[Election-Methods] IRV-Tournament

[Dave Ketchum]

On Thu, 17 Jul 2008 18:23:20 +0300 Juho wrote:
> On Jul 17, 2008, at 16:12 , Bruce R. Gilson wrote:
> 
> 
>>My beef with Condorcet methods is
>>that you need to have a cycle-resolving procedure (which you don't in
>>any other system, except in the case of exact ties
> 
> 
> Unfortunately other methods need to resolve the cycles just as well  
> (not necessarily with an explicit "sub-procedure" but one way or  
> another in any case).
> 
> 
>>and I feel that having a cycle-resolving procedure that the
>>voters both UNDERSTAND and ACCEPT AS FAIR may not be easy to do.
> 
> 
> If the method is presented to the voters as containing first the  
> "normal procedure" and then an exceptional "cycle-resolving  
> procedure" then the voters may get worried due to not understanding  
> what the exceptional procedure exactly means. But of course this is  
> just psychological, not related to if the winners in that method are  
> good or not.

This can be said more clearly than is, too often, done:
      For each candidate pair, x and y, as many voters as choose rank 
either x>y or y>x.
      The one candidate winning each of its pairs, by being liked by 
more voters than the opposing candidate, is elected.
      If no such candidate, we have a cycle such as A>B>C>A in which:
           Each of A/B/C would win over all others (d-z)
           Each would win over at least one other cycle member.
           Each would lose to at least one other cycle member.
           Thus there is a near tie.  A lottery among them would be 
reasonable, but Condorcet normally tries to do better - though we can 
debate, before such an election, over exactly how we shall proceed.
> 
> Other methods like IRV also need to break the same cycles. The  
> breaking of the cycles is not explicitly visible in the IRV procedure  
> description. That hides the breaking from the voters and may keep  
> them more satisfied (or more ignorant of the cycle-resolving  
> process). One would need to also avoid giving out any detailed  
> information about the cast votes in IRV if one wants to hide the  
> cycles, since otherwise the media can point out that there was a  
> cycle and demonstrate how it was resolved (in favour of some  
> candidate that all do not like and that would have lost to someone in  
> pairwise comparison).
> 
> Condorcet is btw not a very good classification of methods that have  
> an explicit cycle-resolving procedure since some Condorcet methods  
> don't have it. E.g. Minmax just finds the candidate whose worst  
> defeat is least bad in one step (without any explicit cycle-resolving  
> phase/procedure). Depending on how the results are announced some  
> clever voters or media may however find out also in Minmax that there  
> was a cycle of opinions.
> 
> It is also quite easy to "UNDERSTAND and ACCEPT AS FAIR" some basic  
> Condorcet methods like Minmax(margins) since it elects simply the  
> candidate that needs least number of additional votes to win all  
> others. (On the other hand all methods tend to have cases where one  
> can at least disagree on which winner is the best. Clear agreements  
> and understanding of the target utility function is needed when the  
> election method is chosen.)

The last statement needs emphasis.  How cycles are to be resolved for 
a location had better be decided before Condorcet is used in an 
election there.
> 
> Juho
-- 
  davek at clarityconnect.com    people.clarityconnect.com/webpages3/davek
  Dave Ketchum   108 Halstead Ave, Owego, NY  13827-1708   607-687-5026
            Do to no one what you would not want done to you.
                  If you want peace, work for justice.