I'm looking for both degenerate cases/extremes, as well as real/realistic data. The more samples the better.
I have something I wrote that generates "random" ballots by giving each candidate a random point in space (an x y and z value within a fixed volume), and then each voter would get a random point in space, and each ballot is put in order based on distance from the voter's point to each candidate's point. This was helpful, but I don't think it was entirely realistic (it sort of assumed that voters were exceptionally rational), and very rarely produced elections without a condorcet winner. There are various things that can be adjusted to make it less regular and possibly more realistic (such as changing the boundaries of the voter space and candidate space, or randomly tweaking the distances prior to sorting them)....but still, its just a guess as to what is realistic. I'd like something better if its out there.
-rob
---- election-methods mailing list - see http://electorama.com/em for list info
