Wednesday, December 4, 2013

Is the BCS Really Wrong?

I lived in Ohio for a long while.  While there I made friends who I will keep for the rest of my life.  Some of these friends are Ohio State fans who are unhappy that OSU is not (as of this moment) slated to play in the National Championship Game even though they have the longest winning streak in college football.

Being an inquisitive college football fan, I thought about this stuff.  The complexities of the current state of college football ranking is difficult to explain as evidenced by my futile attempts to explain the Coaches Poll, the AP Poll and the BCS formula to my wife.  In the end she asked the question that every person not sponsoring a Bowl game or profiting from a Bowl game has asked: why isn't there just a playoff?

Because.  That's why.  And the "Plus 1" that the BCS will move to next year is kind of like putting a band-aid on an infection and claiming that it's cured.  So, regardless of whether or not D1 college football will do what every major sport on the face of the planet has done and have a playoff, the question of ranking bias will exist albeit (with a 16 team playoff) an easier to stomach question of who should be 16 vs. 17 as opposed to who should be 2 vs 3.

So I did what any self respecting inquisitive person would do: I thought about it for a few days and then gave up.  The numbers and complexities are too large.  When is a win a legitimate win vs. a win against a "lower quality" opponent?  What would an algorithm to decide rank look like?  Then I found a blog and a partially implemented solution in C# and decided to finish it and fix it.

This method is not without its flaws.  I eliminated every team that hadn't played at least 9 games against teams in D1 football.  That means that the loss that Florida had against Georgia Southern didn't hurt Florida. The loss to Georgia Southern was not going to be the only loss they had (clearly). That also means that Alabama's win against Georgia State didn't count.  I think this is a fair trade-off as it is kind to bad teams and harsh to good teams.

This has room for improvement.  I still am counting games against conferences with absolutely terrible average ratings.  Here are the average ratings by conference (on a scale of 1-100 with 100 being the best)

1. SEC: 70
2. Pac 12: 67
3. Big 10: 61
4. Big 12: 58
4. ACC: 58
6. AAC: 41
7. Sun Belt: 37
7. Independents: 37
9. Mountain West: 36
10. MAC: 33
11. C-USA: 31

The algorithm is as follows which is effectively how Google stack ranks pages:

1) Every team starts with a rank of 100
2) Iterate over every team.
3) Add the rank of every team they beat together
4) Divide by the number of games they played
5) Go back to step 2 and iterate 25 times

The standings are as follows (and I'm sorry Ohio State, the math and the worthless wins against Illinois (33) and California (23) don't lie):

1. Auburn (100)
2. Missouri (97)
3. Alabama (96)
3. Arizona State (96)
5. Ohio State (95)
6. Florida State (93)
7. Stanford (91)
7. South Carolina (91)
9. Oklahoma State (89)
10. Baylor (88)
11. Oregon (86)
12. Michigan State (85)
13. Clemson (84)
14. UCLA (81)
15. Georgia (80)
16. Oklahoma (79)
17. LSU (78)
17. UCF (78)
19. Duke (77)
20. Northen Illinois (76)
21. Wisconsin (75)
21. USC (75)
21. Texas A&M (75)
21. Miami (FL) (75)
25. Washington (74)

So even if the SEC bias exists, it's correct.  Top to bottom they're strong and they beat their non-conference foes more (at least this year).  The one really surprising thing to me is the Sun Devils who are tied for 3rd despite their two losses because they have beaten UCLA, Wisconsin, USC, and Washington. This may or may not be a flaw in the system and is a better comparison of conference strength as opposed to bowl records.  A bowl game is a non-conference (ideally) game played in January.

OSU fans can take comfort in the fact that if I give them a win next week against MSU, their rank jumps to 98.  Either Auburn or Missouri will lose and, surprisingly, the only game that will hurt OSU in this formula is the Stanford v. ASU game but that game will not matter from a BCS standpoint.

I could fix this by tweaking the formula which is the mess the BCS formula is in.  There is no system, computer, human, or otherwise that can replace a good old fashioned tournament.

Even my wife agrees.