I was wondering why Oz had such a problem with this, then I went and found where Ohio State ranks on this.

Yeah Ozcat went out of his way to repudiate my rankings without giving them so much as a trial run. I also suspect butthurt.

(Butthurt: an inappropriately strong negative emotional response from a perceived personal insult.)

Thursday night action:

By the "Predictor" rating, "Golden_Mean" rating, and OVERALL rating Sagarin had Virginia Tech beating Miami last night. (His ELO_Chess rating, a contributor to the overall rating, had Miami 8 spots ahead of Virginia Tech)

What actually happened was a Hurricane 24 pt victory on the road.

And had Ozcat given a closer look at my rankings, he would've seen the U at #18 and Va Tech at #33.

So after only 2 games:

- Sagarin: 1-1
- Me*: 2-0

*The point of my most recent string of posts is to weaken Sagarin's approach by demonstrating simple, clearly defined, measurements that can effectively replace his rankings.

I went over to the MAC Message board to get some feedback.

Long story short I made some modifications to the rankings. There wasn't a big shakeup, but I described my modifications and the big individual movers in the post below copied from that site. The original post has been updated:

---------------------------------------------------------------
=================================================
MileHighBronco Wrote:
I found it somewhat odd that, for example, a 6 - 1 Michigan State team is ranked not only worse than a 3 - 4 Michigan team but behind a 4 - 3 Toledo team. Central Michigan one spot behind MSU? Ahhh, no. These are just examples I spotted. There are many that just don't seem to make any sense but then I agree with Axe about small sample size of the data.
=================================================

I agree with you guys that there are are teams that currently have a placement that appears to be mis-aligned with the current level of talent that team puts on the field. Another example of this is 3-4 Texas, ranked #91. (Texas has wins over my #110 North Texas, my #87 Kansas, and my #52 Iowa State). Texas has zero (0) WVWTs or wins vs teams with a 0.500 record. Both Iowa State and Kansas were able to earn those accomplishments, Texas has not (yet). But don't look for Texas to sit at the bottom all season.

As far as the numbers go, there's just no way for me to justify Michigan State in the Top 32, let alone Top 16. Their only loss is to Oregon by 19. So what? You know who just lost to Oregon by 19 last night? My #80 Cal.

I made a minor inclusion to the spreadsheet this morning that help teams like Michigan State. Currently teams receive a one time Strong Loss Bonus (SLB) if their only losses are to Top 16 and Top 32 teams by 1 possession (T16 = 1pt; T32 = 0.5pt). I added two additional loss bonuses. Teams who only have losses to Top 16 teams get 0.25 pts. And teams who only have losses to T32 teams get 0.125 pts. Each tier is halved. And teams can only claim one (1) loss bonus total. Therefore teams receive one (1) 1pt, 0.5pt, 0.25pt, 0.125pt, or 0pt loss bonus.

What was affected by this inclusion? The biggest movers were Michigan State up 6, Minnesota up 4, Oregon State up 4. Fifteen (15) other teams received new loss bonuses, on the whole, the majority of teams move 0-1 spots, and no more than 6. The original post has been updated with these changes.

I then made a slightly more significant inclusion to my model. Previously I awarded Incremental Point Fraction Bonuses (IPFB) for indirect achievements (wins against team with winning records that also had wins against teams with winning records). Before this morning I only awarded the IPFBs for wins against winning teams. Now I award the same IFPB whether or not the team you beat has a winning record. (IFPBs are explained in more detail in the Bobcat Attack thread).

What was affected by this inclusion? There were 68 occurrences of this new IFPB. But not a big overall rankings shake up. Big individual risers were Oklahoma State up 16. Notre Dame up 11. Cal up 17. Minnesota up 17. Michigan State up 10 more (16 total from the first described modification). The average spot change was ~4.

Long story short, people can complain all they want about Michigan State or Texas OPPA++ rankings, but in reality, those teams quite literally haven't accomplished anything significant enough to warrant Top 32. I have no idea how Jerry Rig Jeff concludes Michigan State is #11. It seems to defy all numerical evidence. But I don't doubt that they can prove themselves to be a Top 32 OPPA++ team in a couple weeks.

If I went into OPPA to duct tape Michigan State to the Top 32 now, the integrity of my system would be compromised.

After the described modifications, Michigan State is now ahead of Michigan, Toledo, and Central Michigan (the conflicts MileHigh mentioned). Emotionally I agreed with MileHigh before the modifications, but now the numbers justify that.

Any other potential mis-alignments? (You've found two so far MileHigh)

My OPPA++ rankings correctly picked the winner of 36 of 46 FBS match-ups (78%) this week. Sagarin correctly picked 37 of 46 FBS match-ups (80%) this week. Seventy-four percent of our picks this week matched.

The statistical advantage of using Sagarin's overall ratings instead of my OPPA++ rankings this week to predict win/loss was +0.7%.

The results speak for themselves.

One week sample.

Key result: Ohio at WMU. Not concerned, in my invalid opinion, about the rest. What's this all say about making us better?

What you are doing here is fine. It's just of little interest to me. Fair enough; I'll leave it be and hope you enjoy it and hope you prove accurate.

Think about it, though and I think that you'll find a conference championship game to be more than a one week sample.

Thank you for the compliment Optimist.

I appreciate the proposed solution. My emphasis though is on the word "easy". And I don't see an easy way to convey my conclusions without handing over the inner workings of a model I created from scratch. I do not feel compelled to do that (even with a password).

I've discussed very openly precisely what affects a team's OPPA++ ranking.

I believe I've demonstrated an advanced understanding of this topic. For the remaining doubters of my simple alternative to Jerry Rig Jeff I will leave you with this:

"Occam's razor (also written as Ockham's razor and in Latin lex parsimoniae) is a problem-solving principle devised by William of Ockham (c. 1287–1347). It states that among competing hypotheses, the one with the fewest assumptions should be selected. Other, more complicated solutions may ultimately prove correct, but—in the absence of certainty—the fewer assumptions that are made, the better.

....

For each accepted explanation of a phenomenon, there is always an infinite number of possible and more complex alternatives, because one can always burden failing explanations with ad hoc hypothesis to prevent them from being falsified; therefore, simpler theories are preferable to more complex ones because they are better testable and falsifiable.[1][10][11]"

http://en.m.wikipedia.org/wiki/Occam's_razor

Objective point scores for teams currently range from 0pts (SMU) to 10.77pts (Mississippi State). OPPA stands for Objective Points Per Attempt. So what you're seeing is the raw score divided by the number of attempts. (10.77/7 = 1.539)

The most points you can earn from a win is 2 (1pt plus the max IPFB of 1pt). So far LSU has the only 2pt victory for beating then #2 (and still #2 Ole Miss). This is a relative award. Theoretically more than one team can earn the max 2 pt victory. The max score is given to the team with a win over the team with the most wins vs winning teams to lose. A win against a team that only has 1 WVWT would be worth 1.2 pts currently (and will lose value as the season progresses because other teams push the max pt limit further).

I will see if I can put together some sort of visual on pt scoring in the coming weeks.

I've genuinely enjoyed this process so far. We'll see how the message board discussion evolves. But don't expect to see a file dump on here any time soon.

Really?! Arizona at #3 lost you?

Was it Arizona's October 2 victory at Oregon by 7 points that lost your trust? Or was it Arizona's one loss that exposed them? (lost by 2 to my then #8 USC; game included an Arizona failed 2 pt conversion)

If anything Michigan State in anyone's Top 10 throw's me for a loop. Michigan State is 6-0 against teams with losing records and 1-1 against teams with winning records. I guess Arizona would've been better off beating Nebraska than Oregon (who Michigan State lost to).

P.S.

OPPA++ has Arizona at #10 this week. For my own enjoyment I would like to know where you think the other plot flaws are.

Situation,

Having grown up in penn state country, I was shocked at where you had them placed in your initial rankings. I thought they would be blown out by OSU - shoulda paid attention to your numbers.

I got a mouth full from the cubicle farm and the doors last Friday about that.

There's probably over a 90% chance Penn State already has 2 WVWTs that will hold through the season (placing them top half FBS). If Akron proves they can cobble together a few more wins Penn State will have a third. They're favored (in my book) to pick up a potential 4th WVWT against Maryland. That would place Penn State firmly in the Top 30 end of year.

SEC was 7-3 and two of those wins were against SEC teams. Two of the losses....1 to the national champion and 1 to the eventual #6 team (Oklahoma). Third loss Ga to Neb with Murray watching from sidelines.

Of note:
The Big Ten is 6-15 on or after Jan. 1 the last four seasons, including a 4-9 mark against the SEC.

The Big Ten finished with a 4-10 record in the Rose Bowl in the BCS era.

The Big Ten is 5-21 against Top 10 nonconference teams since 2006.

The Big Ten gets the respect it deserves - which is little and none.

Now the Pac may be a different story -- 6-3 last year.