My system is "retrodictive" vs "predictive." By that, I mean it measures performance up to this point in time, but is not meant to claim one team will beat or lose to another based on the rankings. There are apparently some hefty mathematical reasons why retrodictive systems are not good at predicting, just as predictive systems are not so good at measuring past performance. I don't know the math, but I've read enough lately to know that smarter people than me have proved it.
So I'm contemplating a large change to my system by adding a predictive component. It will probably be the SRS from here:
I see no need to reinvent the wheel.
I don't see what I'm doing as trying to predict the future, but I can see that there is a desire for that type of ranking system. The thing is, FBS will never have an objective system until it goes to a college hoops style system with automatic bids and at large berths determined mainly by a computer. There are more than a few statistical and mathematical treatments out there that can handle that side.
I want to see who is truly deserving of that consideration, and that means measuring past performance. Very much "What have you done for me lately?" and less "Who would win between these two teams?" This is in line with the "body of work" argument you hear on the TV shows all the time: Who have you played? Did you beat the teams you were supposed to beat? Are the teams you beat any good? How about the teams they beat?
Keeping all of that in mind, here is my new top 25 without the predictive component:
25. Kansas State - sorta backing in, carried by 2nd level points
24. Washington - definitely backing on after that disappointing loss to Colorado
23. Va Tech - 8-3 with a win over a 7-4 Pitt
22. Boise St - convincing win, SoS hurts them
21. Miami (FL) - lost to FIU yet still here; I cannot explain that other than SoS
20. Auburn - just got hurt by playing an FCS team. In November. #byeweek
19. Louisville - top 10 Sos
18. Alabama - OW% 0.473, played Western Carolina (-1 1st level pt) #byeweek
17. OK St - 2-2 vs ranked, 4 straight wins
16. Florida - another tough SoS
15. Iowa - beat a very not bad Illinois
14. Minn - bounced back from the Iowa loss
13. USC - 8-4, pretty convincing win over UCLA, better than their record
12. Utah - well, you gotta play good teams to acquire those 1st & 2nd level pts
11. Wisconsin - beat Purdue, yawn
10. UM - resurgent offense? Beat a not awful Indiana
9. Baylor - waiting for their chance at revenge
8. ND - sneaky good, solid SoS
7. Oregon - got completely outplayed, almost zero chance at the playoff
6. OU - got a little help from the refs at the end, allowed TCU to come back
5. PSU - score looked better than they played, turnovers helped
4. LSU - it was a conference game at least
3. Clemson - still amazing to me with their soft SoS
2. UGA - A&M isn't great, but that Dawg D is something special
1. tOSU - win 35 or 42 to 0 without the turnovers
It really makes very little difference if I actively penalize teams for playing FCS schools or just zero them out. I am constantly surprised by some of the results my spreadsheet spits out.
Here it is with the SRS predictive ratings added.
Any modification I make to the entire list does almost nothing for Bama or Auburn. The only way to make the Tide look good is to completely ignore their lack of competent opponents. This is not Bama's fault, as schedules are set years in advance and you can't predict the strength of a program that far out. But it still needs to be acknowledged: Alabama has played far worse teams than any of the top 10 teams. Their lone exception is LSU, who beat them and hung 46 points and 600 yards on them. Keep in mind that Utah is suffering from the same issue (except they lost to USC).
I'm still playing with my system, so there will be changes ongoing.
No comments:
Post a Comment