[EM] A Generalization of the Condorcet Criterion

[Forest W Simmons]

I believe that (except for the margins vs. wv controversy) most 
Condorcet fans agree on how to pick the winner in a three way single 
winner race: if there is no CW, break the cycle at the weakest link.

Now suppose there is a candidate X who wins by this standard in every 
three-candidate subset of the form {X,Y,Z}, where Y and Z are from 
among the other candidates in the race.  In other words, X either beats 
both Y and Z, or there is a cycle in which X's defeat is minimal.  Then 
we could say that X is a kind of generalized CW, or GCW.

GCW Criterion:  If a method M always chooses the GCW when there is one, 
then the method M satisfies the GCW criterion.

Comments?  Has this been discussed before?
  
Forest