[EM] NPV vs Condorcet

[Kristofer Munsterhjelm]

Dave Ketchum wrote:
> The first N*N matrix  below is what I was talking about - it takes the 
> information from the FPTP votes and records that as if bullet votes.
> 
> Note that this example matrix is complete only for only three 
> candidates. If there were seven candidates the matrix would be bigger, 
> showing A, B, and C ranking over the other four.
> 
> On Wed, 22 Oct 2008 18:18:00 +0200 Kristofer Munsterhjelm wrote:
[snip]
>> That's the "consider bullet voters" idea. The other one is to count 
>> the plurality vote locally, so you get:
>>
>> 100: Able > Baker > Charlie
> 
> BUT, the FPTP voters could not express such thoughts.

That's right, they could not. This would simply be the result of the 
Plurality election (for that district), turned into a preference 
ordering of magnitude equal to the number of voters in that district. 
Such a generalization would work no matter what method you'd use (even 
IRV), but it'd be suboptimal since it compresses down the data of all 
the voters into a single preference, whereas with FPTP-as-Condorcet, 
you'd get the data directly.