[EM] uncovered set

[Kristofer Munsterhjelm]

fsimmons at pcc.edu wrote:
> River, Schulze, and Ranked Pairs all give the win to D in the scenario
> 
> 40 D>B>C>A
> 30 A>B>C>D
> 30 C>A>D>B
> 
> because in the pairwise beatpath D beats B beats C beats A, all of the defeats are 70 to 30, and all 
> other defeats are weaker.
> 
> But alternative D is covered by A.

> Even though we cannot have monotonicity with independence from covered alternatives, we can have 
> monotone methods that choose exclusively from the uncovered set.  Copeland is such a method, but it 
> has a high rate of tied outcomes, and fails clone independence.

Presumably, my second order Copeland variant would be more decisive. 
However, unless it is 2-1-0 or greater (two points or more for a win), 
it would not be monotone. I think it elects uncovered candidates, but I 
only base that on performance with regards to my test program, and that 
can't prove compliance, only disprove it.

Speaking of River, I know (and we discussed) that it meets something 
called ISDA which is almost-but-not-quite independence of uncovered 
alternatives. Perhaps it would be possible to make a method that meets 
ISDA and elects from the uncovered set (unlike River itself does).