[EM] range voting - better than Condorcet for purpose of generating Condocet winners...

[David Cary]

--- Warren Smith <wds at math.temple.edu> wrote:

> concerning what happens with an honest+strategic voter mix under
> range voting:
> 
> http://rangevoting.org/StratHonMix.html
> 
> the interesting findings include
> 1. despite the worry that strategic voters occurring in ways
> correlated with politics might penalize the honest voters, 
> the sim shows that range is still top of the heap versus 
> the other voting methods even in situations designed to 
> exhibit this effect.
> 

Based an my review of the simulation program at:
    http://rangevoting.org/IEVS/IEVS.c
(the version I downloaded was internally identified as version 2.58),
it appears that the strategic voters in the simulation are uninformed
in several regards:
1.  They don't know anything about the preferences or voting of the
other voters,
2.  They use suboptimal "zero-information" strategies, and
3.  The strategies are tailored to the form of voting, but not
necessarily each particular voting method.

So while an "honest" Approval voter will approve candidates that have
higher than average utilities, a "strategic" Approval voter can
invert preferences.  As a result, "strategic" Approval voting, and
even fully "strategic" Range voting have higher simulated Bayesian
Regret than "honest" Approval voting.

Similar suboptimal strategies seem to be used for the ordinally
ranked ballot methods.

Studying "zero-information" strategic voters, and maybe even
suboptimal strategies, may have some value, but more care should be
given in making broad statements about election methods based on such
peculiar scenarios.  

-- David Cary



 
____________________________________________________________________________________
Bored stiff? Loosen up... 
Download and play hundreds of games for free on Yahoo! Games.
http://games.yahoo.com/games/front