(UPDATE: Re-rank based on new, additional, criteria)

So last week I chopped up historical college football win-loss data back to the start of Solich's OHIO career.

This week I decided to work with the data a little more and develop an objective ranking system of FBS teams.

Award Criteria:

- Fraction of 1 Point (Pt) Equivalent to Team Win Percentage (less than or equal to 1 Pt total)

- 1 Pt awarded for each win vs a team with 0.500 Record

- 1 Pt awarded for each win vs a team with a winning record + Incremental Point Fraction Bonus (IPFB)

-IPFB can add up to 1 pt per win over winning record teams with win(s) over other winning teams. This is a relative point fraction bonus distributed over the complete range of wins vs winning teams. For example, a number of teams such as USC, Boston College, and Arizona have wins against a team that has three (3) unique wins against other winning teams. USC, BC, Arizona, et al, lead the FBS with this achievement and currently receive a full 1 Pt Incremental Point Fraction Bonus. Florida has only 1 win over a team who also beat 1 team with a winning record. Florida currently receives 0.33 IPFB for that win. This is a variable incremental bonus that will remain a function of results as the season progresses.

(UPDATE: Additional Criteria)

-"Strong Loss Bonus": 1 Pt awarded total to teams with zero losses. 1 Pt awarded for teams that only have losses to Top 16 teams by one possession. 1/2 Pt awarded for teams that only have losses to Top 32 teams by one possession. Only one SLB can be recieved. Therefore teams either have 1 Pt, 1/2 Pt, or 0 SLB. (SLBs are based on rankings before SLBs are awarded. Teams are then re-ranked)

-Scores are now ranked "per attempt" (or per game played)

-Scores are referred to as Objective Points Per Attempt++ (OPPA++). The two pluses represent the IPFB and SLB bonuses

I would be happy to discuss my methodology further with anyone who expresses interest. I'm certainly open to criticism. I really just did this for fun, no ulterior motive here. I personally think Jeff Sagarin produces rankings with logically indefensible biases in favor traditional power conferences. Similar biases exist in the other computer ranking systems I'm familiar with so I decided to make my own.

In only the most limited sense do I expect these rankings to serve as a predictor of future games. I don't believe predicting outcomes is a fruitful task (kind of like working on a time machine). There's quite a lot of football left to play.

I expect some will look at these mid-season ranking and raise an eyebrow at Colorado State at #11 or Alabama at #17. I believe this order is defensible by the criteria defined above. Every team is evaluated by the same blind standard.

(Take at notice how Saragin and the other computer rankings go out of their way to weigh down non-power conference teams and prop up power conference teams).

RNK --- OPPA++ --- SAGARIN ---- TEAM

1 ------ 1.500 ------- 4 ------- Mississippi State (6-0)
2 ------ 1.333 ------- 1 ------- Mississippi (7-0)
3 ------ 1.083 ------- 28 ------- Arizona (5-1)
4 ------ 1.069 ------- 3 ------- Auburn (5-1)
5 ------ 1.027 ------- 9 ------- Oregon (6-1)
6 ------ 1.000 ------- 10 ------- Florida State (7-0)
7 ------ 0.900 ------- 20 ------- UCLA (5-2)
8 ------ 0.888 ------- 19 ------- Southern California (5-2)
9 ------ 0.861 ------- 14 ------- Kansas State (5-1)
10 ------ 0.861 ------- 5 ------- Georgia (6-1)
11 ------ 0.857 ------- 31 ------- Marshall (7-0)
12 ------ 0.816 ------- 26 ------- Missouri (5-2)
13 ------ 0.806 ------- 7 ------- TCU (5-1)
13 ------ 0.806 ------- 22 ------- Utah (5-1)
15 ------ 0.789 ------- 2 ------- Alabama (6-1)
15 ------ 0.789 ------- 15 ------- Notre Dame (6-1)
17 ------ 0.778 ------- 48 ------- Penn State (4-2)
18 ------ 0.777 ------- 56 ------- Colorado State (6-1)
19 ------ 0.771 ------- 13 ------- Ohio State (5-1)
20 ------ 0.769 ------- 43 ------- Boise State (5-2)
21 ------ 0.739 ------- 17 ------- Clemson (5-2)
22 ------ 0.736 ------- 23 ------- Arizona State (5-1)
23 ------ 0.682 ------- 45 ------- Minnesota (6-1)
24 ------ 0.653 ------- 38 ------- Miami (FL) (4-3)
24 ------ 0.653 ------- 59 ------- Nevada (4-3)
26 ------ 0.653 ------- 50 ------- UCF (4-2)
27 ------ 0.650 ------- 6 ------- Oklahoma (5-2)
28 ------ 0.626 ------- 51 ------- Kentucky (5-2)
29 ------ 0.625 ------- 12 ------- LSU (6-2)
30 ------ 0.616 ------- 36 ------- Duke (6-1)
31 ------ 0.605 ------- 35 ------- Virginia Tech (4-3)
32 ------ 0.582 ------- 54 ------- Virginia (4-3)
33 ------ 0.578 ------- 52 ------- Georgia Tech (5-2)
34 ------ 0.558 ------- 53 ------- Pittsburgh (4-3)
35 ------ 0.556 ------- 57 ------- Arkansas State (4-2)
36 ------ 0.549 ------- 39 ------- East Carolina (5-1)
37 ------ 0.537 ------- 61 ------- Northwestern (3-4)
38 ------ 0.531 ------- 29 ------- Louisville (6-2)
39 ------ 0.528 ------- 55 ------- Boston College (4-3)
40 ------ 0.510 ------- 44 ------- BYU (4-3)
40 ------ 0.510 ------- 33 ------- South Carolina (4-3)
40 ------ 0.510 ------- 60 ------- Utah State (4-3)
43 ------ 0.503 ------- 8 ------- Baylor (6-1)
43 ------ 0.503 ------- 18 ------- Nebraska (6-1)
45 ------ 0.491 ------- 11 ------- Michigan State (6-1)
46 ------ 0.490 ------- 30 ------- Tennessee (3-4)
47 ------ 0.483 ------- 24 ------- West Virginia (5-2)
48 ------ 0.472 ------- 49 ------- Memphis (3-3)
49 ------ 0.469 ------- 16 ------- Texas A&M (5-3)
50 ------ 0.442 ------- 58 ------- North Carolina (3-4)
51 ------ 0.435 ------- 42 ------- Iowa (5-2)
52 ------ 0.422 ------- 66 ------- Iowa State (2-5)
53 ------ 0.415 ------- 68 ------- Toledo (4-3)
53 ------ 0.415 ------- 69 ------- Louisiana Tech (4-3)
55 ------ 0.388 ------- 113 ------- Air Force (5-2)
56 ------ 0.367 ------- 67 ------- Houston (4-3)
57 ------ 0.361 ------- 25 ------- Wisconsin (4-2)
58 ------ 0.354 ------- 112 ------- Central Michigan (4-4)
59 ------ 0.347 ------- 80 ------- Indiana (3-4)
59 ------ 0.347 ------- 70 ------- Michigan (3-4)
59 ------ 0.347 ------- 93 ------- Navy (3-4)
59 ------ 0.347 ------- 27 ------- Florida (3-4)
63 ------ 0.340 ------- 40 ------- Maryland (5-2)
64 ------ 0.320 ------- 97 ------- Akron (4-3)
64 ------ 0.320 ------- 21 ------- Stanford (4-3)
66 ------ 0.319 ------- 46 ------- Oregon State (4-2)
67 ------ 0.306 ------- 64 ------- Cincinnati (3-3)
68 ------ 0.281 ------- 34 ------- Oklahoma State (5-2)
69 ------ 0.281 ------- 81 ------- Rutgers (5-2)
70 ------ 0.279 ------- 71 ------- Washington State (2-5)
70 ------ 0.279 ------- 108 ------- Texas-San Antonio (2-5)
72 ------ 0.278 ------- 96 ------- Western Kentucky (2-4)
73 ------ 0.252 ------- 121 ------- Wyoming (3-4)
73 ------ 0.252 ------- 101 ------- Illinois (3-4)
75 ------ 0.250 ------- 102 ------- Louisiana-Lafayette (3-3)
76 ------ 0.245 ------- 41 ------- Washington (5-2)
77 ------ 0.245 ------- 114 ------- Bowling Green (5-3)
77 ------ 0.245 ------- 95 ------- Middle Tennessee (5-3)
79 ------ 0.229 ------- 139 ------- OHIO (4-4)
80 ------ 0.224 ------- 99 ------- Western Michigan (4-3)
80 ------ 0.224 ------- 63 ------- California (4-3)
82 ------ 0.214 ------- 85 ------- Purdue (3-5)
83 ------ 0.204 ------- 32 ------- Arkansas (3-4)
83 ------ 0.204 ------- 65 ------- Syracuse (3-4)
83 ------ 0.204 ------- 73 ------- Texas Tech (3-4)
86 ------ 0.197 ------- 92 ------- Northern Illinois (5-2)
87 ------ 0.184 ------- 115 ------- Ball State (2-5)
87 ------ 0.184 ------- 123 ------- Kansas (2-5)
89 ------ 0.172 ------- 130 ------- FIU (3-5)
89 ------ 0.172 ------- 90 ------- Fresno State (3-5)
91 ------ 0.156 ------- 47 ------- Texas (3-4)
92 ------ 0.156 ------- 125 ------- Florida Atlantic (3-4)
93 ------ 0.139 ------- 86 ------- San Jose State (3-3)
94 ------ 0.132 ------- 107 ------- South Alabama (4-2)
95 ------ 0.129 ------- 106 ------- UAB (4-3)
96 ------ 0.111 ------- 79 ------- Temple (4-2)
97 ------ 0.104 ------- 136 ------- Louisiana-Monroe (3-3)
98 ------ 0.088 ------- 118 ------- Hawai'i (2-5)
98 ------ 0.088 ------- 163 ------- UNLV (2-5)
98 ------ 0.088 ------- 143 ------- New Mexico (2-5)
101 ------ 0.083 ------- 83 ------- Rice (3-3)
101 ------ 0.083 ------- 145 ------- Texas State (3-3)
101 ------ 0.083 ------- 126 ------- UTEP (3-3)
104 ------ 0.082 ------- 82 ------- San Diego State (4-3)
105 ------ 0.073 ------- 176 ------- Idaho (2-6)
106 ------ 0.063 ------- 75 ------- NC State (4-4)
107 ------ 0.061 ------- 142 ------- Buffalo (3-4)
107 ------ 0.061 ------- 116 ------- South Florida (3-4)
107 ------ 0.061 ------- 134 ------- Southern Miss (3-4)
110 ------ 0.041 ------- 161 ------- Army (2-5)
110 ------ 0.041 ------- 91 ------- Colorado (2-5)
110 ------ 0.041 ------- 210 ------- Eastern Michigan (2-5)
110 ------ 0.041 ------- 146 ------- North Texas (2-5)
110 ------ 0.041 ------- 135 ------- Tulane (2-5)
110 ------ 0.041 ------- 119 ------- Vanderbilt (2-5)
110 ------ 0.041 ------- 144 ------- Wake Forest (2-5)
117 ------ 0.031 ------- 160 ------- Massachusetts (2-6)
117 ------ 0.031 ------- 195 ------- New Mexico State (2-6)
119 ------ 0.028 ------- 156 ------- Connecticut (1-5)
120 ------ 0.020 ------- 191 ------- Georgia State (1-6)
120 ------ 0.020 ------- 173 ------- Kent State (1-6)
120 ------ 0.020 ------- 189 ------- Troy (1-6)
120 ------ 0.020 ------- 165 ------- Tulsa (1-6)
124 ------ 0.016 ------- 164 ------- Miami (OH) (1-7)
125 ------ 0.000 ------- 181 ------- SMU (0-6)

Don't know what type of modification could be used to remove the clear errors. No way after the Badgers destroyed the Falcons can BGSU be ranked that much higher than the Badgers. So too OHIO; seems that OHIO gets a BIG boost by losing to KY and Marshall while only having one "quality win" cough--over the Zips. Kent and Idaho are bottom feeders.

Nice to see O$U which lost at HOME to a 4-3 team and because of schedule has not beat anyone that the top tier MAC teams could not have also beat, is ranked out of the top 25.

Can UoMichigan turn their program around in the next five years or have the last few years put them in the permanent pretender class?

This is a similar idea to the pairwise system in College Hockey. The formulas are available and transparent and every team knows what they need to do at the end of the year to get into the tournament. It's really a great system and I wish college football had something similar

"actually matters"

(Reads the words "actually matters"; laughs out loud)

Can you further clarify or expand on the fact that losing doesn't matter in your ranking system? At some point if losses pile up, the ranking has to go down doesn't it?

I like that Miami is dead last both years.

I hate rankings and everything about them in the same way I hate the "judging" sports like Gymnastics & Figure Skating. I've said for years there should just be a tournament of conference champions and we burn every rating system to the ground (even though in reality we would still need to seed them, unfortunately).

Thank you for the compliment.

However, I argue there should be no middle ground when ranking college football teams. That's where mistakes get made, and justifiable talent gets lost in the haze.

Common sense and the eyeball test is why I stated Sagarin and his ilk do not measure teams by the same stick.

It's one thing to look at Alabama and say, "wow that teams out of place at #17." (I have little doubt Alabama will finish higher than that. And big picture in my dream world of a 16 team playoff, I have little doubt Alabama gets into the playoff pool.)

It's another thing to give Alabama these tests and not give these tests to every single team with the same level of diligence.

What is a greater mis-alignment?

- Ranking Alabama at #17 instead of Sagarin's #2
OR
- Ranking Nevada at Sagarin's #59 instead of #24

Nevada has lost to 3 of my top 16 teams by one possession. Yet Sagarin values them as a middle of the road college football team. Preposterous. Whatever jerry rigging Jeff Sagarin is doing for Alabama he is not doing for Nevada. That's the injustice.

And that injustice is always present when we operate under the false pretense of middle ground ranking systems (a combination of data and sporadic subjective interference).

Question: When referring to Sagarin's ratings, do you mean the ELO, the Predictor, or the average of the two?

I'm referring to his "rating" which as defined on the USA Today pulication:

"The overall RATING is a synthesis of the three different methods, with more total weight to the two completely SCORE-BASED methods and thus should be a good predictor in its own right."

I'm assuming the three methods referenced are the Predictor, ELO, and Golden Mean.

Agree on both scores:

1. Tournament with every FBS conference champion and six at-large, makes a total of 16. This takes four extra games, which could be incorporated into the current bowl system fairly easily.

2. "Judging sports" -- I just can't stand them. Every time there is an Olympic figure skating competition on the tube my wife and I get into an argument over this. She loves it, and I hate it. I can appreciate the beauty of the skating in its own right, but not what seems to me capricious scoring by the judges and what I call the "professional dilettante" (I know it's an oxymoron, but that's the point) attitude of the announcers.

1. Yes, I do believe these tests should be performed on every team.

2. To a degree, I believe most teams are ranked by the same standards. However, there are predetermined (intentionally or not) factors that give a certain perception to certain teams. Conference affiliation, returning starters, coaching, schedule, etc. But some of those factors are pretty important when ranking teams.

For example, let's say Auburn goes 11-1, but doesn't make it to the SEC Championship because their only loss is to 12-0 Mississippi State. That means Auburn would end up beating anywhere between 5-7 ranked teams, depending on which teams finishes where (currently South Carolina and Texas A&M are unranked). Four of those games they will have won were on the road.

Then let's say there is an 11-1 Big 12 Champion, 12-1 Big Ten Champion, 12-1 PAC 12 Champion, all of whom have less impressive resumes than Auburn. Statistically, you may not take Auburn or because they didn't win their conference. But look at who they beat, where they beat them and how they finished the season. How could you keep them out of a 4-team playoff? Of course you wouldn't keep them out of a 16-team playoff, but that's not what is in place right now.

I only bring this up because some people (not you) feel like the SEC gets a leg up on other conferences simply based on reputation and notoriety. Well, there's a reason for that reputation: it's because the SEC is the best conference in college football, and it's really not that close. Florida State won a national championship because they are modeled to be an SEC team, they just happen to play in the ACC.

3. I think my disagreement simply lies that you can't just rank a team strictly based on statistical analysis. You think there shouldn't be a middle ground between stats and visual, I think there can be.

With the same diligence is the operative word. Its an important distinction and you left that out.

Let's say someone watched Alabama 3 times on national TV and they want to make a conclusion on Alabama from that experience. I don't necessarily have a problem with that approach so long as the same person who is deriving conclusions watches 3 Nevada games, and 3 games of everyone else. It would take one person most of the work week to watch every FBS game (at an accelerated pace). The "eye test" is an invalid approach because teams will be left out of the eye test analysis. And when they are, the data forever skews results of the teams that were given the eye test away from teams that weren't given the eye test.

Statistically I'm arguing that the conference championships are irrelevant. I also don't care if a games on the road, in Atlanta, or in 2 feet of snow.

I am looking at who Auburn beat. And I'm also looking at who every one else beat by the same standard. If Auburn secures enough of those WVWTs, they'll be there, don't worry.

As I mentioned in the thread about WVWTs. My hindsight evaluation revealed that the team in college football with the most end of year WVWTs played in the National Championship game every year of the data I reviewed back to 2005 (played in but not necessarily won). Every MAC team that had the most WVWTs relative to the MAC played in the MACC every year back to 2005 (played in but not necessarily won).

Also I agree the SEC is the strongest conference. But I don't think the strength is coming from all the teams and all the reasons a common opinion may have. I look to the teams in that conference that get WVWTs. They produce results.

For the reasons stated above. Visual analysis cannot assess every team with the same diligence. We both believe it is fair to judge every team by the same standard. If you believe applying the same diligence to whatever method you'd like to every FBS team is important then I recommend you abandon your position on middle ground rankings. Otherwise we can agree that I will disagree.

From the win-loss data I've reviewed since 2005 I don't see any evidence that suggests margin of victory is a useful indicator of outcomes.

Sagarin has hardly created a crystal ball.

My theory is that Jerry Rig Jeff uses margin of victory as duct tape in his algorithms. I don't doubt that his model works better with margin of victory. I'm just not impressed with his model all together. And I take exception to some of the unpalatable rankings he puts on the market (such as Jeff ranking 20+ FCS teams ahead of a 4-4 OHIO team (Sagarin 139 LOL). Give me a break Jeff. Are you going to tell me Alabama would beat the Browns too?)

But I'll go further to say his model doesn't produce more useful results than mine (and to clarify, mine isn't going to win you any money either).

I ran through last week's prediction by Jeff (Oct 11 rankings vs week of OCT 18 results).

Jeff predicted 33 out of 49 FBS match-ups (67%).

I ran my week 7 rankings as well. And again, my rankings aren't intended to predict the future; simply rank teams by an objective measure. My week 7 rankings predicted 28 out of 48 FBS match-ups (58%) (SMU and UNT were tied in my rankings and played).

For the most part we both predicted the same outcomes for each match-up.

We flip flopped on a number of games. He had FSU over Notre Dame. I had Arizona State over Stanford. He thought Florida was going to win. I "knew" (just kidding) Missouri was going to win. Jeff picked UMass correctly over EMU. I did not. The flip flops were all over the map (top 10 to bottom 10).

Yeah Jeff predicted the outcome of 5 additional games that my rankings didn't. When I was right and he was wrong the margin of victory was 14.4. When he was right and I was wrong, the margin of victory was 9. He missed 21 pt and 29 pt blowouts that I got right. Meanwhile I missed five one possession games that he got right. What does this tell you? I don't know.

Is a 9% difference significant? I say no because the difference is not actually 9%. You cannot apply that statistical advantage to any game you choose. We have independent variables at play. You pull a random game out of a hat you have to calculate the probability that we don't match predictions and the probability that he is right. Since there's only a 30% chance we don't match, the advantage the Sagarin ratings would give you over my rankings is week 7 is only ~2.5%.

Long story short, Sagarin can't predict the future (with or without margin of victory data). But you already knew that.