more feedback on Rob's Condorcet engine web page

[Steve Eppley]

I made a return visit to Rob L's website 

   http://www.eskimo.com/~robla/politics

to examine his recent changes to his Condorcet demo.  It's improved.
Most importantly, the Condorcet result now matches the Condorcet
variation we in EM have advocated: calculating each candidate's score
as its maximum opposition in any of its pair-defeats, and electing
the candidate with the minimum score.

The Condorcet engine is now usable; it's more than a demo.  Groups
which want to use it can prepare two files while offline: a file of
candidates and a file of voters' preference orders.  (The files need
to be in the format acceptible to Rob's engine.)  Then someone can
surf to Rob's engine's front-end page and paste the two files into
the data entry form, press the "submit" button, save the results
page, and forward it to the members of the group.  I've posted this
procedure to the Alliance for Democracy, along with a description of 
the two file formats.

Still, I have some more suggestions:

1. Rob's /politics/ page seems to imply that Condorcet's method is a
form of proportional representation, since it's indented under PR.

2. There's a mistake in the pairwise results page: the description of 
the Condorcet tie-breaker omits that only pair-defeats are included 
when calculating each candidate's largest opposition.

3. I like the new use of color to distinguish pair-losses from
pair-wins, but the use of the same two colors in other elements of
the matrix makes the distinction less clear.  I suggest Rob use
those two colors only to indicate wins and losses (and use a third
color for ties).  One exception: I think the largest loss row should
use the same color as the pair-losses.

Instead of using sky-blue and pink as background colors to
distinguish wins from losses, Rob might want to consider using black
and red as foreground text colors.  The metaphor is financial
accounting, where positives were written in black ink and negatives
in red ink.  A pairing lost by the candidate above would have its 
cell drawn in red, and maybe the largest loss row should be in red 
too.

4. I entered a sample poll where the candidates were A, B, and C.  
The matrix uses the terms A and B in each cell; obviously this can 
be confusing.  Perhaps something more intuitive can be used in each 
cell, such as a left-arrow and an up-arrow.  (I had a different 
suggestion in my previous message about this, but implementing that 
one would involve more labor.)

---Steve     (Steve Eppley    seppley at alumni.caltech.edu)