MitchalH wrote, in part..
 
>   Also, while KRACH is apparently always the best at matching with past
> results (those from which the rankings are derived), I have found in the
> last year that the CCHP has always been better at predicting future results.
>
>   Does anyone know why that would be true?
 
I'll take it a step further.  Some mathematicians would say that the best
ranking scheme is one that describes the data (past results) best.  I
say there are other possibilities that may or may not match that criteria
 
Which team has had a "better" season?
 
Which team is "best" going into the playoffs?
 
Which team will win a H2H matchup? How confident of the prediction
are you? By how much?
 
Should a ranking system be "good" at the top, middle and bottom of
the rankings, or is just the top important?
 
Given a division of top and bottom teams, how important are games
with bottom teams or top teams, wrt the ranking?
 
If we can agree that our ranking is a "tournament selection" ranking of
the best teams at the moment of selection, then I'll offer that RPI
(flawed as it may be) tweaked by PWR (with more and bigger warts of
its own) may be a better selector than any of the other systems shown
here.
 
On a related note, some of the conferences have a post-season
consolation game.  How should this game affect the rankings? It's
normally between two top teams, so the winner should increase
strength of schedule and rating?  It is a game very close to (NCAA)
tournament selection, so should be important? But it's not for 1st place
and it's between teams that have probably already played several
times, so not necessarily a real measure of the teams involved wrt
everyone else in college hockey (IMHO).
 
A better measure might involve "How does each ranking system fair
for the last 10 games of each team?", yet the strength of schedule for
these games will have varied widely ... so widely so as to make the
results meaningless, IMHO.
 
Anyway, these have been interesting discussions.  My last humble
opinion for this e-mail is: everyone is almost right and everyone is
almost wrong.  All of the solutions almost solve the problem. No one
understands the problem.  :-)
 
Good luck to all still playing, wayne
 
 
Wayne T. Smith                             mailto:[log in to unmask]
Systems Group - UNET                 University of Maine System
Co-owner of the College Hockey mailing lists
 
HOCKEY-L is for discussion of college ice hockey;  send information to
[log in to unmask], The College Hockey Information List.