[EM] Strategic Voting and Simulating It

[Raph Frank]

On Fri, Oct 17, 2008 at 1:32 AM, Kevin Venzke <stepjak at yahoo.fr> wrote:
> Hi Raph,
>
> --- En date de : Jeu 16.10.08, Raph Frank <raphfrk at gmail.com> a écrit :
>> I think Warren has some simulations where the voters are
>> spread over
>> 1-2 axes.  (He can comment).  My understanding is that
>> there are lots
>> of different distributions.
>
> Ok. That is better. But you still have the problem that it's open to
> endless debate, what exactly the realistic simulation method is.

Right, Warren's proposal is to try lots of different variations.  A
method that scores best under lots of different assumptions is likely
to be best.

Ofc, even then, he may not have covered enough search space.

> But this ignores the fact that parties still want to try to win the
> election. If they back candidates at random, they could conceivably hold
> on to frontrunner positions, but they wouldn't generally win, so they
> don't do this.

One option here is to do what parties actually do and hold a plurality primary.

In fact, lots of different primaries election types could be tested.

> It is still a problem to take this interpretation of FPP as a starting
> principle to measure *all* rank ballot methods.

The advantage is that it can be easily applied to voters with random utility.

It automatically splits them into 2 groups.

> I am not sure I've seen the other thread but I'll look for it.

It was the suggestion that you pick 2 candidates as the top 2 and then
test if it is stable by allowing each voter to change his vote one at
a time.

> Perhaps... I've never written a simulation to study nomination incentive
> specifically, but I have written e.g. a FPP simulation, in which
> voters stop voting for a candidate (in the polls leading up to the
> election) when the calculated benefit to the vote disappears. And in
> FPP there is no way for the benefit to come back (in contrast to, say,
> Approval, which in my simulations of the same sort had the potential to
> never arrive at stability).

If there isn't a condorcet winner, then you get instability.  I
remember running sims on the "Rank your favourite of the top 2 and all
you like better" strategy and it is unstable, if there isn't a clear
condorcet winner.  This is a representation of approval's condorcet
seeking behaviour.