[Election-Methods] Strategy/polling simulation for simple methods

[Kevin Venzke]

Hi Steve,

--- Steve Eppley <SEppley at alumni.caltech.edu> a écrit :
> What poll question is Kevin asking his simulated voters?  Who is your 
> favorite?  How would you vote if the election were held today?  What is 
> your order of preference?  How much would you pay to elect X if it were 
> a contest between only X and Y? 
> 
> Wouldn't the wording of the poll question, and the allowed answers, make 
> a difference?
> 
> Regards,
> Steve

Sorry if this wasn't clear. Each poll is an election under the given
method. They/I collect data from the polls that is needed according to
whatever best strategy I have been able to formulate for that method.

There is plenty of room to doubt the predictive use of these simulations,
of course. For one thing the voters don't attempt to manipulate polls; they
vote in every poll as if it were the actual race.

I actually did manage to implement a type of Condorcet method, by requiring
voters to vote in each pairwise contest, but not requiring them to rank
transitively. This reduces the ballot to a set of independent binary
decisions. You can easily see favorite betrayal and burial, but I believe
to a huge extent this is caused by "correlation implies causation" errors
by the voters. For each pairwise contest they are comparing (observed)
expectation when each candidate wins that contest.

The methods I implemented were Condorcet//RandomCandidate and MinMax. These
run quickly, and on this computer I seem to be lacking the code for more
sophisticated methods.

Kevin Venzke


      _____________________________________________________________________________ 
Ne gardez plus qu'une seule adresse mail ! Copiez vos mails vers Yahoo! Mail http://mail.yahoo.fr