[EM] simulation graphs

Kevin Venzke stepjak at yahoo.fr
Wed Dec 20 16:51:45 PST 2006


Hi,

Thanks for running VFA, Brian.

I've been reinventing the wheel myself the last few days, so I can
withdraw the suggestion that you should run the code for VFA with poll.

(My version can take up to 10 candidates. I've only been doing 160x160
graphs with under 500 voters, though. And (thus) my sim doesn't produce
very fine boundaries. Today I got bmp exporting working, so perhaps I
should look for a way to post some.)

I come to the conclusion that raphfrk's Approval code does actually
need a fix. I tried to reproduce the zoomed-out graphs, and a couple of
differences I got:

For "Four Corners," rather than a partial swastika, I get an X shape
of solid win regions, with the top, bottom, left, and right triangles
at the borders as tie regions. (By "tie region" I mean a region where
two candidates win about as often.)

For 3A, the blunted green wedge is a shared green-red zone, as is all
the area to the right and above. I don't have a pure green win area at
all.

For 3B, I have the right half of the green region as a blue-green
tie region.


A couple of other neat things I implemented: You can compare two methods'
results. For instance, you can pick TopTwoRunoff as the "base" method,
and then watch the FPP plot. Where the two methods agree, the pixel is
dark gray. Where they don't agree, FPP's choice is colored normally.

I also implemented the ability to compare two methods' Condorcet
efficiency, which was the first thing I was asking myself when two 
methods disagreed. For instance, you could compare zero-info Approval 
and IRV, and see one color where Approval is agreeing with Condorcet 
against IRV, and another where IRV agrees with Condorcet against 
zero-info Approval. (Everything else is gray.)

I tried out another zero-info approval strategy that makes sense in
this setting... "Approve every candidate nearer to yourself than the
center of the smallest rectangle that contains all the candidates."
This would be more resistant to the introduction of multiple similar
candidates.

Kevin Venzke


	

	
		
___________________________________________________________________________ 
Yahoo! Mail réinvente le mail ! Découvrez le nouveau Yahoo! Mail et son interface révolutionnaire.
http://fr.mail.yahoo.com



More information about the Election-Methods mailing list