[EM] Losing Votes, Improved and unimproved Condorcet

Michael Ossipoff email9648742 at gmail.com
Sat Dec 15 19:13:00 PST 2012

I don't know the properties or criterion-compliances of the
losing-votes proposals. They, and the practice of choosing among the
IC-unbeaten candidates by ordinary pairwise count (disregarding equal
ranking), appear to have a good chance of maybe being better than ICT
and Symmetrical ICT, but I don't know. With only an initial look at
those new proposals, it isn't really possible to say anything.

Chris's example:

35 A
10 A=B
30 B>C
25 C

A>B 35-30 (ignoring the 10 A=B ballots unlike my proposal, according
to which A>B 45-40)
B>C 40-25
C>A 55-45

Symmetrical ICT chooses A.

By Symmetrical ICT's "beat" definition, all three candidates
cyclically beat eachother, and A wins by top-count.

The top-count isn't a perfect solution--when it loses
clone-independence, for instance. But it's a simple and natural way to
avoid chicken dilemma: When everyone is beaten, all below-favorite
support is withdrawn, leaving defectors without support from other

I don't know the properties or criterion-compliances of the new
methods that I mentioned above.

Symmetrical ICT meets FBC, CD, legitimate Condorcet Criterion, and
0-info Later-No-Help, and is genuinely without the chicken dilemma.

Kristofer spoke of going from IRV to (presumably traditional
unimproved) Condorcet, or going directly to Condorcet.  --the
presumption being that traditional unimproved Condorcet is already
agreed to be the final destination.

Going from IRV to traditional unimproved Condorcet wouldn't be any
significant improvement.

Stick with Approval or Score. In addition of FBC, Later-No-Help,
Participation, Consistency and IIAC, they have the further advantage
of being enactable, and feasibly handcountable.

Mike Ossipoff

More information about the Election-Methods mailing list