Mike M:
>I almost let this go by, but it can't be left alone.  Brian, the Union
>analogy is completely inappropriate as far as Croce used it!  To remind
>everyone, he said Union beat Dartmouth which beat RPI which beat BU which
>beat Maine, therefore Union is better than Maine.  Of course, he is so far
>off the mark, it's even worse than the garbage he printed the first time.
 
Actually, he said Union > Dartmouth > RPI > UNH > BU > Maine.
 
>1) If these were the only games played, then yes, this is likely how TCHCR
>  would order the teams (Union higher than Maine).  But these are only four
>  of 426 results right now!  And as the season goes on, things even out,
>  such that if this happened on the first day of the season, it sure isn't
>  going to be how the teams would be ranked now.  That's one of the reasons
>  TCHCR isn't released early in the season.
 
Also, by stating that Maine is the consensus #1 team (flaw: #1 in his poll,
 which
is not relevant when talking about "power ratings"), one can read into his Union
chain example that a computer rating could have Union as the new #1 team in the
country.
 
Yes, Bob's Computer Rating [BCR (^:], perhaps, that only considers 5/426 = 1.2%
of the data. Guess who #2 would be in BCR?
 
>2) As you go farther and farther down the connections, results of a third
>  team have less and less of an effect on a team's rating.  This is
 another.....
 
I can't explain it any better than this, so I won't add anything here.
 
 
>Believe it or not, this very situation is one of the things considered by
>the NC$$ selection committee.  It looks at your opponents, and your
>opponents' opponents, in determining strength of schedule.  And it is very
>likely that in some situations, discrepancies like this will occur among
>teams fighting for, say, the last seed - and the committee has to somehow
>reconcile the differences.  Would Croce like to barbecue the selection
>committee, too?
 
THE COMMITTEE is using the same thing the basketball people have been using
for years. It is called the RATING PERCENTAGE INDEX, and it is a mathematical
combination of winning percentage, opponent's winning percentage, and opponent's
opponent's winning percentage. The RPIndex is a ranking of the 45 Division I
teams and is only a small portion of the data that THE COMMITTEE considers when
doing their poll, and soon enough, when selecting the 12 teams for the tourney.
 
For more information on computer ratings in college basketball (both Sagarin's
and the RPI [as they call it]), I refer you to the 1/20 article in The Sporting
 News
by Andy Katz.
 
And sorry, RPI/Union fans, you would not be any happier with your teams' showing
in the RPIndex.
 
Keith