[EM] [Condorcet] Re: IRV vs Condorcet

[robert bristow-johnson]

On May 27, 2010, at 1:09 PM, Jameson Quinn wrote:

>
>
>
> now ask yourself the question whether or not Condorcet satisfies  
> these criteria (assuming a CW exists).
>
>
>
> Of course it does, because you only included the anti-strategy  
> criteria which it does pass.

they're more like basic principles of a fair and predictable (in  
behavior) election system.  they're almost axiomatic.  if one were to  
take issue with any one of those 6 principles,

> But what do you call this:
>
> Hypothetical true preferences:
> 39.4: D>H=C
> 30.8: H>C>D
> 27.6: C>H>D
> 2.5: NOTA
>
> H is the condorcet winner. But if just 8% from the C voters instead  
> vote C>H=D (or 4% of them vote C>D>H, if equal rankings aren't  
> allowed), then there is no Condorcet winner, and C could win.

so, you're saying that if 4% of the voters bump their preference of H  
down, then H no longer wins outright.  that might be expected.

i'm gonna scale this to integer quantities (dunno what to do with NOTA):

     394: D>H=C
     308: H>C>D
     276: C>H>D

     the "NxN" matrix in my preferred triangular form:

                H>D  584
                D>H  394

                H>C  308     C>D  584
                C>H  276     D>C  394


now 80 C voters hypothetically change their mind to:

     394: D>H=C
     308: H>C>D
     196: C>H>D
      80: C>H=D

                H>D  504
                D>H  394

                H>C  308     C>D  584
                C>H  276     D>C  394

i dunno, but i still see H as the Condorcet winner.  what am i doing  
wrong, Jameson?


> My APV proposal does very well on this and other scenarios.  
> Specifically, for this scenario, the pure strategy which is closest  
> to being a trembling-hand equilibrium is the "good" situation where  
> H wins in one round (Not true of Approval, Bucklin, margins  
> Condorcet, Range; IRV is the only "major" system I know of which  
> passes this test). And it is monotonic, and, unlike IRV, unilkely to  
> fail to find a centrist Condorcet winner.

again, even though Condorcet seems to favor the centrist over either  
extreme, that is not the reason Condorcet is fairer than methods (like  
IRV) that favor the centrist less.  the main reason the Condorcet  
winner (if one exists) should be elected to office is, compared to any  
other candidate, the CW is the candidate the majority of voters prefer  
when asked to choose between the two.  it is the simplest extension of  
the concept of "simple majority" rule from the two-candidate context  
(where we all agree how the votes should be counted) to the multi- 
candidate context.  i don't think centrism is such a bad byproduct of  
Condorcet, but i would still be plugging Condorcet even if it tended  
to favor the centrist less.

> In fact, I can't think of a single scenario where the pure strategy  
> closest to being trembling-hand equilibrium doesn't give an very- 
> arguably "right" answer in one round for APV. And I can easily get  
> IRV to give the wrong answer, so APV is the only system I know of  
> which pass this test.


what is the "trembling-hand equilibrium" strategy?  i dunno what that  
is.

bestest,

--

r b-j                  rbj at audioimagination.com

"Imagination is more important than knowledge."