so you know what time of the year it is. No, not the holiday season! It's time for people to complain about the Bowl Championship Series (BCS)!
This year the main complaint is that Alabama, not Oklahoma State, gets to play Louisiana State (LSU) in the title game. Some voters in the Harris Poll don't even have Oklahoma State in the top 3! This has led some to call out certain voters in the Harris Poll. But before we go crying "conspiracy," it's best to do some data analysis. So I'm going to use Minitab Statistical Software to give us some descriptive statistics and break down the votes in the Harris Poll.
First things first. There are 115 voters, and everybody voted LSU #1. They were the only undefeated team this year, so there is no conspiracy there.
But what about #2? Let's see what proportion of the voters put Alabama and Oklahoma State at #2.
Out of the 115 voters in the Harris Poll, 80 of them ranked Alabama second. We see that the 35 voters that didn't have Alabama at #2 ranked Oklahoma State there instead. So Alabama and Oklahoma State were the only teams to be ranked #2. Let's keep digging!
If a voter had Alabama at #2, it should be because they feel they are the better team. But they probably also feel that the conference they play in is better. Alabama plays in the Southeastern Conference (SEC). So you would expect voters that had Alabama #2 to also vote the other SEC teams higher. Let's see how voters ranked them.
Above is the average ranking for Arkansas, South Carolina, and Georgia broken down by where the voter ranked Alabama. The first row tells us that voters who didn't rank Alabama #2 on average ranked Arkansas 7.657. Voters who had Alabama second ranked Arkansas higher, at 6.987. The same is true of South Carolina and Georgia. So voters who had Alabama #2 did rank the SEC teams higher than voters that had Oklahoma State at #2.
What about the converse? Did voters who had Oklahoma State at #2 rank Big 12 teams higher?
We see the same thing here. Voters who had Oklahoma State #2 also voted other Big 12 teams higher. Especially Kansas State. So no matter which team the voters put at #2, they seemed to stay consistent with the rest of their ballot.
But what about voters who didn't even have Oklahoma State in their top 3. Can we find anything out about them? First, let's see how many there were.
So there were 16 people that didn't have Oklahoma State in their top 3. Did these 16 people have an agenda? Well, 13 of them ranked Stanford at #3. Stanford has the same record as Oklahoma State, and their only loss came to #5 Oregon. It doesn't seem like a stretch to think that Stanford is a better team than Oklahoma State.
And maybe those 16 people just don't think very much of the Big 12. Let's see how they voted other Big 12 teams.
We can see that the voters who left Oklahoma State out of the top 3 were also unkind to the other Big 12 teams. Kansas State and Baylor (Oklahoma State's 2 best wins) were ranked over 2 spots lower by voters who left Oklahoma State out of the top 3. However, it is odd that they had Oklahoma ranked higher. This shows some inconsistency in their ballots, but not enough to cry conspiracy.
So after breaking down the numbers, it doesn't appear like there was any conspiracy in the Harris Poll. If anything, Oklahoma State won a bunch of voters over in the final poll because they finished ranked 3rd. In the previous week's Harris Poll, they were ranked 5th!
What's more likely is that there were 5 different teams that finished the season with 1 loss, and how do you choose which one is the best? None of them played each other and the only common opponents between any of the teams were Tulsa and North Texas. With 115 different voters, you're bound to get some vastly different opinions!
And speaking of opinions, this statistician finds it ironic that the sport that demands only two teams be selected to compete for the championship is also the sport that has the smallest sample of games from which to judge who the best teams are. Each team only plays 12 or 13 games, and in half of those they play an opponent that is over-matched! That gives people very little data to make a decision with. Of course you're going to have disagreements! Someday they'll make changes to college football. But until then, the data seem to say don't hate the voters, hate the system!
Photograph by RonAlmog. Images licensed under Creative Commons Attribution ShareAlike 2.0.