[EM] A "Runoff" type method for resolution of cycles incondorcet

[Eric Gorr]

At 10:30 AM -0400 4/27/04, Ken Taylor wrote:
>  > At 10:21 AM -0400 4/27/04, Ken Taylor wrote:
>>  >I'm dropping the weakest candidate, as defined by number of first
>>  >choice votes, which causes all their defeats of other candidates
>>  >to be dropped.
>>
>>  What will you do when two or more candidates are tied for least first
>>  choice votes?
>>
>
>I'm not sure. There are probably several different ways to deal with this,
>and I don't think it's very relevant to evaluating the procedure in general.

If you were to drop all candidates tied for least first choice votes, 
it seems possible that your method could not claim to be cloneproof. 
Two or more clones could evenly split the vote and then all be 
eliminated even if they, in combination, had > 50% of the vote, 
implying that one of the clones should win.

>What does IRV do if there are two candidates tied for last place in a round?

There is no standard.

If you want to see an IRVist cringe, just ask them what they think it 
should be.
It's something they generally don't like to talk about as it destroys 
the illusion of IRV being a simple method. The answer does matter as 
it can change who wins.

In genuine implementations, the two common methods is to either 
select one at random or to drop all of them. Basically, the reason 
why these methods are selected is that there is no expectation that 
this event will occur or that it will change the winner. However, 
this expectation is false and I have several genuine examples 
demonstrating this 
(http://www.ericgorr.net/library/tiki-index.php?page=BallotArchives).

http://demochoice.org attempts to decide based on prior round vote 
totals and, if it cannot, selects randomly.

Of course, one could also employ Steve Eppley's excellent Random 
Voter Hierarchy to make the decision.

>What does one do in SSD if there are two defeats that are tied for smallest
>magnitude?

SSD is not cloneproof, so we really only care about CSSD, which is 
equivalent to the Beatpath Winner method and is well defined.



-- 
== Eric Gorr ========= http://www.ericgorr.net ========= ICQ:9293199 ===
"Therefore the considerations of the intelligent always include both
benefit and harm." - Sun Tzu
== Insults, like violence, are the last refuge of the incompetent... ===