[BostonCard]

Not to speak for CTCard, but in the past he has raised two objections, both of which are reasonable:

1) Offering a "statistical" analysis based on no discernible data, since no team has played yet this year.  There are, of course, some correlations from season to season, but that doesn't account for departed or new players.  As I understand it, some analyses look at the number of returning players, and some look at recruiting rankings (though those rankings are based on subjective assessments of recruits).

2) Black-box analyses where the methodology is not disclosed.  If one doesn't know the method, or even the variables used as input, there is no way to determine what the potential flaws are.

BC

EDIT: As best I can tell, CTCard does not object to the use of statistical analyses, provided that a) they are based on real data, and b) they are transparent about methodology.

[BostonCard]

Not to speak for CTCard, but in the past he has raised two objections, both of which are reasonable:

1) Offering a "statistical" analysis based on no discernible data, since no team has played yet this year.  There are, of course, some correlations from season to season, but that doesn't account for departed or new players.  As I understand it, some analyses look at the number of returning players, and some look at recruiting rankings (though those rankings are based on subjective assessments of recruits).

2) Black-box analyses where the methodology is not disclosed.  If one doesn't know the method, or even the variables used as input, there is no way to determine what the potential flaws are.

BC

EDIT: As best I can tell, CTCard does not object to the use of statistical analyses, provided that a) they are based on real data, and b) they are transparent about methodology.

[washingtonismoney]

They are reasonable. But I think approaches that blend different ratings can shake out some of the uncorrelated error there. So if the statistical models say Stanford is in the 10-20 range, I think that's helpful and somewhat worthwhile.

[washingtonismoney]

They are reasonable. But I think approaches that blend different ratings can shake out some of the uncorrelated error there. So if the statistical models say Stanford is in the 10-20 range, I think that's helpful and somewhat worthwhile.

[CTcard]

Not to speak for CTCard ...

You speak well. Perhaps I can hire you as my official internet spokesman. Well, as long as I don't have to pay anything in real money. How about payment in upvotes when the promised, long-dreamed-of new software comes through? 

But I think approaches that blend different ratings can shake out some of the uncorrelated error there.

How do you know the blending doesn't just amplify the inherent noise?
Why should the blending be restricted to "statistical ratings" rather than judgmental ratings?
-------

BC got my basic concern down pat.

I'm certainly all for looking at data - and I think my posts here reflect that. I am all for statistical analysis of that data. But, I am against what I consider misuse of statistics, one of which is believing statistics can create information where it doesn't exist in the input data. 

It's misuse of statistics that causes so many people to go on about "lies, damn lies, and statistics" or that "you can prove anything with statistics". Combating such nonsense is best done by combating misuse of statistics - or at least being careful in noting what they can and cannot tell you.

[CTcard]

Not to speak for CTCard ...

You speak well. Perhaps I can hire you as my official internet spokesman. Well, as long as I don't have to pay anything in real money. How about payment in upvotes when the promised, long-dreamed-of new software comes through? 

But I think approaches that blend different ratings can shake out some of the uncorrelated error there.

How do you know the blending doesn't just amplify the inherent noise?
Why should the blending be restricted to "statistical ratings" rather than judgmental ratings?
-------

BC got my basic concern down pat.

I'm certainly all for looking at data - and I think my posts here reflect that. I am all for statistical analysis of that data. But, I am against what I consider misuse of statistics, one of which is believing statistics can create information where it doesn't exist in the input data. 

It's misuse of statistics that causes so many people to go on about "lies, damn lies, and statistics" or that "you can prove anything with statistics". Combating such nonsense is best done by combating misuse of statistics - or at least being careful in noting what they can and cannot tell you.