[EM] Condorcet - a review

[Dave Ketchum]

Trying for the high points.

While Condorcet concedes to Range ability to indicate strength of  
preferences, it's ranking gives voters unconditional ability to  
indicate, for any pair of voters, which is liked better (I expect  
Condorcet to permit multiple equal ranks, though I know some who  
disagree choke).

Voters can rank:
      (A, or A and B each)>(each and every other candidate).  This  
leads me to suspect claims that Condorcet does not  produce a winner  
often - MANY races have only one or two strongest candidates, for  
which Condorcet easily identifies the winner (ties can be solved by  
chance).
      (A and B and C each)>(each and every other candidate).  This  
includes most cycles.
      When above has A>B and B indirectly above A via B>C>A, we have a  
cycle.  Note that this requires compatible patterns of voting for no  
one voter can express all of A>B>C>A.
      Significant voting patterns with more than three leading  
candidates - rare, but must be planned for.

The Condorcet X*Y array both identifies winner when there is no cycle,  
and what votes the cycle members got.  Resolving a cycle from its  
members' X*Y data is obvious and will be expected by voters - though  
what is best method is debatable.
      Resolving wins from those members but demanding additional data  
may be doable, but deserves explanation for the additional effort.   
Remember that while all the votes for some districts may be available  
in one place, other districts will have voting in many locations.
      Letting a non-member of the cycle win deserves voters DEMANDING  
an explanation as to why.

DWK