[EM] POLL: References (was Re: Poll, preliminary ballots)

[Chris Benham]

>
> Here's a list of the methods: 
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005767.html
> Minmax(wv):
> 	https://electowiki.org/wiki/Minimax_Condorcet_method
> 	This method takes ranked ballots.
> 	Based on the ballots, consider each one-on-one matchups between
> candidates. Elect the candidate who has the most comfortable lead
> against the candidate who does best against him, one on one.
I find that last sentence to be vague and nonsensical. The linked-to 
electowiki description is much better:
> *Minmax(winning votes)* elects the candidate whose greatest pairwise 
> loss to another candidate is the least, when the strength of a 
> pairwise loss is measured as the number of voters who voted for the 
> winning side. 

> Smith//DAC:
> 	This is a composite method.https://electowiki.org/wiki/Composite_methods
> 	https://electowiki.org/wiki/Descending_Acquiescing_Coalitions
> 	It takes ranked ballots.
> 	First eliminate every candidate not in the Smith set. Then perform the
> DAC algorithm to determine the winner.
> 	The DAC algorithm is complex, with the result being a cloneproof method
> otherwise close to Plurality.
No, the person who wrote this is confusing it with DSC.

46 A
44 B>C
10 C

DSC elects A but DAC elects C.   The " descending acquiescing 
coalitions"  are AC 56 (disqualifying B), BC 54 (disqualifying A) 
leaving C as the winner.  The largest "solid coalition" is A 46, 
disqualifying B and C leaving A as the winner.

DSC meets Later-no-Harm but fails Later-no-Help (thereby giving it a 
silly random-fill incentive) while with DAC it is the other way round.  
So DAC has a quite strong incentive to truncate and (in this version 
that allows above-bottom equal-ranking) to equal rank at the top.

Both LNHs are incompatible with Condorcet, so the "Smith//..." kills 
that.  But Smith//DAC is a Condorcet method that has very strong 
resistance to Burial, at the expense of a strong truncation incentive.


> Essentially, the DAC algorithm works like this: For each ballot, for
> each set of candidates a voter acquiesces to, increase that set's count
> by one. Then sort the sets in order of count, descending. Starting at
> the top with the current viable set being every candidate, eliminate
> from the viable set every candidate who is not in the set at the list's
> current position, unless that would eliminate everybody. The candidate/s
> remaining once you've traversed the whole list is/are the winner/s.
> 	A voter acquiesces to a set if he doesn't rank any candidate outside
> the set strictly above a candidate within the set.

"Disqualify" is a much better choice of word than "eliminate", which 
usually means "drop from the ballots and carry on as if though the 
candidate had never been on them". How I put it in a November 2023 post 
here:

> By itself DAC (one of Woodall's inventions) meets Later-no-Help and Participation (both incompatible with Condorcet) but can behave very oddly and badly in the presence of one or two weak should-be-irrelevant candidates (which is why I'm wary of Smith,DAC).
>
> So defining it thus: Voters rank from the top, equal ranking and truncation is fine.  Eliminate and drop from the ballots all the candidates not in the Smith set.Ballots "acquiesce" to a candidate or set or subset of candidates (a
> "coalition") if they vote no other (outside the set or subset) candidate strictly above any of them.
>
> Number all the possible coalitions according to how may ballots acquiesce to them. Start with the highest-numbered and disqualify all the candidates not in it.Proceed to the next-highest numbered that contains any not-yet
> disqualified candidates and disqualify those not in it, and so on until one candidate is left undisqualified.

I'm sorry I didn't get on to this earlier. Also I think when we set up the poll and nominated the methods, there should have been more clarity about balloting rules. For example I quite like IRV and Benham (and I suppose Woodall) with unlimited strict ranking from the top. But with restricted ranking as is apparently typical if not universal in the US (and/or equal above-bottom ranking allowed without a complicated procedure I like that is definitely not worth the trouble) then I like those methods far less.

It might have been better to have separate polls for the different ballot types and restrictions, and another poll on what ballot types and rules we like. I find limiting the absolute number of candidates a voter can vote above bottom to be abhorrent, but limited-slot rating ballots (or ballots with a restricted number of "ranking levels" but no limit to the number of candidates the voter can put at each level) are fine or not-too-bad for a lot of methods.


Chris Benham


On 27/04/2024 1:20 am, Kristofer Munsterhjelm wrote:
> On 2024-04-26 16:34, fdpk69p6uq at snkmail.com wrote:
>> Can someone point me to the description of how to participate in this 
>> poll and what the nominees are?  Or just re-post the instructions 
>> clearly with an obvious subject line, so people can find it?  There 
>> are dozens of messages in a bunch of different threads and I don't 
>> see clear instructions anywhere.
>
> Here's a list of the methods: 
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005767.html
>
>
> and a description of most of them: 
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005802.html
>
> Those not described there are:
>     Margins-Sorted Approval: 
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005650.html
>     Double Defeat, Hare: 
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005708.html
>     Margins-Sorted Minimum Losing Votes: I couldn't find a direct 
> description, closest one I found is 
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005785.html
>     Gross Loser Elimination: This is Raynaud(gross loser). See 
> https://electowiki.org/wiki/Raynaud
>     Max Strength Transitive Beatpath: 
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005716.html
>
> The poll text is here:
> http://lists.electorama.com/pipermail/election-methods-electorama.com/2024-April/005934.html 
>
>
> The poll asks for a ranked ballot (truncation and equal-rank are 
> allowed), and an Approval ballot.
>
> -km
> ----
> Election-Methods mailing list - see https://electorama.com/em for list 
> info
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20240427/567ada94/attachment.htm>