[EM] Realistic strategy questions

[Kevin Venzke]

Hi Jameson,

_______________________________
De : Jameson Quinn <jameson.quinn at gmail.com>
À : Kevin Venzke <stepjak at yahoo.fr>
Cc : em <election-methods at electorama.com>
Envoyé le : Mardi 25 février 2014 15h17
Objet : Re: [EM] Realistic strategy questions

>>
>>Jameson, does your strategy #3 "fully strategic" mean Warren-style
>>"use favorite betrayal and burial no matter what"? Or is it more of an
>>approval-style "everybody gets an A or an F" thing, under the methods
>>that allow it at least?
>
>In strictly ranked Condorcet, it means FB and burial. In non-strict
>ranking, I think it should be just FB (for margins) or just turkey-raising
>(for WV), that is, compression-to-ties on one side, and inversion on
>the other. But I'm open to suggestions on that count.

Not sure I follow your ideas there. I think the stereotypical "bad" case
for WV is lots of compression with lots of truncation. (I have also seen
the criticism that WV has a random fill incentive, rather than truncation
incentive, but I think it's doubtful that a margins advocate, at least,
would choose that particular attack.)

To my mind the stereotypical margins strategy is Warren-style voting. But
that's definitely a caricature. I would not necessarily vote like that.

There are also those who think all Condorcet methods would see lots
of burial all the time. Or favorite betrayal all the time. I don't envy
you the task of having to select which extreme strategies deserve to be
tested!

>
>>
>>Does your polling system do a round of Range voting, or something like that?
>>
>
>A round of honest votes, using the same system. In some scenarios, those
>honest results are then subjected to a media bias which disfavors arbitrary
>"non-major" candidates according to one of two soft scales.

Hmm, so all supported methods need to be able to pick a second place finisher.

I could see that polling in this way could really hurt some methods' scores,
when the original sincere outcome is very far from "ideally strategized."
Maybe it would be easy enough and of enough interest to try also with using
raw utilities (or a sincere Condorcet method, which I'd think would be more
realistic) for the polls, to simulate a common means of voters' evolving
awareness of the frontrunners.

The "frontrunners are completely arbitrary" scenario is also amusing and
informative, and it sounds like you're planning to try something like that.

>>
>>My concern would of course be the possibility of supposing behavior that
>>doesn't happen to make sense in a specific scenario. I see that you have some
>>conditions to try to avoid this. But naturally if "strategic" voters
>>strategize poorly (either too much or too little, or not in the right way)
>>or "unevenly" poorly (among the various methods and scenarios) this could
>>bias a method's final score (either positively or negatively!).
>
>Right. My goal is to find some set of models (each of which is a probability 
>measure over scenarios) that I can call "realistic", and explore relative system 
>quality under such models. So I'm not aiming for ideal strategy, except insofar 
>as there are significant numbers of real-world voters who strategize ideally. 

But that's the tricky thing isn't it? It's very hard to say in advance or in
general what it means to strategize ideally under a given method. You could
completely nail it with your strategic Range implementation, but then have only
a murky comparison with any method you can't nail down quite as well.

I'm just noting the difficulty of finding the apples to apples comparison.

Kevin