-- [ From: Kepler * EMC.Ver #2.5.02 ] --
 
John notes:
 
>  At any rate, HEAL also has the
> possibly undesirable feature of punishing teams which play strong
schedules, in
> that a loss to any team has the same effect on your rating.  (In KRACH,
the
> strength of the team you lost to is a serious mitigating factor:  losing
to a
> team ranked far ahead of you was pretty likely given the other information
> available, so it doesn't tell the system anything it didn't already know.)
 
I would opt for a system which produced the following reward spectrum:
 
Beat great team > Beat awful team > Lose to great team > Lose to awful team.
 
(Ties?, you ask.  Um, yes, well... Arguably, a win is always worth more than
a tie.  Arguably, tying a 29-0-1 team counts for more, not less, than
beating a 0-29-1 team.  Lots of things are arguable.  Tying a great team >
tying an awful team, at least.)
 
But underlying all of this: having three or four (or five, or six...)
systems with somewhat different methodology, and yielding somewhat different
results, is not a bad thing.  Nobody would argue that statistical measures
are an infallible predictor of the future -- or even a complete summary of
the past.  And I can't believe I'm the only person who dutifully surveys all
of the measures this week, picks out my team's ranking in each, and ten
minutes later recalls the one which puts them highest.  After all, that's
the "best" measure, of course.  ;-)
 
Here as elsewhere, diversity is its own reward.  Good luck to everyone in
the post-season.
 
-- Greg
 
 
 
 
 
--
 
*  *  *  *  *  *  *  *  *  *  *  *  *  *  *  *
*  Greg Berge
*  Portland, Oregon
*  [log in to unmask]
*  www.spiritone.com/~kepler
*
 
HOCKEY-L is for discussion of college ice hockey;  send information to
[log in to unmask], The College Hockey Information List.