On Wed, 20 Feb 2008, Jorge Arroyo wrote:
The article is interesting, but with such a low number of voters as we get,
is there a method that is non-exploitable?
No method is not-exploitable; unless there's only one voter (and then it's
a dictatorship.)
Is there an article that states the minimum number of voters each method
need to get reliable results?
I'm not sure how to answer this. More is always better, but I'm not
certain what sort of metric you could use to measure "reliability".
Applied to the Icehouse contest what if each judge were to give points
from 1 to 5 to 5 entries, how would it work? Would it be fair?
Depends what you mean by "fair" :)
I feel (again, this is MHO), that the Condorcet criteria is a very
important one for an election method to satisfy; most point-based methods
(and instant runoff voting and plurality voting) don't meet it. The trick
with Condorcet is that it isn't complete: Condorcet alone only tells us,
in this election, that Martian Gunslingers shouldn't win; it can't decide
among the others. It was a VERY close election. Ranked pairs is a
so-called "Condorcet completion" method, a tie-breaker if you will, that
tries to pick its way through the sticky minefield of a Condorcet
"circular ambiguity"; at that point, the only gurantee is that a lot of
voters will be pissed off ;)
--
Dale Sheldon
dales@xxxxxxxxxxxxxxxxx