Re: Constitutional amendment: Condorcet/Clone Proof SSDvotetallying

[Andrew Pimlott <andrew@pimlott.net>]

On Sat, May 31, 2003 at 02:23:07AM +1000, Anthony Towns wrote:
> On Fri, May 30, 2003 at 10:48:06AM -0400, Nathanael Nerode wrote:
> > In a pure Condorcet system with a default option, lying about the 
> > acceptability of A *doesn't* help B.  
> 
>     40 C A B
>     30 A B C
>     20 B C A
> 
> C defeats A 60:30, A defeats B 70:20, B defeats C 50:40; the weakest
> defeat is dropped, C wins.  If the ABC voters switch to:
> 
>     30 B C A
> 
> then B wins -- lying about the acceptability of A *does* help B, even in a
> pure Condorcet case (well, pure CpSSD -- although even in pure Condorcet,
> B winning is probably preferable to no result).

This is the point that I noted complicates the analysis:  Strategic
voting can work in Condorcet+CpSSD.  It just works less often than
in the proposed system.  If you carry out more examples, I think
you'll conclude that it works in appreciably fewer cases, though I
don't know how to quantify this.

Eg, sincere:

    9 ABD       A>B 12:8
    6 BAD       A>D 15:5
    3 DAB       B>D 15:5
    2 DBA

B can swap and make D>A 11:9, but in Condorcet/CpSSD A still wins.
I think this is related to the Strong Defensive Strategy Criterion.

Andrew