Laughing at Ratings Discussion Over Coffee

[visitingcock]

Had some good laughs this morning along with a couple of coffee spews. Why do y’all give a crap where sports writers rank you? They don’t matter. It’s just a popularity poll. Some of you think you won’t be a top four team. THE HUMAN POLLS DON’T MATTER

What matters is NET. You're # 2.

In the ratings without human bias:
You’re number 1 in RPI.
You’re number 2 in Massey.
You’re number 2 in warren Nolan’s ELO

And in my theoretical Boneyard ballot based on Quad One wins (minus all losses), you are tied for Number Two - but get the tie breaker for having played the most Quad One games

1. SCar 9-0. = +9

2. UConn 10-3 = +7
3. Utah 9-2= +7

4. Indiana 9-3= + 6
5.. Stanford 9-3= +6

6. LSU 4-0= +4

7.
Notre Dame = + 2
Iowa = + 2
Colorado = + 2
tOSU = + 2
Duke = + 2
Maryland +2
Michigan + 2

14.
UNC +1
UCLA +1
Columbia +1
Villanova +1
FSU .+1

19.
NC State 0
USF 0
Gonzaga 0

Everybody else has more losses than Quad Ine wins.

 

[Topdawg]

Had some good laughs this morning along with a couple of coffee spews. Why do y’all give a crap where sports writers rank you? They don’t matter. It’s just a popularity poll. Some of you think you won’t be a top four team. THE HUMAN POLLS DON’T MATTER

What matters is NET. You're # 2.

In the ratings without human bias:
You’re number 1 in RPI.
You’re number 2 in Massey.
You’re number 2 in warren Nolan’s ELO

And in my theoretical Boneyard ballot based on Quad One wins (minus all losses), you are tied for Number Two - but get the tie breaker for having played the most Quad One games

1. SCar 9-0. = +9

2. UConn 10-3 = +7
3. Utah 9-2= +7

4. Indiana 9-3= + 6
5.. Stanford 9-3= +6

6. LSU 4-0= +4

7.
Notre Dame = + 2
Iowa = + 2
Colorado = + 2
tOSU = + 2
Duke = + 2
Maryland +2
Michigan + 2

14.
UNC +1
UCLA +1
Columbia +1
Villanova +1
FSU .+1

19.
NC State 0
USF 0
Gonzaga 0

Everybody else has more losses than Quad Ine wins.

I share your point of view. The AP poll is not important. What matters is that UConn continues to build back to strength as injured players return and rejoin the rotation. We deserve to get a good tournament seeding that respects our body of work. I would also love to see a Championship game with South Carolina. There is much basketball to be played between now and then!

 

[CL82]

Why do y’all give a crap where sports writers rank you?

It’s something to talk about.

 

[Urshurak]

When it becomes bracket time, what are the things they look at? Do they look at the AP poll / Coaches poll? If I remember right, I thought that the net rankings were not looked at as much.

Formula of some sort? Like Top 25 wins + SOS + conference record, etc. ?


I hear what you are saying, just not sure I understand exactly hat matters when it comes to tournament seeding.

To be honest, I kind of like what Geno said last year. Something along the lines of it doesn't matter, we have to win 6 games in any case.

 

[Centerstream]

Every rating/ranking is based upon human content. Every algorithm is "written" using human inputs and various weighted (ie. conference membership) factors. For example, which Conference is rated #1, #2 etc. But I have posted this before and have always been blasted for it. But having spent 25+ years as a software tester and PT coder, I think I know a bit about it... bash away...

 

[maple]

Had some good laughs this morning along with a couple of coffee spews. Why do y’all give a crap where sports writers rank you?

I like to complain

But otherwise yes, you are right.

 

[JellyBean]

You know folks gotta complain about something.

 

[DrummerDee]

Had some good laughs this morning along with a couple of coffee spews. Why do y’all give a crap where sports writers rank you? They don’t matter. It’s just a popularity poll. Some of you think you won’t be a top four team. THE HUMAN POLLS DON’T MATTER

What matters is NET. You're # 2.

In the ratings without human bias:
You’re number 1 in RPI.
You’re number 2 in Massey.
You’re number 2 in warren Nolan’s ELO

And in my theoretical Boneyard ballot based on Quad One wins (minus all losses), you are tied for Number Two - but get the tie breaker for having played the most Quad One games

1. SCar 9-0. = +9

2. UConn 10-3 = +7
3. Utah 9-2= +7

4. Indiana 9-3= + 6
5.. Stanford 9-3= +6

6. LSU 4-0= +4

7.
Notre Dame = + 2
Iowa = + 2
Colorado = + 2
tOSU = + 2
Duke = + 2
Maryland +2
Michigan + 2

14.
UNC +1
UCLA +1
Columbia +1
Villanova +1
FSU .+1

19.
NC State 0
USF 0
Gonzaga 0

Everybody else has more losses than Quad Ine wins.

Finally, a voice of reason. I don't get the angst over AP either..

 

[hard2please]

dont get too cocky

 

[visitingcock]

When it becomes bracket time, what are the things they look at? Do they look at the AP poll / Coaches poll? If I remember right, I thought that the net rankings were not looked at as much.

Formula of some sort? Like Top 25 wins + SOS + conference record, etc. ?


I hear what you are saying, just not sure I understand exactly hat matters when it comes to tournament seeding.

To be honest, I kind of like what Geno said last year. Something along the lines of it doesn't matter, we have to win 6 games in any case.

The ONLY ratings they look at are NET. They do not look at AP Coaches poll rankings at all . ( though they mmight have subliminal effectThey prety much follow NET but can adjust. The NET has a built-in SOS That has too much analysis in my opinion.

The committee can adjust the NET according to some rather loose guidlines that permit them to adjust it according to conference strength, opinions of the regional advisors, etc., etc, etc. In 2021, South Carolina was #5 at the end of the season. The committee bumped them up into the top four - where they actually ended up at the end of the tourney. They also play around with groupings (1-4, 5-8, 9-16, etc to try to balance the regions so that a one seed (Say Indiana) isnt paired in a region with a two seed they have played (say Iowa twice).

But, the NET is the guide that they use to try and produce the best pairings

 

[CL82]

But having spent 25+ years as a software tester and PT coder, I think I know a bit about it... bash away...

Physical therapy coder? Why does physical therapy need to be in code? I mean, I think it should be an open discussion between the therapist and the patient.

 

[visitingcock]

Every rating/ranking is based upon human content. Every algorithm is "written" using human inputs and various weighted (ie. conference membership) factors. For example, which Conference is rated #1, #2 etc. But I have posted this before and have always been blasted for it. But having spent 25+ years as a software tester and PT coder, I think I know a bit about it... bash away...

To some extent that is true. Early season rankings like Masseys have no data to enter on teams. They relate back to the previous data and supposedly don.t have enough new year data until after about ten games. However, the new data remains tinged by the old data. And, an algorithm can have the bias of its author. I have often wondered about an algorithmic bias for a factor that overrates Oregon since 2020. (I was biased against them that year.)

I think the algorithms are too complex. That's why i stick with a simple Quad One minus losses analysis. But yeah, that has a flaw in that I start with the over-complex NET rating that determines which opponents are quad one teams.

 

[Bone Dog]

The ONLY ratings they look at are NET. They do not look at AP Coaches poll rankings at all . ( though they mmight have subliminal effectThey prety much follow NET but can adjust. The NET has a built-in SOS That has too much analysis in my opinion.

The committee can adjust the NET according to some rather loose guidlines that permit them to adjust it according to conference strength, opinions of the regional advisors, etc., etc, etc. In 2021, South Carolina was #5 at the end of the season. The committee bumped them up into the top four - where they actually ended up at the end of the tourney. They also play around with groupings (1-4, 5-8, 9-16, etc to try to balance the regions so that a one seed (Say Indiana) isnt paired in a region with a two seed they have played (say Iowa twice).

But, the NET is the guide that they use to try and produce the best pairings

If you’re going to keep quoting me about the NET, I’d like a tribute plaque, or a mention, or at least a donut — I prefer Boston Creme by the way.

Yes, we’re on exactly the same page about how the committee will use the NET. It’s the rating tool we all hate early on but learn to love in February.

 

[OldSchooler]

I don't drink coffee.

 

[CL82]

I think the algorithms are too complex. That's why i stick with a simple Quad One minus losses analysis. But yeah, that has a flaw in that I start with the over-complex NET rating that determines which opponents are quad one teams

I have to admit that actually seems like pretty good data so long as you have enough games. If you feel like posting your net quad one rankings on a weekly basis, I think it would be interesting.

 

[CL82]

I don't drink coffee.

I like turtles.

 

[SVCBeercats]

Had some good laughs this morning along with a couple of coffee spews. Why do y’all give a crap where sports writers rank you? They don’t matter. It’s just a popularity poll. Some of you think you won’t be a top four team. THE HUMAN POLLS DON’T MATTER

There is one pollster that matters, the Division I Women's Basketball Championship Sport Committee. They choose the best 32 teams selected on an at-large basis. Then there is their task (poll?) of placing teams in bracket slots. Everything else is background noise. Like this post.

AP and coaches polls are a farce. They have not seen all of the teams play. At best they use box scores (stats) for the unseen teams. At worst they use opinions, not unlike most everyone on this board. The computer based polls see the stats (quantitative data) for all of teams' games but have no grasp of the human aspects of team play (qualitative data). Plus humans and their opinions construct these computer systems. It is all imperfect but we like bitch about the polls and computer systems. We like to correct their ratings with our far more accurate opinions. It is fun.

 

[diggerfoot]

Every rating/ranking is based upon human content. Every algorithm is "written" using human inputs and various weighted (ie. conference membership) factors. For example, which Conference is rated #1, #2 etc. But I have posted this before and have always been blasted for it. But having spent 25+ years as a software tester and PT coder, I think I know a bit about it... bash away...

I don’t think I blasted you in the past, since I agree with the assertion that subjectivity affects the creation of these metrics. Yet they are standardized through their creation. You don’t change the algorithms in accordance with changing opinions. If flawed you create better ones down the road, but you do not subjectivity change them based on the whim or eye test of the moment.

A coach or writer’s poll can be based on such arbitrary things as visibility and the opinion of what matters can change from each poll to the next.

If you want to fault Massey for being subjectively flawed I’m with you. If you are making a case that everything subjective is equally flawed, the evidence contradicts that mightily.

 

[Bone Dog]

The design of every rating scheme is always subjective — you guys are right about that @diggerfoot and @Centerstream — but not in the same way as the dreaded/revered eye test. A rating tool like NET isn’t biased toward a particular team the way eye tests can be. It may be biased toward a style of play or some other vision of what the game might look like. In this respect, it may (for a few years) tend to favor a conference or two. But I imagine these things even out over time, as long as you stick to a particular tool.

 

[Bama fan]

I like turtles.

"Give me a bowl of turtle soup, and make it snappy." I think it is an old Three Stooges thing.

 

[visitingcock]

Sports journalists like to write headlines.

today's AP might have LSU at number two so they can write headlines that say 1 vs 2.

There is one pollster that matters, the Division I Women's Basketball Championship Sport Committee. They choose the best 32 teams selected on an at-large basis. Then there is their task (poll?) of placing teams in bracket slots. Everything else is background noise. Like this post.

AP and coaches polls are a farce. They have not seen all of the teams play. At best they use box scores (stats) for the unseen teams. At worst they use opinions, not unlike most everyone on this board. The computer based polls see the stats (quantitative data) for all of teams' games but have no grasp of the human aspects of team play (qualitative data). Plus humans and their opinions construct these computer systems. It is all imperfect but we like bitch about the polls and computer systems. We like to correct their ratings with our far more accurate opinions. It is fun.

Which is why, of course, I created this thread

 

[Glenn]

For the Men's game, like the focus on Quad 1 wins. But for the women, where the talent drop-off is significant i prefer a finer tooth comb. Moreover, the Quad system is designed for the more marginal teams. The first quadrant goes through away games of teams ranked 75, for example. Here is an alternative breakdown--which does not take into account home/away unfortunately. But has a finer grain.

 

[Bald Husky]

Bracketology should be interesting. I think UConn would hold on to the 1 line, but Stanford should drop to a 2 line replaced by LSU or Iowa.