I've seen it in the College Hockey News.  Today, it was recited by the
Maine SID as the "Computer Ranking out of Bowling Green" to a Friends of
Maine Hockey Luncheon.  Keep it up Keith!
 
Keith, I've a question for you ...
 
You normalize the rating so that the top team is 100 and bottom 0. This
has the unfortunate side-effects of masking how well the top and bottom
teams are doing relative to anything (such as last week or last year or
...).  Another effect is to make numeric comparisons between various
rating weeks meaningless.  I.e., "We were 7 points behind 3rd place last
week, but only 5 points behind this week".
 
I've had this question for a long time, but have not asked about it
since I don't really know a solution.  Would it be (non-optimally)
possible to calculate the effects of the first and last place teams win
and lose (respectively) all games and do a normalizing calculation on
that.  Now that I "hear" this suggestion it becomes evident that it is
too complicated and there may not be a numerical way to do it.
 
... a second question ...
 
Is the program and database for some machine type available for those
wanting to experiment (with the data or algorithms)?  This is *not* a
request for you to do anything to make them available, Keith!