[EM] A cloneproof fpA-fpC method?

[Kristofer Munsterhjelm]

On 2024-10-25 10:12, Chris Benham wrote:
> What does "fpA-fpC" mean?

First preference for A minus first preference for C.

i.e. in a three-candidate election that's an ABCA cycle, elect the 
candidate A whose first preferences minus the preferences of the one 
beating him is maximized. (If there is a CW, just elect that CW.)

This has similarly low manipulability as the Condorcet-IRV hybrids while 
being monotone. Later I found a connection between low manipulability 
and electing from the resistant set. And sure enough, both 
three-candidate IRV and fpA-fpC elect from this set.

The problem is that it's only defined for three candidates. So I've on 
and off been playing with extensions that would generalize to more 
candidates while retaining monotonicity and some of the strategy resistance.

But none of the generalizations have been cloneproof; even my best 
attempt so far (Friendly Cover) is only independent of twins (the way 
minmax is). This extension, however, might just be cloneproof and have 
some nice Smith-independence properties as well.

Unlike the Condorcet-IRV methods, it (probably) doesn't elect from the 
resistant set in general, but it does with a Smith set of three or 
fewer. I still haven't found a monotone method that always elects from 
the resistant set, but I'm not going to discard partial results like 
this method for that reason.

-km