Talk:Copeland's method

From Electowiki
Jump to navigation Jump to search

Resisting cloning[edit source]

I think it might be possible to make Copeland resist cloning by repeatedly eliminating the candidate with the lowest Copeland score until the remaining candidates all have the same Copeland score. I proved in the "Criteria" section that Copeland passes ISDA, so doing this at the very least doesn't eliminate Copeland's Smith-efficiency. BetterVotingAdvocacy (talk) 01:34, 1 March 2020 (UTC)

Can't you clone on the loser end? You arrange the ballots so that there's a sensitivity to initial conditions depending on whether A or B gets eliminated, and then you make A and B a tie so that there's 50% chance of X winning, and 50% chance of Y winning in the end, whereas before the cloning, someone wins with certainty.
The reason you can't do that in IRV is because IRV is essentially blind to later preferences until someone is eliminated. But finding out whether something like that is impossible in loser-elimination Copeland would require more thought. Other loser-elimination methods like Coombs definitely aren't cloneproof (see e.g. Warren's Yee diagrams). Kristomun (talk) 10:42, 1 March 2020 (UTC)

Issue with simplifying proofs[edit source]

User:Kristomun, I just want to point out that the version of Copeland discussed in this article is "wins - defeats", not just wins. That is the main reason why I had to make the proofs so long as they were. I do not mind if you only want to provide proofs for the simpler "wins" version, but it'd probably be best to mention that in the article. BetterVotingAdvocacy (talk) 00:05, 16 April 2020 (UTC)

That's a good point, so I've added some text to show that the argument also holds for wins-losses. Kristomun (talk) 11:15, 16 April 2020 (UTC)