Rob, --- Rob LeGrand <[EMAIL PROTECTED]> a écrit : > Kevin, could you post an example that gives a favorite-betrayal > incentive under margins but not winning-votes?
I can do a bit better than this. I can try to explain why we would expect margins to offer more FB incentive than WV. The common scenario is that your sincere ranking is A>B>C, and this results in an A>B>C>A cycle in which B>C is the weakest link, so that C is elected. It's possible that the only way to get B to win is to reverse the direction of the A>B win. In this case, WV, margins, and DMC all may offer FB incentive, since changing your A>B vote to B>A is twice as fast as A=B at switching around which one of the two wins. In MMPO, this scenario doesn't exist, because there's no such thing as "reversing the direction of a win." Essentially every candidate has some "win" over every other candidate, so all you can do is weaken opposition. The methods are quite different in the case where it is sufficient to *weaken* the strength of the A>B win without needing to reverse it. In WV and MMPO (that is, MinMax(pairwise opposition)), there is never favorite betrayal in this case, because changing the vote A>B to A=B is no more effective than changing it to B>A. Every voter making this change reduces the A>B strength by 1 vote. But in margins, favorite betrayal always works twice as fast as equal ranking. Changing A>B to A=B shrinks the margin by 1 vote, while changing it to B>A shrinks the margin by 2 votes. Sometimes B>A is going to be the only way to weaken the defeat sufficiently. In DMC, neither A=B nor B>A is effective. The only ability you have to weaken the A>B win (assuming you can't reverse it) is to disapprove A! (My FBC simulation doesn't envision favorite betrayal of that kind, so I would have to rethink it in order to find statistics for DMC relative to WV and margins.) Kevin Venzke ___________________________________________________________________________ Appel audio GRATUIT partout dans le monde avec le nouveau Yahoo! Messenger Téléchargez cette version sur http://fr.messenger.yahoo.com ---- Election-methods mailing list - see http://electorama.com/em for list info