Bobcat Attack - Ohio Football - OUT OF DATE: Week 11 OPPA++ Rankings (SEE FINAL REV)

Ohio Football Topic

Topic: OUT OF DATE: Week 11 OPPA++ Rankings (SEE FINAL REV)

Page: 2 of 2

The Situation

General User

Member Since: 7/13/2010

Location: Columbus, OH

Post Count: 957

mail

The Situation

mail

Posted: 11/4/2014 11:57 AM

settings_applications edit_note

L.C.

I agree that your described approach is a stronger way to rank the bottom half. I do however expect the top half to be affected as wins over, say 3-5 Texas would be worth the same as 3-5 UL-Monroe.

At some point stretching the band for what is a notable accomplishment works against you (Im not saying what youve described is at that point). For example, the opposite extreme would call all wins equal. I think most here can agree that all wins are not equal (but I know that's not what you were getting at).

I awarded loss bonuses for one possession losses to top teams because they're actually not that common against top teams, unless of course you are a top team. I'm not at my computer, but I will check today, loss bonuses mostly help sort top half teams that lost to top half teams. Incremental Point Fraction Bonuses help sort the bottom half (along with win percentage).

Each step is ranked side by side to see what kind of affect IFPBs or Loss Bonuses have on the overall rankings.

I will modify the rankings some time this week to include a third segregation similar to what you recommended.

I think the strength of segregating only two pools is that by the time you get into the top pool, many of the variables at play have been neutralized. What I mean by that is no matter what schedule you played, if you're in the top half, you've earned your seat. The issue that some seem to have with that is, there are teams who didn't get a fair opportunity to earn their seat because the deck was stacked against them. But that's life. No one is giving out chips for uncashed potential.

Last Edited: 11/4/2014 12:02:55 PM by The Situation

L.C.

General User

Member Since: 9/1/2005

Post Count: 10,584

person

mail

L.C.

mail

Posted: 11/4/2014 12:12 PM

settings_applications edit_note

To a certain extent the problem is using a discrete function to approximate a continuous function. Overall team quality is continuous, while lumping them into groups "winning record" "losing record" maps them onto a discrete function with only two possible values.

Another interesting shortcoming in your system is that a team could have no wins over anyone with a winning record, yet have a winning record itself. We know this happens from looking at bowl teams. Is a win over that team really worth more than a win over some other team that might have a losing record solely as a function of the teams they played, but which actually does have some wins over teams with winning records?

Another modification you might consider would be to, before processing the teams at all, drop all wins over FCS teams. Next you could drop all wins over teams with no wins at all (or wins only over FCS teams). Beating winless teams and FCS teams should not count for anything, and not towards pushing a team into the "winning team" category.

..and yes, I know that teams in the top half would also be affected. That doesn't bother me. A win over 3-5 Texas, or 3-5 Monroe should be worth more than a win over 1-7 Kent, for example. The value of wins over mediocre teams should be a lot less than the wins over winning teams, though, so even though the teams in the top half might move a little, they wouldn't move at all. Thus, the more significant effect would be a more reasonable sorting of the bottom half.

Also, yes, I know that my proposed modification will move Ohio down substantially since most of its wins are over bad teams, while some of the teams that are ranked below them may have beaten some teams with "near winning" records.

Yet another possible change might be to do the intial calculation exactly as you have, but then do a second iteration, and add "bonus points" to teams with wins over teams in your top 10, lesser bonus points to teams in your "top 30", and still less bonus points over teams in your "top 50". That might improve sorting in your top half by giving teams some extra credit for especially good wins.

BTW, system simulation and stochastic process were some of my favorite courses, long, long ago, when I was studying operations research.

Last Edited: 11/4/2014 12:23:56 PM by L.C.

The Situation

General User

Member Since: 7/13/2010

Location: Columbus, OH

Post Count: 957

mail

The Situation

mail

Posted: 11/4/2014 12:46 PM

settings_applications edit_note

L.C. wrote:expand_more

+1

Do you see this Monroe?

This is what strong feedback looks like.

L.C.,

The two groups aren't discrete as any victory at any point in the season can score your team points in a full (basically) continuous range from 0-2 points. I refer to them as segregated groups but each team can pass on credit for their wins vs winning teams to teams that beat them. For example, currently, beating (4-5) Texas would be worth 0.6 points (because of their indirect wins), while beating (3-5) Louisiana Monroe would be worth 0.0 points. Beating Mississippi State would currently be worth 2 points.

In many ways I've already addressed what you've recommended but there may be some appreciable improvements from awarding slightly more credit for beating teams 1 or 2 games below 0.500.

The biggest challenge in all college football ranking systems is overcoming the inherent circular logic:

Q: How good was the win?

A: It depends on how good the team is.

Q: Well how good was the team?

A: Well you'll have to tell me how good their wins were.

Q: Well how good are the teams they played?

The affects of this circular logic can be mitigated through iteration.

Also I forgot to mention, the perceived quality of a team is often a victim of circumstance. Week after week there are match ups where one team that lost quite reasonably could have been the victor if they were scheduled as the home team. This is a butterfly effect x1000000.

Last Edited: 11/4/2014 12:55:34 PM by The Situation

L.C.

General User

Member Since: 9/1/2005

Post Count: 10,584

person

mail

L.C.

mail

Posted: 11/4/2014 1:01 PM

settings_applications edit_note

The Situation wrote:expand_more

... For example, currently, beating (4-5) Texas would be worth 0.6 points (because of their indirect wins), while beating (3-5) Louisiana Monroe would be worth 0.0 points. Beating Mississippi State would currently be worth 2 points.

In many ways I've already addressed what you've recommended but there may be some appreciable improvements from awarding slightly more credit for beating teams 1 or 2 games below 0.500.
...

Correct. By doing this you have already picked up a lot of the information, especially for teams in the top half. I think think that adding some credit for beating near-winning teams would improve the order of the bottom half of the ranking, but you need to keep the award low enough that it doesn't significantly change the top half.

PhiTau74

General User

PT74

Member Since: 8/6/2010

Location: Columbia, SC

Post Count: 458

person

mail

PhiTau74

mail

Posted: 11/4/2014 1:50 PM

settings_applications edit_note

The Situation wrote:expand_more

PhiTau74 wrote:expand_more

I said South Carolina's defense was terrible, maybe you have a reading comprehension problem.

You used South Carolina's offensive production against a common opponent (UK) in an attempt to demonstrate the type of numbers they'd put up on OHIO. I presented the final score against OHIO's common opponent as counter-evidence to introduce doubt that South Carolina would actually score that many points.

I comprehended. And responded.

PhiTau74 wrote:expand_more

Get a reality check and the idiot said they barely beat Furman? Since when is 41-10 close?

I imagine arguing with you in real life is quite difficult. I say this because we conveniently have an exact record of what was said and not said right here in this thread.

And unfortunately for you, I did not say that.

PhiTau74 wrote:expand_more

Get your green glasses off and come back to reality that the MAC sucks this year.

These rankings literally have all but 2 MAC teams ranked #75 or lower, including a 6-2 NIU team at #96.

I have this sinking feeling that nothing short of omitting the names of G5 schools will satisfy you.

PhiTau74 wrote:expand_more

I didn't go to South Carolina so I have no skin in the game but I know football. I watch SEC games every week and there is a reason 1-2 weeks ago that the SEC had 4 of the top 5 teams in college football.

Like you, I believe the SEC is the best conference. My rankings this week have the SEC in 3 of the top 4 overall positions. But unlike you, I believe I present my reasoning in a coherent fashion.

Let's look at the "top" team in the MAC. N. I'llinois lost to a bottom third SEC team Arkansas 52-14 this year. If USC can score 36 per game against the SEC then they can score 50 per game against the MAC and the the Arkansas games shows how low the MAC is this year. The MAC has had some really good teams the last 5 years but they are extremely weak this year. I'm not saying they would score 50 per game, no one will ever know but they certainly could put 50 up on any MAC team on any given day. South Carolina was preseason top 10, they suck because they have zero defense. Common opponent is one game?

The Situation

General User

Member Since: 7/13/2010

Location: Columbus, OH

Post Count: 957

mail

The Situation

mail

Posted: 11/4/2014 2:41 PM

settings_applications edit_note

What is your justification for claiming Northern is the top MAC team? I currently have them ranked #96, 6th in the MAC, 4 spots away from 8th.

The MAC is down this year.

Ball State and Toledo would've been tougher games.

GoCats105

General User

GC105

Member Since: 1/31/2006

Location: Seattle, WA

Post Count: 7,823

person

mail

GoCats105

mail

Posted: 11/4/2014 3:59 PM

settings_applications edit_note

Situation,

I probably need to go back and look at your original calculation techniques, but does a bad loss get taken account of at any point? I'm just thinking about scenarios like losing to a team with a losing record at home, when the home team has a very good winning record. I'm assuming this would just play into normal calculation and how the loss affects the overall score?

GoCats105

General User

GC105

Member Since: 1/31/2006

Location: Seattle, WA

Post Count: 7,823

person

mail

GoCats105

mail

Posted: 11/4/2014 4:02 PM

settings_applications edit_note

PhiTau74 wrote:expand_more

The Situation wrote:expand_more

PhiTau74 wrote:expand_more

I said South Carolina's defense was terrible, maybe you have a reading comprehension problem.

PhiTau74 wrote:expand_more

Get a reality check and the idiot said they barely beat Furman? Since when is 41-10 close?

PhiTau74 wrote:expand_more

Get your green glasses off and come back to reality that the MAC sucks this year.

PhiTau74 wrote:expand_more

Like you, I believe the SEC is the best conference. My rankings this week have the SEC in 3 of the top 4 overall positions. But unlike you, I believe I present my reasoning in a coherent fashion.

Off topic a bit, but Arkansas is not a lower third SEC team. I would put them somewhere in the middle. They are unfortunate to be in the West Division.

The Situation

General User

Member Since: 7/13/2010

Location: Columbus, OH

Post Count: 957

mail

The Situation

mail

Posted: 11/4/2014 6:54 PM

settings_applications edit_note

GoCats105 wrote:expand_more

I do not deduct points for any losses. For example, last week ECU and UCF took bad losses on the road. In those cases their OPPA scores would've only been marginally been affected win or lose because the teams they lost to weren't worth any significant ranking points.

There's certainly an opportunity to grade bad losses. But it really jumps into a world of assumptions that I wouldn't enjoy getting in to. And at some point, my ranking system just becomes Sagarin part II.

UpSan Bobcat

General User