This is regarding the "Steve G performance-indicated computer rating
system".
 
The marvelously lucid Ralph N. Baer asked me a couple of good questions.  In
answering him, I noticed a flaw (failure to re-normalize) in my formula for
"Y".  I promised Ralph I'd investigate it in the evening, and get back.
Last night I discovered, and today Ralph confirmed, that I'd reinvented RPI
with 50-25-25 weightings (plus the difference that RPI doesn't include the
results of games vs opp when figuring opp win %, but mine does).
 
Ralph reminded me of a thread that ran about a year ago, and even sent me
some of his posts (either he archives religiously, or is a pack rat that
would give any sysadmin with server space limitations nightmares!).  I need
to do a little more cerebral work on the properness of including or
excluding games vs opp when figuring opp win %, and investigate the
convergence of opp(opp(opp(.....)win % around 0.500 (meaning, at what point
is is too close to be worth one more 'opp'?).  I'll get back...
 
Steve "now with a Christmas project" G
 
HOCKEY-L is for discussion of college ice hockey;  send information to
[log in to unmask], The College Hockey Information List.