[EM] attempt at exhaustive computer sim of 3-candt left-middle-right scenario

Warren Smith wds at math.temple.edu
Sat Jul 8 13:47:57 PDT 2006


A PRELIMINARY draft paper, which attempts an exhaustive simulation of
left-middle-right 3-candidate election scenarios, is 
   http://math.temple.edu/~wds/homepage/lcr.pdf
It is only 3 pages long.

The reason it is a "preliminary" draft is, I need to recheck the computer program
for bugs.

It compares a lot of different voting methods.
All the different Condorcet flavors all become the same method in the 3-candidate 1-dimensional
case (which is all this paper considers) and similarly a lot of
other method-variants also become the same, so hopefully you can't complain to me
that I have not considered your favorite method, because I have.

The comparison includes the recent "asset voting", and "IRV with optional dropout" ideas
as well as a scenario with a 50-50 mix of near-honest and strategic range voters.

Anyhow, among the methods tried, (near-honest) RANGE and the 50-50 honest-strategic
range mix, clearly outperformed all other methods, but like I say this is a PRELIMINARY
conclusion.

You may wish to comment on the paper.  To do so email warren.wds &at^ gmail.com.
You may also wish to confirm it yourself by writing your own program.
As I describe in the paper, everything is deterministic and done in exact
arithmetic with no roundoff errors.
Therefore, if you write your own program to perform the same calculation, you ought to get
EXACTLY the same results.  That would in fact be a very good and desirable check
for the purpose of spotting bugs in (either) of our programs.

I will repost at some later point when and if I get confiedent enough to anoint this as
a "final" draft, so caveat emptor for now.

Warren D. Smith
http://www.rangevoting.org



More information about the Election-Methods mailing list