[EM] Range Voting "unbeatable"?

Warren Smith warren.wds at gmail.com
Mon Aug 31 06:10:22 PDT 2009


>What you could do is take a "poll" and have 10 random voters.  You
then work out optimal assuming that they are the electorate.

--there is no such thing as "optimal strategy" in games with >=3
players. Game theory breaks down.  So, in general, this cannot be
done.  The only way to do it is to add to game theory some other
ingredient, such as some model of how the other voters act
(which will thenbe, in fact, false, since it isn't the way YOU are acting!).

--------

Also, to reply to (yet another) confused claim by Jameson Quinn, he
follwoing K.Venzke had the wrong notion that
(i) range voting was "absolutely unbeatable" with honest voters using
Bayesian Regret as yardstick
(ii) which is false, and a counterexample is the BRBH voting system in my paper
(iii) so then JQ replied this was only under some unrealistic model
(called RNEM in this case)...
(iv)  which completely missed the point that of course, under pretty
much any other model, some other voting system would have beaten range
voting's Bayesian Regret.   The point of RNEM model was not its
realism (which was poor to middling), but rather that it was
sufficiently simple that you could work a lot of things (such as the
BRBH voting system, and its Regret) out as explicit formulas.



On 8/31/09, Raph Frank <raphfrk at gmail.com> wrote:
> On Mon, Aug 31, 2009 at 7:07 AM, Kristofer
> Munsterhjelm<km-elmet at broadpark.no> wrote:
>> Could your third point be done, for very small electorates, by use of
>> minimax game tree algorithms like the one used in computer chess? The
>> objective for each voter would be to get his own candidate to win (and
>> for
>> rated methods, to have the winner maximize his individual utility).
>> Minimax
>> requires perfect information, so that's a flaw, but it should give a
>> bound,
>> as it were, because the voters can't know more than perfect information,
>> only less.
>
> What you could do is take a "poll" and have 10 random voters.  You
> then work out optimal assuming that they are the electorate.
>
> The actual election would then be slightly different as you include
> all of the voters.
>
> However, can minimax be applied in a single step "game"?
>


-- 
Warren D. Smith
http://RangeVoting.org  <-- add your endorsement (by clicking
"endorse" as 1st step)
and
math.temple.edu/~wds/homepage/works.html



More information about the Election-Methods mailing list