Well, since Condorcet is (to the best of my knowledge) never used in public elections and rarely used in private groups, I don't know whether the 0.5 votes each is a standard convention or not. It wouldn't change the margins, so it would be applicable to margins methods. Certainly it would be necessary for winning votes methods.
No, absolutely not. When you add half-votes, the results for winning votes become identical to margins. The whole point of winning votes is you only count the votes FOR the candidate in a pairwise contest.
Think about it - if you add a half vote for every equal ranking, then the pairwise results will ievitably end as 50+(m/2)% against 50-(m/2)%, where m is the margin. So you haven't changed the margin total (like you said), but you've made the winning votes total a linear function of the margins. So any method that resolves cyclic ties by comparing defeat-strength will do the same thing in winning votes with half-votes for ties as it would in margins (with or without half-votes).
I'm pretty sure you don't mean to start a winning-votes vs. margins debate, so I'll keep this brief - using winning votes to evaluate defeat-strength gets rid of the strategic incentive to insincerely truncate your ballot in nearly all situations.
Maybe there's also a strategic advantage from intransitive rankings, and Donald wants voters to have access to that option by allowing them to vote A>B, B>C, C>A.
Well, I don't think that's what Donald wants. Moreover, it seems absurd to give an individual voter the ability to submit an ambiguous ballot. Cyclic ties can make sense for an electorate, but they can't make sense for a sane individual.
-Adam
---- Election-methods mailing list - see http://electorama.com/em for list info
