The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.

In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.

By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?

On Tuesday, November 15, 2022 at 10:22:42 AM UTC-8, JGibson wrote:
> The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.
>
> In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.
>
> By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?

You see, that's the thing about college football. And you can't use the Group of 5 Argument either, in this case.

To me, the criteria should be as follows:

1) All undefeated teams in recognized FBS.
2) After that, you rank the conference champs for the rest of the spots.

Otherwise, you get what we have now: One completely dominant conference that, if you wanted the "four best teams", I could say Georgia, Tennessee, LSU, Alabama and dare anybody (including Ohio State and Michigan) to say otherwise.

Mike

> > By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?
> You see, that's the thing about college football. And you can't use the Group of 5 Argument either, in this case.
>
> To me, the criteria should be as follows:
>
> 1) All undefeated teams in recognized FBS.
> 2) After that, you rank the conference champs for the rest of the spots.
>
> Otherwise, you get what we have now: One completely dominant conference that, if you wanted the "four best teams", I could say Georgia, Tennessee, LSU, Alabama and dare anybody (including Ohio State and Michigan) to say otherwise.

The problem with your method is, Army (or a newly-independent Navy) schedules 10 cupcakes plus the other two service academies, and they have a decent shot at getting in.

The only NCAA tournament I can think of that doesn't depend on the "eye test" in any way to fill its field is men's ice hockey - and the method it uses doesn't really work in a sport like football where, with very few exceptions, teams don't play each other more than once in a season.

On Tuesday, November 15, 2022 at 11:35:47 AM UTC-8, The NOTBCS Guy wrote:

> The problem with your method is, Army (or a newly-independent Navy) schedules 10 cupcakes plus the other two service academies, and they have a decent shot at getting in.

Which is the method most outside the SEC DO attempt with their non-conference schedules, at least in the Power 5, unless ESPN walks up with a large wad of cash.

The problem is, what happens when it's pretty clear the four best teams are beating the shit out of each other in one conference?

Mike

On 2022-11-15, JGibson <james.m.gibson@gmail.com> wrote:
> The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.
>
> In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.
>
> By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?

I don't think predictor should even be a factor. Straight ELO win/loss, because
when it comes down to it, that's what wins at the end of the year. The W.

--
Being against torture ought to be sort of a bipartisan thing.
-- Karl Lehenbauer

On 2022-11-15, The NOTBCS Guy <don.p.del.grande@gmail.com> wrote:
>> > By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?
>> You see, that's the thing about college football. And you can't use the Group of 5 Argument either, in this case.
>>
>> To me, the criteria should be as follows:
>>
>> 1) All undefeated teams in recognized FBS.
>> 2) After that, you rank the conference champs for the rest of the spots.
>>
>> Otherwise, you get what we have now: One completely dominant conference that, if you wanted the "four best teams", I could say Georgia, Tennessee, LSU, Alabama and dare anybody (including Ohio State and Michigan) to say otherwise.
>
> The problem with your method is, Army (or a newly-independent Navy)
> schedules 10 cupcakes plus the other two service academies, and they
> have a decent shot at getting in.

There has to be a certain level that gets you considered. Something like
why BYU didn't really win the national title back in 1984 (beating a 6-6
Michigan team is not a qualifying win).

>
> The only NCAA tournament I can think of that doesn't depend on the "eye test" in any way to fill its field is men's ice hockey - and the method it uses doesn't really work in a sport like football where, with very few exceptions, teams don't play each other more than once in a season.
>

--
Being against torture ought to be sort of a bipartisan thing.
-- Karl Lehenbauer

On Tuesday, November 15, 2022 at 7:42:08 PM UTC-5, Con Reeder, unhyphenated American wrote:
> On 2022-11-15, JGibson <james.m...@gmail.com> wrote:
> > The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.
> >
> > In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.
> >
> > By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?
> I don't think predictor should even be a factor. Straight ELO win/loss, because
> when it comes down to it, that's what wins at the end of the year. The W.
>

That's the reason I'm highlighting FPI/SOR and Dolphin - because they make both rankings available. Sagarin and Massey no longer make their BCS-type rankings any more. Both Dolphin standard and ESPN's SOR actually have TCU #1. Colley actually has Georgia #1.

On Tuesday, November 15, 2022 at 7:59:17 PM UTC-5, JGibson wrote:
> On Tuesday, November 15, 2022 at 7:42:08 PM UTC-5, Con Reeder, unhyphenated American wrote:
> > On 2022-11-15, JGibson <james.m...@gmail.com> wrote:
> > > The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.
> > >
> > > In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.
> > >
> > > By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?
> > I don't think predictor should even be a factor. Straight ELO win/loss, because
> > when it comes down to it, that's what wins at the end of the year. The W.
> >
> That's the reason I'm highlighting FPI/SOR and Dolphin - because they make both rankings available. Sagarin and Massey no longer make their BCS-type rankings any more. Both Dolphin standard and ESPN's SOR actually have TCU #1. Colley actually has Georgia #1.

The other thing is that straight up ELO gives you debacles like Oklahoma being #1 in 2003 right after their 35-7 loss to K-State in the Big 12 championship game, so the Sooners play in the BCS title game instead of USC vs. LSU.

On 2022-11-16, JGibson <james.m.gibson@gmail.com> wrote:
> On Tuesday, November 15, 2022 at 7:59:17 PM UTC-5, JGibson wrote:
>> On Tuesday, November 15, 2022 at 7:42:08 PM UTC-5, Con Reeder, unhyphenated American wrote:
>> > On 2022-11-15, JGibson <james.m...@gmail.com> wrote:
>> > > The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.
>> > >
>> > > In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.
>> > >
>> > > By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?
>> > I don't think predictor should even be a factor. Straight ELO win/loss, because
>> > when it comes down to it, that's what wins at the end of the year. The W.
>> >
>> That's the reason I'm highlighting FPI/SOR and Dolphin - because they make both rankings available. Sagarin and Massey no longer make their BCS-type rankings any more. Both Dolphin standard and ESPN's SOR actually have TCU #1. Colley actually has Georgia #1.
>
> The other thing is that straight up ELO gives you debacles like
> Oklahoma being #1 in 2003 right after their 35-7 loss to K-State in
> the Big 12 championship game, so the Sooners play in the BCS title
> game instead of USC vs. LSU.

That goes to what I call qualifying and disqualifying events. Doesn't
matter if you're 12-0 (or 11-1) if you don't have a qualifying (good victory)
event. Likewise a bad loss without more good victories.

The NCAA Final Four selection committee sort of does this. I know they're
trying to get humans out of the process, but we serve a function from
time to time.

--
How far can you open your
mind before your brains
fall out?

On Wednesday, November 16, 2022 at 3:23:00 AM UTC-8, Con Reeder, unhyphenated American wrote:

> That goes to what I call qualifying and disqualifying events. Doesn't
> matter if you're 12-0 (or 11-1) if you don't have a qualifying (good victory)
> event. Likewise a bad loss without more good victories.

To me, if 12-0 isn't good enough to get in the Playoff, then either there are five or more 12-0s or the team and conference should not even be considered FBS. At the latter point, the only "bowl" that 12-0 team should be going into is the FCS Playoffs.

Mike

> I don't think predictor should even be a factor. Straight ELO win/loss, because
> when it comes down to it, that's what wins at the end of the year. The W.

What about home field advantage? Shouldn't that play some part?

Here are my top 25 when I use a variation on Colley that does not include the two "free" games (a 1-point loss to NFL All-Stars and a 3-point win over Little Sisters of the Poor, assuming all other games are won by 1 point) and includes all FCS teams separately (Colley ranks FCS teams first, then groups them so that each group has approximately the same total number of games played against FBS teams), counting all non-Division I opponents as a single team:
1. Georgia
2. TCU
3. Tennessee
4. Ohio State
5. Alabama
6. USC
7. LSU
8. Clemson
9. Michigan
10. Oregon
11. Utah
12. UCLA
13. Penn State
14. Washington
15. Mississippi
16. North Carolina
17. Kansas State
18. Oregon State
19. Notre Dame
20. Mississippi State
21. Texas
22. Oklahoma State
23. Florida
24. Florida State
25. Central Florida
Note that the #35 team in the country is Sacramento State, which is higher than both Arkansas and Auburn. Somebody at the Pac-12 office probably just saw this and is on the phone to the AD right now...
(Somewhere in Sacramento: "Only if Davis comes in as well")
(Somewhere in Berkeley: "But...but...we've managed to avoid having to play men's basketball in Davis for, what, 50 years now?" Seriously, for a long time, up through somewhere in the mid-80s, one of Cal's two allowed "exhibition" games was against Davis; the bylaws require that Cal host the game for it not to count against Cal's schedule limit. Even after Davis joined D1 in 2007 or so, I don't think they have played each other outside of Berkeley..)

On 2022-11-16, The NOTBCS Guy <don.p.del.grande@gmail.com> wrote:
>> I don't think predictor should even be a factor. Straight ELO win/loss, because
>> when it comes down to it, that's what wins at the end of the year. The W.
>
> What about home field advantage? Shouldn't that play some part?
>
> Here are my top 25 when I use a variation on Colley that does not include the two "free" games (a 1-point loss to NFL All-Stars and a 3-point win over Little Sisters of the Poor, assuming all other games are won by 1 point) and includes all FCS teams separately (Colley ranks FCS teams first, then groups them so that each group has approximately the same total number of games played against FBS teams), counting all non-Division I opponents as a single team:
> 1. Georgia
> 2. TCU
> 3. Tennessee
> 4. Ohio State
> 5. Alabama
> 6. USC
> 7. LSU
> 8. Clemson
> 9. Michigan
> 10. Oregon

A zero-loss Michigan lower than a one-loss Tennessee and a Clemson
punked by Notre Dame? Certainly with the undoubtedly strong Michigan
OOC schedule there....oops. Good here.

--
The minimum wage law is most properly described as a law saying
employers must discriminate against people who have low skills.
-- Milton Friedman

> >> I don't think predictor should even be a factor. Straight ELO win/loss, because
> >> when it comes down to it, that's what wins at the end of the year. The W.
> >
> > Here are my top 25 when I use a variation on Colley that does not include the two "free" games (a 1-point loss to NFL All-Stars and a 3-point win over Little Sisters of the Poor, assuming all other games are won by 1 point) and includes all FCS teams separately (Colley ranks FCS teams first, then groups them so that each group has approximately the same total number of games played against FBS teams), counting all non-Division I opponents as a single team:
> > 1. Georgia
> > 2. TCU
> > 3. Tennessee
> > 4. Ohio State
> > 5. Alabama
> > 6. USC
> > 7. LSU
> > 8. Clemson
> > 9. Michigan
> > 10. Oregon
> A zero-loss Michigan lower than a one-loss Tennessee and a Clemson
> punked by Notre Dame? Certainly with the undoubtedly strong Michigan
> OOC schedule there....oops. Good here.

You're the one who wants a non-margin of victory system.

> > What about home field advantage? Shouldn't that play some part?

Actually, if I use 1.4/1/0.6 like some versions of RPI do (i.e. a home win is worth 0.6 of a neutral win, and an away win is worth 1.4), I get...Holy Cross ranked ahead of both LSU and Michigan.

It's a little better with 1.3/1/0.7, but Sacramento State and Holy Cross are both ranked ahead of Notre Dame and Texas.

On Wednesday, November 16, 2022 at 11:10:40 AM UTC-5, The NOTBCS Guy wrote:
> > >> I don't think predictor should even be a factor. Straight ELO win/loss, because
> > >> when it comes down to it, that's what wins at the end of the year. The W.
> > >
> > > Here are my top 25 when I use a variation on Colley that does not include the two "free" games (a 1-point loss to NFL All-Stars and a 3-point win over Little Sisters of the Poor, assuming all other games are won by 1 point) and includes all FCS teams separately (Colley ranks FCS teams first, then groups them so that each group has approximately the same total number of games played against FBS teams), counting all non-Division I opponents as a single team:
> > > 1. Georgia
> > > 2. TCU
> > > 3. Tennessee
> > > 4. Ohio State
> > > 5. Alabama
> > > 6. USC
> > > 7. LSU
> > > 8. Clemson
> > > 9. Michigan
> > > 10. Oregon
> > A zero-loss Michigan lower than a one-loss Tennessee and a Clemson
> > punked by Notre Dame? Certainly with the undoubtedly strong Michigan
> > OOC schedule there....oops. Good here.
> You're the one who wants a non-margin of victory system.
> > > What about home field advantage? Shouldn't that play some part?
> Actually, if I use 1.4/1/0.6 like some versions of RPI do (i.e. a home win is worth 0.6 of a neutral win, and an away win is worth 1.4), I get...Holy Cross ranked ahead of both LSU and Michigan.
>
> It's a little better with 1.3/1/0.7, but Sacramento State and Holy Cross are both ranked ahead of Notre Dame and Texas.

Where do those weights come from?

If the sd of games is 14 and the home field advantage is 3, the home team has a 58.8% chance against a peer. I reckon this means the weights should be 1.21/1/0.85 so there is the same expected value at home or away (as 1.21*(1-0.588)= 1*0.5= 0.85*0.588). But I am not savvy in the ways of RPI.

> > Actually, if I use 1.4/1/0.6 like some versions of RPI do (i.e. a home win is worth 0.6 of a neutral win, and an away win is worth 1.4), I get...Holy Cross ranked ahead of both LSU and Michigan.
> >
> > It's a little better with 1.3/1/0.7, but Sacramento State and Holy Cross are both ranked ahead of Notre Dame and Texas.
> Where do those weights come from?
>
> If the sd of games is 14 and the home field advantage is 3, the home team has a 58.8% chance against a peer. I reckon this means the weights should be 1.21/1/0.85 so there is the same expected value at home or away (as 1.21*(1-0.588)= 1*0.5= 0.85*0.588). But I am not savvy in the ways of RPI.

Ask the people who use RPI where they get the numbers from.

Using 1.21 / 1 / 0.85, I get:
1. Georgia
2. TCU
3. Tennessee
4. Alabama
5. Clemson
6. Ohio State
7. USC
8. LSU
9. Oregon
10. Utah
11. Michigan
12. North Carolina
13. Washington
14. Mississippi
15. UCLA
16. Penn State
17. Kansas State
18. Oregon State
19. Notre Dame
20. Texas
21. Mississippi State
22. Florida State
23. Oklahoma State
24. Sacramento State
25. Central Florida

On 2022-11-16, The NOTBCS Guy <don.p.del.grande@gmail.com> wrote:
>> >> I don't think predictor should even be a factor. Straight ELO win/loss, because
>> >> when it comes down to it, that's what wins at the end of the year. The W.
>> >
>> > Here are my top 25 when I use a variation on Colley that does not include the two "free" games (a 1-point loss to NFL All-Stars and a 3-point win over Little Sisters of the Poor, assuming all other games are won by 1 point) and includes all FCS teams separately (Colley ranks FCS teams first, then groups them so that each group has approximately the same total number of games played against FBS teams), counting all non-Division I opponents as a single team:
>> > 1. Georgia
>> > 2. TCU
>> > 3. Tennessee
>> > 4. Ohio State
>> > 5. Alabama
>> > 6. USC
>> > 7. LSU
>> > 8. Clemson
>> > 9. Michigan
>> > 10. Oregon
>> A zero-loss Michigan lower than a one-loss Tennessee and a Clemson
>> punked by Notre Dame? Certainly with the undoubtedly strong Michigan
>> OOC schedule there....oops. Good here.
>
> You're the one who wants a non-margin of victory system.

Which is why the one-loss teams ahead of Michigan offends the eye.

But it doesn't offend the sensibilities again because of my bugbear,
which is a schedule that offers no real difficulty. If you meet that
threshold, then MOV moves me not.

--
"The formula for achieving a successful relationship is simple: you
should treat all disasters as if they were trivialities but never
treat a triviality as if it were a disaster." -- Quentin Crisp

On Tuesday, November 15, 2022 at 1:22:42 PM UTC-5, JGibson wrote:
> The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.
>
> In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.
>
> By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?

I have a simple, accurate two (2) step process to
determine the top four (4) college football teams.

Step #1:

The Ohio State University Buckeyes are number one
and you let the crowd squabble over the next three teams.

Step #2:

If the Alabama Crimson Shit Kickers are on the list
you look for all the crooked crap and preferential
treatment the SEC pulled to get them there.

Irish Mike

On Saturday, November 19, 2022 at 7:57:40 AM UTC-8, Irish Mike wrote:
> On Tuesday, November 15, 2022 at 1:22:42 PM UTC-5, JGibson wrote:
> > The problem I see is that when people say "best 4 teams", they are usually talking about something predictive. But, look at any predictive rating, and TCU is far down. But what about most deserving? Here's a look at two rating systems that actually break down both.
> >
> > In FPI, TCU is ranked 16th (!). But in the strength of record on the FPI page, which asks the question, what is the likelihood of this record, TCU is ranked 1st. Similar are the Dolphin rankings where his predictive rankings have TCU 10th but his standard rankings (which uses PF/PA to set the schedule strength but then only uses W-L against that) also have TCU 1st.
> >
> > By the supposed criterion of "best 4 teams", TCU might not be included. But if the season stopped now, wouldn't you want to include TCU in your 4 team playoff?
> I have a simple, accurate two (2) step process to
> determine the top four (4) college football teams.
>
> Step #1:
>
> The Ohio State University Buckeyes are number one
> and you let the crowd squabble over the next three teams.
>
> Step #2:
>
> If the Alabama Crimson Shit Kickers are on the list
> you look for all the crooked crap and preferential
> treatment the SEC pulled to get them there.
>
> Irish Mike

Let's try that again...

On Saturday, November 19, 2022 at 7:57:40 AM UTC-8, Irish Mike wrote:

> Step #1:
>
> The Ohio State University Buckeyes are number one
> and you let the crowd squabble over the next three teams.
>
> Step #2:
>
> If the Alabama Crimson Shit Kickers are on the list
> you look for all the crooked crap and preferential
> treatment the SEC pulled to get them there.

Step #3:

Watch The Ohio State University Buckeyes get rolled by multiple touchdowns, this year by Georgia.

Mike