[EM] simulation graphs, zoomed out

[Brian Olson]

The images on your site are compelling and interesting.

http://www.geocities.com/raphfrk/

I've added random tiebreaking to ApprovalNoInfo in my own code and I'm 
re-running those images. With a little luck my computer will run them by 
morning.

I like how it shows regions where the outcome becomes a random pick 
between two tied candidates, each sufficiently approved of by the 
population. They're being generated directly into the web directory and 
will show up as they complete.

I don't think most other methods generally need such tie-breakers since 
they are less likely to tie given rankings or ratings ballot input. On the 
other hand I suppose I should add that to all the simulated methods out of 
some sense of completeness.

Would it be a reasonable adjustment to the zero info strategy that when 
there an odd number of choices, vote with a 50% chance on a voter's middle 
choice? If the strategy is to vote yes for the top half and no for the 
bottom half, then it makes sense to try and split the middle choice in 
half that way.

Brian Olson
http://bolson.org/

On Wed, 20 Dec 2006, raphfrk at netscape.net wrote:

> I slightly modified the ApprovalNoInfo code and re-ran.
>
> Basically, I added:
> if ( m2 == m && random()%2 == 0 )
> {
> winner=second_place;
> }
>
> This just randomly swaps the winner with the 2nd place candidate
> if they tie rather then consistantly going one way. This gives
> a more balanced result than the previous run.
>
> I added it to a geocities webpage.
>
> www.geocities.com/raphfrk
>
> I only ran 2 sims and only with 50 voters, so there is alot
> of inherent noise. However, some of the 'noise' is just
> blending of 2 edges due to the ties. The result is
> symmetric unlike the previous case.
>
>  Raphfrk
> --------------------
> Interesting site
> "what if anyone could modify the laws"
>
> www.wikocracy.com