Saurian Sagacity Power Rankings
With the end of week 8 we are ready to introduce our first Power Rankings of 2008.
Once again we are utilizing a relatively bias free method that has been just slightly tweaked for 2008. The past two year’s very similar methods resulted in both Florida and LSU as the eventual number ones, and was very useful for betting purposes all season (void where prohibited).
For this season’s version we have the following formula –
((SD Ratio *1.1) + (SO Ratio *0.9) + W/L Record)) * SOS = Power Value
Explained further –
SD Ratio is a ratio of each team’s scoring defense to the best scoring defense. For example USC has the top scoring defense in the nation at 7.8 ppg. New to this year we apply a multiplier of 1.1 based on my work that found a high correlation of scoring defenses to success. So USC, with the best scoring defense gets a 1.1 in this category (1*1.1).
SO Ratio is a ratio of each team’s scoring offense to the best scoring offense. Using the USC example, their offense is 8th overall at 41.5 ppg. The best offense is Tulsa’s at 56.7 ppg. Thus USC’s ratio is .7336 of the best. Then we apply a 0.9 multiplier reflecting the relatively lesser importance of offense to defense, and get 0.66024.
W/L record is a simple calculation of % record, in USC’s case 0.83333 (5-1).
We add the three components together, in our example totaling 2.5936. Then the number is multiplied by the strength of schedule per the NCAA, which in USC’s case is a ratio of 0.692308 (40th nationally having played teams that are 18-15). The end result is 1.795553.
Our Top 25 (with Power Values) –
1 Oklahoma 1.81694
2 Southern California 1.79555
3 Texas 1.73109
4 Missouri 1.63397
5 Georgia 1.63333
6 Boise St. 1.53618
7 Ohio St. 1.53295
8 TCU 1.39306
9 Florida 1.33640
10 Georgia Tech 1.26283
11 Iowa 1.26052
12 Alabama 1.26000
13 Florida St. 1.24606
14 Oklahoma St. 1.23535
15 Penn St. 1.19556
16 Nebraska 1.18725
17 Kansas 1.18296
18 Minnesota 1.17105
19 Virginia Tech 1.16828
20 Kentucky 1.16761
21 North Carolina 1.16691
22 Texas Tech 1.16202
23 Michigan St. 1.15407
24 Utah 1.15174
25 Troy 1.15081
A couple of comments –
First the strength of schedule we are using is an incomplete picture at this point in time. A perfect example is Boise State, 48th nationally, with the teams they have played having a 14-13 record. The teams left on Boise’s schedule are presently 14-23, and Boise’s SOS will end up about 90th. So patience on that part.
Iowa? Freakin’ Iowa? Yeah, that shocked me too. Until I looked at their stats. Their 3 losses were by a combined 9 points to Pitt (by 1), Northwestern (by 5), and Michigan State (by 3). They have given up no more than 22 points in a game this year, and rank 4th nationally in scoring defense. Their current SOS is 27th.
Anyway, food for thought. This is an unfinished product, and will look quite a bit better after a couple of more games.
And if you are interested, other notables include –
31. USF
34. BYU
37. LSU
38. Cal
45. Miami
48. Northwestern
49. Tulsa
53. Wisconsin
57. Notre Dame
69. Auburn
73. Tennessee
75. Oregon
86. Michigan
119. North Texas
UPDATE: Maybe I'm not so crazy after all, at least when you consider this system to what Jeff Sagarin is using (or maybe that just makes me crazier).
Sagarin has Iowa ranked 8th overall per his "predictor" model. Of course, he has Florida number 2...
Mentioned in Austin Murphy's book 
7 comments:
Note:
Mathematically and grammatically, a ratio exists between two things, i.e. the ratio of ____ to ____. Thus, the sentence "SD Ratio is a ratio of each team's scoring defense" is incomplete.
We need it to say "SD Ratio is the ratio of each team's scoring defense to the nation's best scoring defense" or something of the like.
Ditto of SO Ratio.
Also, in the SO Ratio explanation it says we apply a 0.0 multiplier, which presumably is a typo, unless offense is of severely less importance than defense.
--------------------------------
I do have one substantive comment, also.
I feel like using a blanket strength of schedule multiplier has the possibility of muddling some of the data. Presumably it would be better to be able to break the strength of schedule multiplier into parts, similar to the tier system you use in your BlogPoll ballot.
I mention this because it's clearly more telling to be able to examine a team's performance against top-tier competition.
For example, objectively speaking we can see that Iowa has benefited from playing an above-average schedule but not facing an elite team. Meanwhile, Oklahoma State is below them because even though they've beaten a (supposedly) elite team they've also toyed with a bunch of nobodies.
[One could probably pick apart those specific examples, but it's hard to argue with the idea I'm getting at]
I think it would be nice to have some way of weighting wins over elite teams (and, similarly, losses to bad teams) more heavily. That way we can somehow give more credit where credit is due.
First comment - true as stated.I was typing mighty fast. The multiplier is 0.9. Sorry for the obvious errors.
Second comment - I agree, but the whole purpose of my Power Rankings is to remove as much bias as possible (I've taken a look at Colley's SOS). The only purely non-biased SOS I have found is the NCAA's, which takes only your opponent's W/L record. Perfect? Hardly, but free from the mystery that comes from SOS ratings like Sagarin's.
Wow, finally a ranking where USC looks good. I didn't think that was possible. LOL
Iowa, much as I hate to admit it since they're a team full o' thugs, is solid and mark my words, they will give PSU all they can handle.
Upon seeing Iowa---IOWA???---lurking just outside your TOP 10, my first thought was that you've FINALLY drank one too many of those exotic beers, and it destroyed the last of your remaining brain cells.
Then I thought I would expound upon how I thought your system undervalued the importance of a team's actual record; i.e. this is NOT 'fantasy football:' when it's all said and done, the ONLY stat that has ANY real meaning is...W's. As we all-too-often found during the Zook Error, close losses are STILL losses...no matter HOW much you may learn from them.
[rolleyes]
[pound head on table repeatedly]
BUT...
on further review...
I guess I need to consider the MEANING of "power rankings." If you would use them like the BcS polls, I would say at this point your system is flawed at best.
But if you were to use them for "informational purposes only" to, say, decide which side to select with your "short-term investment broker"--where actually WINNING the game is merely ancillary...your system seems to have some potential.
yankeegator
I like it.
As always, it's a different and interesting way of looking at things.
I'm also certain much will change as teams get through the meat of their schedules. Keep up the good work!
Yankee -
No, it's not for merit purposes at all, nor to decide who should play for the BCS. It is merely a tool for deciding who is "better" than who on a weekly basis - "power" if you will.
As you put it, it is all about the "short-term investment broker" decision making.
In other words, when making such an investment decision, Iowa might not be an all together bad pick.
Iowa blows. So what if they always lose close games? If they were any good, they would win some of them. Plus, let's not overstate Michigan State and Northwestern as quality opponents. They are also weak.....despite their records.
Post a Comment