[EM] Ranking single-winner methods

[Stephane Rouillon]

Sorry I took CR for cumulative voting...

Thus, IMHO:

FPTP < Cumulative Voting < Borda < Approval < CR < IRV < Condorcet(s)

and about the reasons I stated to put Cumulative Voting low, it seems CR
would pass crowding and cloning problems. CR just adds the additional strategy

that IRV avoids. It seems good to me Alex.

I did not want to enter the Condorcet methods details, but maybe you have more

other methods you would add...  Of course, maybe all methods cannot easily be
classed
as better or worse and it stays a subjective matter.  For example, complexity
is not one of the criteria I value very much.  Somebody wanting a simple
method more than anything else
could take an exact reverse order than mine... Several person on this list
said
they prefer approval to IRV.

Still I think chosing the single-winner method should be the first iterative
election any
assembly should do after its foundation. I developped an algorithm to make
sure such a process
would converge. Anyone interested? It uses any starting method and after each
election, uses the
winning method for next election. People vote until only one method serves to
elect itself
or it generates a cycle. Candidates outside the cycle are eleiminated before
next vote. The
overall winning method is the first method of the last cycle.

Steph

Alex Small a écrit :

> Stephane Rouillon said:
> > CR is the last method I included in my rankings.
> > Maybe I do not know everything about it.
>
> CR is simple, in fact, Approval is a special case of CR.
>
> You can give each candidates how ever many points you want on some scale.
> The scale can be 0 to 4, 0 to 10, whatever.  Approval is the simplest case
> where it's 0 to 1.  This isn't ranking, because you can give equal scores.
>  Approval actually requires you to give equal scores if there's more than
> 2 candidates.
>
> Many people say CR is equivalent to Approval because if you want to help
> elect a candidate you should give him as much support as possible (maximum
> score).  If you don't want to help elect a candidate you should give as
> little support as possible (score of zero).  I'm not 100% convinced that
> this is always the best way to go (zero or full score for each candidate,
> nothing in between) but I do agree that it's usually a very good strategy.
>
> Since CR tends to encourage the same type of voting as Approval, and since
> it gives more flexibility to those voters who want to vote in some manner
> other than all-or-nothing (give every candidate either maximum points or
> no points), I rate CR just above Approval.
>
> > Approval and Borda seemed to handle clones better than CR that
> > splits a voter weight in my eye.
>
> No.  Borda is more easily manipulable by clones.  CR is immune to clones.
>
> Maybe you're confusing CR with cumulative voting for single-winner races.
> Say, give each voter N votes to distribute among candidates however he
> wishes.  The candidate with the most votes wins.  CR stands for Cardinal
> Ratings, not cumulative voting.
>
> Alex
>
> ----
> Election-methods mailing list - see http://electorama.com/em for list info