Hey all, Rob asked me to send out this method I've created to see what everyone thinks. I call it 'ART' for Approval, Runoff Tally (in the name of a good acronym). The method works as follows: Voters rank each candidate 'Good', 'Acceptable', or 'Bad' (like 321). To find the winner you simply add up the 'Acceptable' and 'Good' scores for each candidate and the top two advance, ensuring whoever wins is amongst the most favorable. Then the winner is selected based on a head-to-head automatic runoff, higher score on more ballots wins. The condorcet winner wins every time in this method except when they are not among the most approved. This is the biggest question I have however, I think that the weight of Approval in this method is stronger than condorcet winning (especially in a method where head-to-head tie scores are possible). The condorcet loser never wins. Please see this sheet <https://docs.google.com/spreadsheets/d/1dHjWfzTHSB-6miAM5mZEskWBVU-sgrVndXH6IM-ZP6I/edit?usp=sharing> to run simulations for a 3 person, single-winner election. Further details and notes on the method are included in the sheet. Instructions for the sheet are simple: Hit delete or backspace on the green box to run a new simulation. If you accidentally change something else please let me know or just revert back to the previous version. Thanks for taking the time to read over this and check out the method. Best, Nate

Hi Nate, Using Approval to select two finalists has been proposed many times over the years, but what has never sat well with me is the theoretical Clone-Loser issue. A candidate who wins by approval, but who can't beat the #2 candidate by approval, could in theory run with a partner try to claim both of the top approval spots. It might not help very often, I would admit. It would probably be a more prevalent strategy in a race where voters are primarily concerned about which party wins and not which candidate. I'm not sure I've heard before, the specific idea to take these approval finalists and do the head-to-head comparison instantly. (With other ways of selecting finalists, yes... For instance staples of mine are pitting the top-ranking winner against the approval winner, or the top-ranking winner against the candidate who is most approved on the ballots not approving the top-ranking winner.) >From a Condorcet standpoint, you're only using one pairwise contest. IRV, too, will use one, once you're at two candidates remaining. So I don't think the Condorcet element is so evident. Kevin Le mercredi 2 décembre 2020 à 21:31:16 UTC−6, Nathaniel Allen <npwning@gmail.com> a écrit : >Hey all, > >Rob asked me to send out this method I've created to see what everyone thinks. I call it 'ART' for  >Approval, Runoff Tally (in the name of a good acronym). The method works as follows: Voters rank  >each candidate 'Good', 'Acceptable', or 'Bad' (like 321). To find the winner you simply add up the  >'Acceptable' and 'Good' scores for each candidate and the top two advance, ensuring whoever wins  >is amongst the most favorable. Then the winner is selected based on a head-to-head automatic >runoff, higher score on more ballots wins. The condorcet winner wins every time in this method except >when they are not among the most approved. This is the biggest question I have however, I think that >the weight of Approval in this method is stronger than condorcet winning (especially in a method where >head-to-head tie scores are possible). The condorcet loser never wins.

Hey Kevin, Thanks for the feedback! I wonder if the Clone-Loser situation still applies in a circumstance where more preference can be expressed. For instance, in this method "Approved" simply means combined 2's and 3's. Thoughts? Best, Nate On Wed, Dec 2, 2020 at 9:23 PM Kevin Venzke <stepjak@yahoo.fr> wrote: > Hi Nate, > > Using Approval to select two finalists has been proposed many times over > the years, but what has never sat well with me is the theoretical > Clone-Loser issue. A candidate who wins by approval, but who can't beat the > #2 candidate by approval, could in theory run with a partner try to claim > both of the top approval spots. It might not help very often, I would > admit. It would probably be a more prevalent strategy in a race where > voters are primarily concerned about which party wins and not which > candidate. > > I'm not sure I've heard before, the specific idea to take these approval > finalists and do the head-to-head comparison instantly. (With other ways of > selecting finalists, yes... For instance staples of mine are pitting the > top-ranking winner against the approval winner, or the top-ranking winner > against the candidate who is most approved on the ballots not approving the > top-ranking winner.) > > From a Condorcet standpoint, you're only using one pairwise contest. IRV, > too, will use one, once you're at two candidates remaining. So I don't > think the Condorcet element is so evident. > > Kevin > > > Le mercredi 2 décembre 2020 à 21:31:16 UTC−6, Nathaniel Allen < > npwning@gmail.com> a écrit : > >Hey all, > > > >Rob asked me to send out this method I've created to see what everyone > thinks. I call it 'ART' for > >Approval, Runoff Tally (in the name of a good acronym). The method works > as follows: Voters rank > >each candidate 'Good', 'Acceptable', or 'Bad' (like 321). To find the > winner you simply add up the > >'Acceptable' and 'Good' scores for each candidate and the top two > advance, ensuring whoever wins > >is amongst the most favorable. Then the winner is selected based on a > head-to-head automatic > >runoff, higher score on more ballots wins. The condorcet winner wins > every time in this method except > >when they are not among the most approved. This is the biggest question I > have however, I think that > >the weight of Approval in this method is stronger than condorcet winning > (especially in a method where > >head-to-head tie scores are possible). The condorcet loser never wins. >

I'm not sure what you have in mind there. The fact that the 2/3s only differ in effect for the "second phase" (the pairwise contest) is practically the source of the issue I would say. If one attempts the nomination strategy, the goal is to grab both finalist positions, before the distinction between 2/3 makes any difference. If the distinction mattered earlier, in some way, perhaps it could be made impossible for the same voters to select both finalists. I do think as well, as it stands, that someone could give you some grief that a candidate who uniquely possesses "3" ratings from more than half the voters could theoretically fail to be a finalist. (Like if voters give out "2" ratings generously.) Kevin Le lundi 7 décembre 2020 à 10:59:28 UTC−6, Nathaniel Allen <npwning@gmail.com> a écrit : >Hey Kevin, > >Thanks for the feedback! I wonder if the Clone-Loser situation still applies in a circumstance where more  >preference can be expressed. For instance, in this method "Approved" simply means combined 2's and  >3's. Thoughts? > >Best, > >Nate > > >On Wed, Dec 2, 2020 at 9:23 PM Kevin Venzke <stepjak@yahoo.fr> wrote: >> Hi Nate, >> >> Using Approval to select two finalists has been proposed many times over the years, but what has never sat well with me is the theoretical Clone->Loser issue. A candidate who wins by approval, but who can't beat the #2 candidate by approval, could in theory run with a partner try to claim both >of the top approval spots. It might not help very often, I would admit. It would probably be a more prevalent strategy in a race where voters are >primarily concerned about which party wins and not which candidate. >> >> I'm not sure I've heard before, the specific idea to take these approval finalists and do the head-to-head comparison instantly. (With other ways of >selecting finalists, yes... For instance staples of mine are pitting the top-ranking winner against the approval winner, or the top-ranking winner >against the candidate who is most approved on the ballots not approving the top-ranking winner.) >> >> From a Condorcet standpoint, you're only using one pairwise contest. IRV, too, will use one, once you're at two candidates remaining. So I don't >think the Condorcet element is so evident. >> >> Kevin >> >> >> Le mercredi 2 décembre 2020 à 21:31:16 UTC−6, Nathaniel Allen <npwning@gmail.com> a écrit : >>>Hey all, >>> >>>Rob asked me to send out this method I've created to see what everyone thinks. I call it 'ART' for  >>>Approval, Runoff Tally (in the name of a good acronym). The method works as follows: Voters rank  >>>each candidate 'Good', 'Acceptable', or 'Bad' (like 321). To find the winner you simply add up the  >>>'Acceptable' and 'Good' scores for each candidate and the top two advance, ensuring whoever wins  >>>is amongst the most favorable. Then the winner is selected based on a head-to-head automatic >>>runoff, higher score on more ballots wins. The condorcet winner wins every time in this method except >>>when they are not among the most approved. This is the biggest question I have however, I think that >>>the weight of Approval in this method is stronger than condorcet winning (especially in a method where >>>head-to-head tie scores are possible). The condorcet loser never wins. >>