[EM] Convergent Allied Comparison: new fpA-fpC generalization, producing a comparison matrix

[Kristofer Munsterhjelm]

On 5/19/23 14:40, Filip Ejlak wrote:
> I've got a new fpA-fpC-compliant method and this time I am very 
> satisfied with its behaviour (after checking some test examples that 
> some other methods failed, though I guess it needs some robust checks on 
> monotonicity). Instead of just pairwise comparing the candidates, it 
> compares the "alliances" of these candidates in order to create a 
> special pairwise comparison matrix (in a more burial-resistant manner). 
> The mechanism tends to break Condorcet cycles, even though there are 
> some rare cases where it doesn't clear /all/ the ambiguity and a 
> completion method has to be used.
> 
> The "agreement" rule was created specifically to make the method 
> cloneproof (but it also had to be strict enough so that a) the burial 
> resistance was not weakened, b) the method didn't stop converging, i.e. 
> achieving stable solutions).
> 
> It's worth noting that - in a given iteration - a covered candidate 
> can't have any allies against the covering candidate, so a covering 
> relation should cause a visibly strong defeat.

Great! That does look rather complex; I'll have to investigate further. 
I guess I really should make those automatic criterion checkers, but 
it's really god that you are picking up fpA-fpC and trying to improve it :-)

-km