I've just realized that I left important words out of the proposal that I just posted. So let me start over: Throw out the least defeat in any 1st order cycle. Then, in any 2nd order cycles (cycle that is a subcycle of a 1st order cycle), throw out the least defeat among the defeats of alternatives no longer defeated in a 1st order cycle. Continue, for any next order cycle, throwing out the least defeat in any next order cycle, among the defeats of alternatives no longer defeated in lower order cycles. *** Now, that's what I meant to say before. In the examples I tried, it seems to avoid the problems of other Condorcet versions, and avoid all subcycle fratricide, while being more decisive than the other recently-discussed Condocret versions. *** According to Markus's recent posting, what Condorcet was proposing is what we call plain Condorcet. If it were to go by votes-against, it would be plain Condorcet(EM). *** It seems to me that all of these votes-against Condorcet versions meet the 2 criteria that say new sincere voters shouldn't be able to defeat their favorite or elect their last choice, as a result of participating in the election. *** It also seems that none of them can, and probably no top quality method can, meet the criterion that says that a new sincere voter (or block of identically-voting ones) can worsen the result for themselves in any way as a result of participating in the election. CVD is fond of making that claim for IRO, by saying that your lower choice votes can never hurt your upper choice ones. But, the word "sincere" is the catch. In IRO, you often worsen the result for youself by sincere voting, compared to what it would be if you voted insincerely. And if you vote insincerely when it seems strategically necessary, then, when you've guessed wrong, then you most assuredly can worsen the result for yourself, and even defeat your favorite, or even elect your last choice. So, when CVD makes that claim, these things should be pointed out. Mike Ossipoff
