[EM] CDTT,IRV (Chris)

[Kevin Venzke]

Chris,

--- Chris Benham <chrisbenham at bigpond.com> a écrit:
> >40 A>B>C (sincere)
> >25 B>A>C
> >35 C>B>A
> >
> >IRV/FPP/DSC order is A>C>B; CDTT is {b}.
> >
> >40 A>C>B (insincere)
> >25 B>A>C
> >35 C>B>A
> >
> >IRV/FPP/DSC order is A>C>B; CDTT is {a,b,c}.
> >
> >Yes, CDTT methods have the same burial problem (and solution) as
> >WV methods. That's one reason I suggest CDTT,RandomBallot.
> >
> Thanks for this clarification, which I'd wrongly hinted wasn't true. But 
> take this really outrageous scenario (one of James G-A's):
> 
> 46: A>B>C
> 44: B>C>A   (sincere is B>>>>A>C)
> 05: C>A>B
> 05: C>B>A
> 
> A is the sincere CW, and the (voted) CDTT is {A,B,C}.  Pairwise 
> Defeat-Dropper(Winning Votes)  elects the Buriers' candidate B, while 
> CDTT,IRV easily elects A.

It's interesting. IRV manages this by eliminating C due to low first 
preferences, whereas FPP and DSC favor the first-preference winner, who
could predictably be the buriers'. CDTT,MMPO elects C, I believe.

It makes me wonder whether the IRV behavior could be duplicated without
failing Mono-raise. (I'd prefer to fail Clone-Winner.)

> With your suggestion CDTT,Random Ballot the Buriers increase the chance 
> of their favourite winning from zero to 44%. This is a huge bargain for 
> them if  their sincere ratings
> gap between their second and last preference is much smaller than the 
> one between their first and second preference. (In any case, here the 
> chance of their sincere last preference
> being elected only rises from zero to 10%).
> So on balance I don't think CDTT,RB really resists Burying better than 
> CDTT,IRV.

No, I wouldn't say that, either. But I would hope that some buriers would
consider the fact that they will cause the third candidate to be elected
with some probability.

Kevin Venzke



	

	
		
_____________________________________________________________________________ 
Découvrez le nouveau Yahoo! Mail : 1 Go d'espace de stockage pour vos mails, photos et vidéos ! 
Créez votre Yahoo! Mail sur http://fr.mail.yahoo.com