Monday, November 02, 2009

Initial Power Rankings

With each team having 8 games played, it is time for our inaugural Power Rankings of 2009.

Here is the formula I will be using –

(SD Ratio + SO Ratio+ 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 Florida has the 2nd Scoring Defense in the nation at 11.0 ppg. This is made into a ratio that compares it to the top Scoring Defense, held by Penn State. The ratio in this case is 0.848

SO Ratio is a ratio of each team’s scoring offense to the best scoring offense. Using the Florida example, our offense is 6th overall at 36.0 ppg. The best offense is Texas at 41.75 ppg. Thus, Florida’s ratio is 0.862 of the best.

W/L record is a simple calculation of % record, in Florida’s case 1.000 (8-0).

We add the three components together, in our example totaling 2.710457. Then the number is multiplied by the strength of schedule per the NCAA, which in Florida’s case is a ratio of 0.849781 (7th in the nation right now). The end result is 2.3033.

It has, in my opinion, done an excellent job in the past of determining the relative strength of the teams. Coincidently, the number one team in each of our past three Final Power Rankings has been the BCS Champion for that year. This has occurred despite the unbiased nature of our ranking system.

Our intial top 25, and their scores -

1 Florida 2.3033
2 Texas 2.0085
3 Penn St. 1.8888
4 Oregon 1.8489
5 Virginia Tech 1.8433
6 TCU 1.8416
7 Alabama 1.8233
8 Oklahoma 1.8063
9 Iowa 1.7644
10 Cincinnati 1.7511
11 Clemson 1.6563
12 Wisconsin 1.5854
13 Arizona 1.5765
14 Ohio St. 1.5496
15 South Fla. 1.5472
16 Oklahoma St. 1.5290
17 Southern California 1.5056
18 LSU 1.4974
19 Miami (FL) 1.4973
20 Pittsburgh 1.4919
21 Boise St. 1.4814
22 Houston 1.4489
23 Georgia Tech 1.4441
24 Texas Tech 1.4098
25 Tennessee 1.4018

UPDATE: I put up this post in haste this afternoon without really examining the results. I would say, having the chance to do so, it looks pretty good to me. For the most part it passes "the eye test".

VT and Oklahoma may seem pretty high, but their SOS has been tough (VT's is nation's best). Alabama is a little lower than I might have expected, but their SOS is 43rd (but sure to improve when they face LSU). Likewise LSU at 18th, with an SOS of 58th.

How about Clemson at 11th? Their 6th overall SOS helps get them there despite 3 losses.

And this system is pretty definitive between who wins the TCU-Boise State battle.

8 comments:

Dan Shanoff said...

Wouldn't it be nice if this simple-yet-effective formula was one of the formulae used in the BCS computer rankings? (Never understood why they went with those six, with no thought of adding/removing/changing it based on new info, etc.)

- Shanoff

Anonymous said...

Well the computer rankings would get better by simply letting them take into account whatever they want to take into account (for instance MoV). Also getting rid of the Billingsley Report would help everything.

Mergz, what happened to favoring defense over offense? I think you used 1.1 * SD and .9 * SO for a little while.

I'm also kinda surprised to see Virginia Tech and Oklahoma ranked so high.

Jams said...

re: Dan's comment... Does taking scoring offense and scoring defense into account separately qualify as using margin of victory? I mean there's only a semantic difference between using average margin of victory itself and calculating average points scored and average points allowed and using them separately.

And anyway, what are the BCS "computer" formulas even allowed to take into account besides wins and losses? Anything?

Regardless, simple but powerful formulas like this one should be favored over convoluted ones that are built and massaged to reinforce human polls anyway.

Mergz said...

Amos,

I re-thought about using the modifiers for offense and defense and decided it was too arbitrary on my part, despite having some statitical basis. Plus it didn't change much (top 10 was the same). So I just dropped it.

Jams,

I would say using SO and SD does take MOV into account, and I don't really have a problem with that.

I think MOV is a red herring. All the top teams run up scores on lesser quality teams, and the human pollsters take it into account. In some cases it is probably valid anyway (Oregon over USC, for instance).

Anonymous said...

Jams, I'm pretty sure some of them take into account home/away/neutral, also when the game was played, and I know Billingsley has bonus points for holding an opponent to under 10, I'm not quite sure why this is allowed. Colley only takes W/L into account (and did so even before MoV was disallowed), with a very clean system (simple concept if you are comfortable with mathematics, he's the only one who completely lays bare his algorithms).

Unknown said...

Margin of victory is _not_ a red herring if you limit the margin to 10 points. Anything more gets you nothing, anything less means you struggled.

No coach (except ones like [NAME REDACTED]) is comfortable with a five-point lead so MOV of less than 10 means they could not put the game away.

Trader Rick said...

cHRIS fOWLER SAYS THERE'S NO CLEAR-CUT NUMBER ONE.

Robert said...

I know it will all shake out in the end, but it's a little hard to rank three loss teams--that means they have lost almost 40% of their games among the top 11 teams in the nation. I'm all for rewarding playing tougher schedules but... I'd like to see a bigger penalty for multiple losses built in so that each loss past the first counts off extra (or something).