Jobst, I'm worried about a kind of incentive for insincere voting:
Consider
x A>B>C y B>C>A z C>A>B
where max{x,y,z} < 50%, x+y+z=100%.
If we do random ballot chain climbing, then the respective winning probabilities for A, B, and C are z, x, and y.
Supporters of A have an incentive (up to a certain point) of voting with the C supporters to increase the probability z of A getting elected.
How do you rationalize this?
Forest ---- Election-methods mailing list - see http://electorama.com/em for list info