[EM] An idealized DSV method in the spirit of XA

[Kristofer Munsterhjelm]

On 11/02/2016 11:46 PM, Forest Simmons wrote:
> Vote score ballots.
> 
> For each ballot B the method is used recursively to determine what the
> average scores of the candidates would be if that ballot were not
> included in the election.  Then the scores of ballot B are adjusted to
> B' so that if it (B') were averaged in (as one additional ballot,
> weighted accordingly) it would improve the previous averages (towards
> the ratings of ballot B) as much as possible.
> 
> The final ratings arre the average scores of the respective adjusted
> ballots.
> 
> This method is almost surely intractable computationally, but it has the
> spiit of XA.  I believe that for a large diverse collection of Score
> ballots XA would give an excellent approximation to the result of this
> recursive method.

This sounds similar to an exhaustive version of SARVO-Range, except that
your DSV method is based on XA rather than on Approval.

It only takes individual strategy into account, not organized strategy.
I guess that for organized strategy, you'd need to find the strong Nash
equilibrium, ESS or something like it.