In my opinion the biggest problem with Copeland is its clone problem: the strongest candidate in a large set of clones has an undue advantage over cloneless candidates.

There are various kinds of clones. The kind that is relevant in the case of Copeland is what we might call a "beat clone set."

A subset B of candidates is a beat clone set if and only if every member of the complement of B that beats any member of B beats all members of B, and any member of the complement of B that is beaten by one member of B is beaten by all members of B.

A clone set is proper if its cardinality is strictly between one and the total number of candidates.

To de-clone Copeland "mod out" all of the proper clone sets one-by-one in any order by replacing each clone set with any of its members.

Apply ordinary Copeland to the resulting set of candidates. If the winner is the representative of a clone set that was "modded out," then recursively apply de-cloned Copeland to that clone set.

That's it.  I don't have time to go through examples right now.

What do you think?

Forest
----
Election-methods mailing list - see http://electorama.com/em for list info

Reply via email to