The recent posts about Mass-Lowell got me thinking that I'd also put too
much weight on their beatings from RPI and BU. When Lowell did the same
to Brown (10-1),I didn't raise Lowell but lowered Brown. Lop-sided scores
and unexpected wins/losses get our attention too much. Recent games also
get too much weight while we forget earlier results. The RPICH ratings
counter-balance these natural inclinations with some good hard numbers.
The RPICH that was recently posted gives some good examples:
 
> 3  New Hampshire       25   16- 8- 1   .6600   .5537   .5074   .5564
> 4  Minnesota           26   15- 8- 3   .6346   .5435   .5163   .5508
> 5  Northeastern        24   15- 6- 3   .6875   .5026   .5276   .5496
> 6  Northern Michigan   26   16- 9- 1   .6346   .5430   .5129   .5493
> 7  Mass Lowell         24   14- 5- 5   .6875   .5099   .5038   .5430
 
Of these teams, only Northern Michigan has gotten comparable ratings
in the polls, and I'm willing to bet that the polls will overreact in
lowering NMU for their recent losses. Minnesota will probably get some
recognition for some big wins in this week's poll, but not that high.
Now what about those HE schools? I've been consistently voting UNH higher
in my hockey-l poll vote than the school has placed. Well, I started to
think, "maybe I'm just overrating them because they always play BU tough"
Looking at RPICH, I'm going right back to those high ratings for UNH.
I also tended to rate Northeastern higher than others, but then with 3
Hockey East teams high in my ratings (BU,UNH,NU) I think I've failed
to give Mass-Lowell their due (and those bad losses just gave me a good
excuse.) With an RPICH rank of 7, I think Mass-Lowell deserves a better
look. How about it Hockey-L?
 
Oh, and it can go the other way. How about those ECAC teams?
 
>16  RPI                 19   12- 5- 2   .6842   .4600   .4852   .5149
>19  Harvard             16   11- 3- 2   .7500   .4310   .4791   .5140
 
What is the lowest Harvard's been rated in the polls? About 7?
    Dave Carroll, BU '73
    [log in to unmask]     (or if path problems ...
    [log in to unmask]