[EM] Re: "Condorcet corresponding to some variant of IRV"

[Chris Benham]

Bjarke,
I have had a closer look at your suggested method and I think it is 
superb. I think the short answers to your questions are yes and yes,but 
no real problem if  I am wrong because in that case we can apply the 
method to the Smith set.
Please tell me if a couple of assumptions I have made are correct. I 
assume that there are no more rounds once one of the candidates receives 
more than half the votes ( even though there are unused second 
preferences which if released could  overtake that candidate). And I 
assume that if there is a tie on the last round, then the winner will be 
the tied candidate who was ahead in the previous round (and if there are 
2 tied candidates they don't runoff).
One of the strengths of IRV is that a voter's lower preferences can 
never harm the prospects of the voter's  higher preferences, and I think 
your method shares that quality, or nearly does.
Here is a ticky example someone gave here earlier this year:
36:  R
 9:   C  >  R
18:  C          ( >  L )
18:  L  >  C
19:  L
Eighteen C voters are "insincerely truncating", and by most of the 
methods proposed for resolving circular ties, thereby cause C to win 
when otherwise C would have lost. (to me the problem should be put the 
other way round, i.e  in the case where those C voters vote sincerely 
there lower prefeerences have caused their favourite to lose.)
But to employ your method,  1st. round: R : 36   C: 27   L: 37 ,   2nd. 
round: R: 45   C: 27   L: 37,  
                                            3rd. round: R:  45   C: 45   
L: 37     As there are no more preferences to distribute, and as R was 
 ahead last round, then I assume that R wins.

Chris Benham