Thanks, Charlie, for typing bits of this in. I've been waiting to
hear what the Colorado press had to say.
 
But I need to correct some things. I will try to stick to the facts,
so I don't piss too many people off this time.
 
>It was a terrible mistake.  CC had 23 wins and was ranked fifth
>nationally in the last computer-based power rankings.
 
Hmm, they must be quoting The College Hockey Computer Rating. I know
the CC press has seen it. Maybe Charlie deleted it, but I hope they
distinguished between TCHCR and RPI in their article. That's why I have
that disclaimer at the bottom of TCHCR every week.
 
>Lucia believes the rating index, while unbiased, was incorrectly
>programmed.
 
Poor choice of words. It sounds like Lucia is blaming the programmer,
who thought the commitee said "50,25,25" but s/he programmed in "25,50,25".
Oops. But I think the reporter just stumbled on the wording.
 
>It was based on 25% performance, 50% strength of schedule
>and 25% opponents' strength of schedule.
>
>"So in effect, the schedule was 75%," Lucia said.
 
That is not entirely correct. You have to look at the range of the
three numbers. From Erik's latest (25/50/25) RPICH, the range of
Winning Percentage goes from .8375 to .0667 (.4444 is the low among the
top 20); Opp Win % from .5686 to .3593 (.4714 in the top 20); OppOpp
Win % from .5280 to .4488 (.4828 in the top 20).
 
Winning Percentage is highly variable--a difference of 4 or more
percentage points might be considered the margin of significance.
OppWin% is bunched up near .5000, so maybe only a 1 percentage point
difference is significant.  And so on for OppOppWin%. You can't just
look at the weights and conclude that schedule made up 75% of your
rating; you have to put the numbers in context.
 
And some background information: the NCAA manual says that selection
for *all sports* is supposed to be based on performance and strength of
schedule, with both weighted equally. So, the NCAA is trying to get a
50/50 mix with its RPI.
 
Something that they didn't mention (and my main beef with the RPI), is
that OWP and OOWP are not very good ways to factor in strength of
schedule.  Too bad, cuz you can make some good arguments along that
line. Playing with the weights will only gain you so much.
 
>For instance, CC faced Michigan Tech seven times.  Tech finised last
>in the WCHA, but the computer wasn't equipped to realize Tech lost
>many one-goal games and gave up the second fewest goals in the lrague.
 
Hmm, BG gave up 3 goals to WMU in the last 2 minutes to lose a game. I
think the computer should be equipped to handle that. (^; Also note
that Tech scored the fewest goals in the league, 93. I am not sure what
giving up few goals has to do with anything, but I won't pursue it
further.
 
>Either plan would work far better than the present method which
>assures only one berth per league.
 
Sorry, but each league is guaranteed *2* spots in the tourney.
 
Keith