Saturday, August 12, 2006

BCS Conclusions

Versions / 1998 / 1999 / 2000 / 2001 / 2002 / 2003 / 2004 / 2005 / 2006 / 2007 / Conclusions

So what have we learned? Let's take a look at a compilation of who would've been in the National Championship game each year according to each version. (Tan teams are official, dark tan teams are projections that I'm 100% certain of mathematically, and the green teams with the * are projections I'm confident calling but not certain of mathematically.)

Florida StateFlorida StateFlorida StateFlorida StateFlorida State
1999Florida StateFlorida StateFlorida StateFlorida StateFlorida State
Virginia TechVirginia TechVirginia Tech*Virginia TechVirginia Tech
Florida StateFlorida StateMiami (FL)*Miami (FL)Florida State*
2001Miami (FL)Miami (FL)Miami (FL)Miami (FL)Miami (FL)
2002Miami (FL)Miami (FL)Miami (FL)Miami (FL)Miami (FL)
Ohio StateOhio StateOhio StateOhio StateOhio State
2006Ohio StateOhio StateOhio StateOhio StateOhio State
2007Ohio StateOhio StateLSULSUOhio State
? ? ?LSUOhio StateOhio StateLSU

Trying to arrange a ranking of the different versions based on which one "worked best" is difficult mainly because there's no consensus as to which two teams should've been in the National Championship game some years. It'll be made a bit easier because there's 7 years which were consensus, meaning every version matched up the same two teams. (I'm including 2007 in this count since it has the same 2 teams in 4 rankings, while the 5th ranking is unknown.) So we only have to look at 2000, 2001, & 2003. We'll give it a try.

In 2000, everybody thought that the Miami should be #2 over Florida State since they defeated them on the field. Miami had the edge in the polls, but Florida State had the edge in the computers. This one basically comes down to how much you believe in the computers, making it a very personal, perspective-driven year to look at. Add to that the fact that the versions themselves were divided, and 2000 doesn't look like it'll be much help with our task. So we go to 2001 & 2003.

(Wait a minute - we're going to judge the effectiveness of all five versions of the BCS by only looking at 2 out of 10 years? Yes, we are. But they're two really good years, if that helps you.)

In 2001, Versions A, B, & C all had Oklahoma playing Nebraska. Most people thought this was an outrage, seeing as how Nebraska didn't even make it to their title game. If somebody other than Nebraska would have gone, the only people complaining about the system probably would have been from Nebraska (and Colorado or Oregon, depending on who was #3 and left out as well).

So Versions A, B, & C aren't as effective here. Version D took Colorado, while Version E took Oregon. Which of those two did better? I'd have to go with Version E for two reasons. First, Oregon smashed Colorado in the Fiesta Bowl that year. (Yes, we're employing hindsight - it can be a good thing.) Second, both polls had Oregon at #2. If we're trying to avoid controversy, simply confirming what most of the voters thought seems the best way to do that. So Version E has the lead in the effectiveness department, just ahead of Version D and quite ahead of A, B, & C.

In 2003, Oklahoma got blown out in their last game, didn't win their conference, but were still kept in the top 2 by four versions. (Version A even kept them at #1.) Version E matched up USC and LSU, the game many people wanted to see, especially after they won their BCS bowl games and split the title.

So it seems that of all the versions, the latest, Version E, would have produced the least-controversial matchups had it been in effect all 9 years. The next best would be Version D, which had Colorado meeting Miami in 2001 but kept Oklahoma in 2003. Versions A, B, and C fill out the ticket at the end - they all had versions of Miami vs Nebraska in 2001 and kept Oklahoma in the top 2 in 2003.

So the major conclusion is that overall, the later changes that the committee made to the BCS formula have improved it's effectiveness and would have eliminated a great deal of controversy if one of the later versions had been in place from the beginning. The earlier changes, from Version A to B to C had little effect on what the overall outcomes and pairings would have been.

It's rather ironic that the most effective version of the BCS (E) is the one that most approximates the system of polls that the BCS was originally designed to replace, making you wonder whether it was all worth it. In addition, this is the 4th year of the BCS using Version E (or E-2 - they're basically the same), which makes it the longest running version by far. I don't think we're going to see any more versions either. The BCS has reached the point where simply changing how the rankings are calculated isn't really going to solve any of the issues that people seem to argue over. The next logical change, if there's going to be one, will happen after the current BCS cycle, sometime after the 2009-2010 season... Get ready for the BCS 2.0!

Thanks for reading - please take the time to leave a quick note with a comment, critique, or correction.

- Ed

2006 < Top > Versions home

No comments: