FAQ (was: Let me introduce myself.)

Mike Ositoff ntk at netcom.com
Sun Oct 4 15:14:03 PDT 1998

> BTW (Mike or anyone else), how exactly does VA differ from the other
> (pairwise) methods?  I had believed that this referred to a tie-breaking
> method, and seemed to be self-explanatory, but from the way it's being
> touted lately I must be wrong about that.

Among the circular tie solutions for pairwise methods are the
versions of Condorcet's method, which compare defeats. Since
Condorcet didn't specify how to measure defeats, 2 measures
are currently proposed, VA & Margins. Other differences in
counting procedure distinguish Condorcet(EM), Smith//Condorcet(EM),
Schulze, etc. So how to measure defeats is one of several
variables, but the most controversial one, because it 
makes all the difference in properties.

Of course the votes-against measure & the margins measure can
be used other than with Condorcet versions too. For instance,
Simpson-Kramer looks at victories (& ties) as well as defeats
(The journal description of that doesn't specify a measure
for defeats either). And margins are used differently in
Young's method, according to one use of that term: It counts
_overall_ margins of defeat, the number of individual pairwise
preferences that would have to be ignored in order to make
a particular alternative unbeaten. Dodgson is similar, but
its definition usually speaks of reversing rather than ignoring
individual pairwise preferences.

Then there are completely different circular tie solutions,
like the Hallett system, said to be equivalent to solving
circular ties by IRO.

When I speak of VA or Margins, I mean any of our Condorcet
procedures, which look at defeats, measuring them by votes-against
or by margins of defeat. I'm fairly sure that's how others are
using the terms.


More information about the Election-Methods mailing list