It seems to me that Iterative Condorcet strictly meets LO2E-2, even without the use of the subcycle rule. Using the symbols in my definition of LO2E-2, if the voters in set M refuse to rank any candidate in set S2, and one of them wins a particular round, then all of them will extend 1st choice status to their next choice that doesn't yet have it, and, since they've all ranked all the S1 candidates over all the S2 candidates, then eventually the winner will be in S1, and that will remain true after all the iterations. At least the winner certainly won't be from S2. Iterative Condorcet is probably easier to explain to the public than the subcycle rule, so, since it gives Condorcet the automatic LO2E-2 compliance of Bucklin, that's a reason toi propose Iterative Condorcet instead of Condorcet with the subcycle rule. Then of course there's the fact that Iterative Condorcet adds a whole additional level of making succesful cheating unlikely & difficult, as compared to ordinary Condorcet (though ordinary Condorcet & ordinary Smith//Condorcet do a good enough job for public elections, it seems to me), and thereby probably makes it possible to say that Iterative Condorcet virtually meets the stronger LOE criteria, the perfect method criteria written by Steve & Rob. I'm pretty much sure that Iterative Condorcet has these benefits, and the main question is how good a method is needed for public elections. A method that, like Condorcet & Smith//Condorcet, gets rid of the significant bulk of the LO2E problem, or a more refined version like Iterative Condorcet, that pushes even closer to idealness. Mike --
