Other problems at Rob Lanphier's Condorcet website

[Rob Lanphier]

On Wed, 18 Dec 1996, Steve Eppley wrote:
> I tried two more examples at Rob's Condorcet Demo site:
> 
> The demo claimed the result of
>    46:1
>    20:2
>    34:3>2>1
> is a two-way tie since two candidates have a largest opposition of 46.  
> 
> But that's not the same method as the Condorcet we've defined here,
> which counts only opposition in pair *defeats*.  Candidate #2's "46 
> opposed" were in a pairwin and should not be counted.  The method Rob 
> has implemented was discussed in EM a little.

This problem will be fixed shortly.  However, I was a bit unclear as to
how ties should be handled.  What I'm about to submit for review only
counts strict losses (as per your out-of-band email to me).

I've also incorporated some suggestions from your other email. 

*  I've added some gawdy looking colors that do help in the visual scan.
   I hope they are intuitive because I don't want to go through the
   trouble of explaining them :)
*  I've changed the caption to read "Pairwise Election Results", as this
   program has grown beyond Condorcet's method, and make it clear when I'm
   talking about Condorcet's method.  
*  I've added a summary line which describes worst pairwise defeat.  I
   still include the win-loss-tie summaries, because these are useful
   in seeing which candidate is undefeated, and for evaluating Copeland's
   method.

These changes were all fairly easy, so I did them at the same time.

The Eskimo CGI script reviewer is going to have to review the script
before you actually see it, though, so it may be a while before it
actually shows up on the site.

> The demo correctly showed all three candidates are in the Smith set,
> but it doesn't go the extra step and show the Smith//Condorcet 
> winner.  That would be a useful addition to the demo, and easy to do.

Easier said than done, though that is in the works.  I'd like to make
arbitrary combinations of methods possible.  I could hack together a
Smith//Condorcet solver, but I'd like to handle the more general case so
that I (or others) could implement the atomic method (Smith, Condorcet,
etc), and then the user of the program can chain them in an arbitrary
order.

> The other example I tried was:
>    46:1
>    20:2        joker
>    34:3>2>1
> 
> The 20 ballots with the "joker" characters were invalidated. I'd
> appreciate it if Rob would modify his code so that extra characters
> at the end of a ballot won't invalidate it.  This would allow users
> to append names or handles to their ballots, which would help voters
> verify their ballots were correctly entered. 

Actually, the comment character is "#" (ala shell scripts and Perl), so
that:
46:1
20:2        # joker
34:3>2>1

...is completely valid input.

I hope this helps.

Rob
---
Rob Lanphier
robla at eskimo.com
http://www.eskimo.com/~robla