[EM] Sims - method terminology

[Kevin Venzke]

Hi Jameson,

--- En date de : Ven 11.6.10, Jameson Quinn <jameson.quinn at gmail.com> a écrit :
>Anyway, however you go, I really have seen that "wait until it's done 
>before I share" is not the right attitude; even one helping hand makes 
>the trivial effort of sharing more than worth it.

I'm not totally sure what you're referring to. The program itself is
fairly simple; the tricky thing is deciding what it should do in the first
place, and what results should be gathered from it.

The main reason I would want to share the source is to allow others to
check for bugs or look for assumptions never stated. I haven't finished
checking for bugs myself (I'm slightly incredulous at the poor performance
of MMPO for example).

--------------------

[begin quote:]

Second, one comment on your method descriptions: 
> > Method  BestC   WorstC  Top     Middle  Bottom
>  Dist    DistN
> > CdlASnc 91.8%   1.1%    87.3%   10.4%   2.2%
>  52.978  1.306

Conditional Approval (sincere). This is an obscure method for which I
recently started using the term CdlA. This is intended to be a three-slot
method where the votes are counted repeatedly and voters add in their
second-slot preferences when the leader after any previous round was a
last-slot (i.e. non-)preference.

Sincerely voting this isn't very realistic and I would've just deleted
its results, except that it placed first here...



I'd thought of this method, and consider it a very good one if you're willing to go with a non-summable, computationally-complex method. I am surprised to hear you say that "sincerely voting this isn't very realistic". True, it is not LNH, so truncation is possible; but I don't see truncation as a dominant strategy here. For instance, if everyone else truncates, sincerely voting can only improve your outcome. In fact, the result if your bloc is the only sincere one is actually probably better for you than the result if your bloc is the only truncators.*


So: if you expect others to truncate, you are sincere; if you expect others to be sincere, your rational vote is to truncate. But since you actually prefer the results of the first case to those of the second, your rational mixed strategy is mostly sincere*. (And since pure strategies are unstable*, I believe that this mixed strategy will be a unique strong Nash equilibrium*, with no need for trembling-hand fiddles to make it unique).


*Much of the above is purely intuitive, so if you have a different impression, I'd love to hear it. (The parts without asterisks are intuitive too, but I have sketchy proofs in my head, so I'm a little more confident of those parts).

[end quote]

My impression is that in effect, despite the unusual algorithm, CdlA
is very similar to some Condorcet methods. My DNA generator thinks CdlA
actually is a Condorcet method, iirc.

Since CdlA isn't monotonic I think its main advantage is the ease of
explanation.

I believe you will have a perfectly typical truncation dilemma when your
side has two near-clones and you expect a candidate on the other side to
be the initial leader.

--------------------

By the way, it has occurred to me that I can attempt to determine what
is a "realistic" set of candidate placements by excluding scenarios that
are not competitive. Unfortunately I'm not producing this data yet.

There is though the possibility that some scenarios are competitive but
not realistic...

Kevin