[EM] DMC, Ties & Eppley's RVH

[Eric Gorr]

Dave Ketchum wrote:
> On  Fri, 02 Sep 2005 14:44:40 -0400 Eric Gorr wrote:
> 
>> Eric Gorr wrote:
>>
>>> The primary reason why the RVH is better then simply selecting a tied 
>>> candidate at random is because it provides a statistical edge to 
>>> those candidates who are preferred by a majority of voters.
>>
>>
>>
>> For a concrete example, let's say that you have a genuine three way 
>> tie, with 100 voters total:
>>
>>
>> 50:a>c>b
>> 50:b>c>a
>>
>>
>> 50 voters prefer a>c
>> 50 voters prefer a>b
>> 50 voters prefer c>b
>> 50 voters prefer c>a
>> 50 voters prefer b>c
>> 50 voters prefer b>a
>>
>> Now, simply selecting one of these candidates at random, provides 'c' 
>> with a ~33% to win the election. Does this make sense? Not 
>> really...after all, 'c' was not the first place choice of _any_ voter. 
>> The RVH recognizes this and would provide 'c' with absolutely no 
>> opportunity to win. It provides 'a' and 'b' a 50% chance to win.
>>
> Not at all clear that that is proper resolution:
> 
>      Looking at A/B, A/C, and B/C, each pair has ties.
>      Looking closer, no one hates C - perhaps reason enough to elect C, 
> since A and B each inspire dislike/hate.
> 
> Not ready for such detailed analysis, I like the 33% odds, and do not 
> need RVH to implement them.

What if the party representating A or B decided to run candidate C?

Suddenly that party would have a 66% chance to win the election in this 
example. Not a good thing.




-- 
== Eric Gorr =============================== http://www.ericgorr.net ===
"Government is not reason, it is not eloquence, it is force; like fire,
a troublesome servant and a fearful master." - George Washington
== Insults, like violence, are the last refuge of the incompetent... ===