[EM] Poll, preliminary ballots

[Kristofer Munsterhjelm]

On 2024-04-24 17:37, robert bristow-johnson wrote:
> On 04/24/2024 11:10 AM EDT Chris Benham <cbenhamau at yahoo.com.au> wrote:
>>
>>   
>>
>>> The RCIPE advocate only asks for a rule change that seems modest and logical within the context of IRV: If the candidate was destined to lose *in IRV* anyway, then eliminate him sooner.
>>
>>   Why? All the IRV non-winners were "destined to lose", and the easiest way to identify them is to complete the IRV count. That seems easier than looking for Condorcet Losers.
>>   
> 
> Finding the Condorcet winner is easier than finding the Condorcet loser.
> 
>>
>>> In exchange, RCIPE achieves quite a lot.
>>
>>   Really? As a modification of IRV how much does it "achieve" in comparison with what it loses? Rescuing the occasional Condorcet winner to make the method a lot more complicated and trash a lot of IRV's popular criterion compliances??
>>   
>>   I can't see how looking for Condorcet losers is any way easier than looking for Condorcet winners. So why don't we just do that (before each elimination, among the remaining candidates) instead?
>>   
>>   That method (Benham) is a Condorcet method and quite a bit simpler to operate than RCIPE. So the argument for RCIPE versus Benham is ...what??
>>
> 
> <applause>  But Chris, BTR-IRV is even simpler than Benham.  What
> advantage does Benham have that BTR-IRV does not have?

Clone independence and strategy resistance. Burial is harder to make 
work under Benham (and Woodall) than BTR-IRV.

I'll soon be posting some strategy figures for the poll methods that my 
simulator implements. I once did a test with BTR-IRV, but for some 
reason I forgot to save my code. From what I recall, its strategy 
resistance under impartial culture is similar to that of Minmax. The 
other Condorcet-IRV hybrids are much more resistant.

-km