[EM] Note on maximally beneficial elimination

[Kristofer Munsterhjelm]

I forgot to say that a "maximally beneficial elimination" method can 
trivially be made ISDA by constraining the elimination order to 
eliminate every non-Smith candidate, i.e. the remaining candidate/s 
besides A himself can't be outside the Smith set. But this is just 
Smith//X and thus probably not monotone (consider e.g. Smith//Plurality 
as proof that Smith//X may be nonmonotone even if X is monotone).

But perhaps a Benham type constraint would preserve monotonicity while 
passing Smith (though it would likely fail ISDA). Let the Benham 
elimination according to some order be:
	- At each step, eliminate the candidate at the end of the elimination order
		UNLESS that candidate beats every remaining candidate pairwise,
		in which case eliminate the next to last instead.

A's score still involves finding some elimination order that maximizes 
the score of A according to the base method once all but a certain 
number of candidates have been eliminated. But maybe this Benham MBE is 
monotone?

If the base method passes Condorcet (e.g. first preference Copeland or 
Friendly Cover) then if W is the CW, every other candidate will have a 
lower score than W when using Benham MBE.

And perhaps some clever adaptation of Tideman's alternative methods 
could get us ISDA. Still no clone independence, though... I don't think?

(It would be something like: the order is constrained to get rid of 
non-Smith first. Then it can eliminate in any arbitrary order until 
there is a distinct Smith set among the remaining candidates, after 
which this Smith set must similarly be eliminated before more arbitrary 
candidates can be eliminated. In either case, A is always spared when 
calculating A's score. This *sounds* like a variant of the Smith//X 
thing and possibly would not be monotone.)

-km