Monday, November 05, 2007

Saurian Sagacity Week 10 Power Rankings

The more I look at this power ranking system, the more I like it. For our methodology see here.

Essentially, we create ratios that equally weight total offense and total defense (weighted by strength of schedule), and win-loss records. A perfect score would be a “3”, which would indicate the best offense and best defense against the toughest schedule, and an undefeated record.

For informational purposes, the top offense right now is Texas Tech (539.5 ypg), the top defense Ohio State (221.2 ypg), and the toughest schedule is NC State (past opponents records 43-20).

Our top 25, with total score (out of a possible “3”), are –

1 LSU 2.4661
2 Oregon 2.3408
3 Ohio St. 2.3034
4 Kansas 2.1009
5 Oklahoma 2.1001
6 South Fla. 2.0846
7 Missouri 2.0495
8 Boston College 2.0399
9 Florida 2.0171
10 Arizona St. 1.9890
11 Boise St. 1.9788
12 West Virginia 1.9731
13 Georgia 1.9363
14 Alabama 1.9247
15 Penn St. 1.9236
16 Virginia Tech 1.9108
17 Michigan 1.9043
18 Clemson 1.8988
19 Texas 1.8825
20 Kentucky 1.8804
21 Rutgers 1.8763
22 Connecticut 1.8727
23 Illinois 1.8717
24 Oregon St. 1.8526
25 Auburn 1.8320

I’m sure there is some “How the hell can USF be 6th?” talk when you look at this, but realize USF –

1. Has the nation’s 2nd hardest schedule (opponent’s records 39-19)
2. Has the 18th best total defense

I think the above represents a pretty good idea of the relative strength of the teams based on what we have seen so far.

Update: Something to note about the rankings here is how they rate relative to the others. Number one LSU with a 2.4461 is 17.3% better than USF at 6th with 2.0846, no small difference.

I have been working on a "predictor" model using these rankings, and have something in the early stages I don't want to reveal quite yet. It has been reasonably predictive in game picking last week.


Anonymous said...

Fattening up on second-tier teams that play other weak teams doesn't make "the nation's 2nd toughest schedule" --

Elon (6-3)
Central Fla. (6-3)
Fla. Alt. (4-4)

Having been suckered in to watching them recently, I'm puzzled as to how they beat West Va. (7-1) or Auburn (7-3), other than those teams figured USF was a "bye" week.

Nick said...

Which is exactly why SoS should be measured itteratively. I suggest you use either A&H or Colley's SOS rankings.

Henry Gomez said...

USF has the 2nd hardest schedule according to NCAA, 28th according to Anderson and Hester, 9th according to Sagarin, and 10th according to ColleyMatrix. Who's to say who is right? Why should we take one over the other. No, the NCAA doesn't look at the records of the opponents of the opponents but I think before we stick a variable into our equation we'd like to know exactly what goes into it.

Mergz said...

I use the NCAA SOS for a reason - it is the only unbiased one, based strictly on opponent's records. The others use various opinion based weightings.

The whole point of our Power Rankings is that is has no biased input - it is simply a team's total offense and total defense weighted by SOS, and their win-loss record. I don't necessarily agree with any single ranking, but they are what they are.

Nick said...

The NCAA ranking is unbiased? Are you sure?

So the 2/3 and 1/3 weightings make perfect sense to you? If so, explain where they come from, and why they make sense.

Henry Gomez said...

Unbiased, meaning that the formula is not a secret and it's not arbitrary. Trying to eliminate bias is possible, completely eliminating it is impossible. At some point you have to decide how you are going to weight the variables. What's your model Nick? Would you use a SoS calculation created by a third party that doesn't explain how it's calculated?

Mergz said...

I wrote about my methodology elsewhere, but essentially what I was seeking was unbiased INPUTS. As Henry points out, no system is totally without bias.

The NCAA's SOS simply ranks teams by their opponents W/L records. Perfect? Perhaps not, but it is at least better than someone deciding what teams have tougher schedules (like Sagarin last year putting essentially every PAC 10 team in his top 20).

The other 3 inputs are total offense and defense (unbiased - it is what it is) and W/L records of the team itself(ditto).

Now my decision part comes in equally weigting all 3 at 1/3 each. Many times the W/L record does not reflect what actually happened on the field (or ND would be 0-8). Is equal weighting them the "best"? Hell I dunno, but until I hear something better, it is what I am using.

And, as I have said, it is transparent - you know what I use. You may not like the outcome, but unlike Sagarin, you know how I got it.