The need to refine ranked-choice-voting counting rules is becoming increasingly obvious as more voters learn about what fans of STAR voting call the center squeeze effect and what math-savvy folks refer to as Condorcet failures. A recent special Alaska congressional election, and a notable Burlington VT mayoral election, have demonstrated that the candidate with the fewest transferred votes is not always the least-popular candidate, especially when the top-three counting round is reached.
Eliminating pairwise losing candidates when they occur will reduce the frequency of such unexpected election results, in which the election winner lacks majority support.
The definition of a pairwise losing candidate is easy for voters to understand. It's a candidate who loses every one-on-one contest against every other continuing (not-yet-eliminated) candidate.
Eliminating pairwise losing candidates will eliminate the center squeeze effect, and will almost always eliminate Condorcet failures.
The following webpage further explains why this optional counting refinement is needed, and how it works:
https://votefair.org/rctab_add_pairwise_counting.html
I have already written lots of the code for this feature. It's in a branch at: https://github.com/cpsolver/RCTabPlus After submitting this issue I'll change the name of that branch to include the assigned issue number.
(In that same repository, the branch named feature_issue_972_count_overvote_when_single_continuing will continue to be kept separate because a pull request for it has been submitted here.)
Of course this refinement is unlikely to be merged into RCTab anytime soon. So when the software is ready I don't intend to submit a pull request for this feature.
Yet eventually a jurisdiction will wisely choose to require majority support for the winning candidate. When that happens the RCTabPlus version of RCTab will serve to exactly specify how an election-system vendor can implement this improvement that ensures the election winner is virtually always supported by a majority of voters.
The need to refine ranked-choice-voting counting rules is becoming increasingly obvious as more voters learn about what fans of STAR voting call the center squeeze effect and what math-savvy folks refer to as Condorcet failures. A recent special Alaska congressional election, and a notable Burlington VT mayoral election, have demonstrated that the candidate with the fewest transferred votes is not always the least-popular candidate, especially when the top-three counting round is reached.
Eliminating pairwise losing candidates when they occur will reduce the frequency of such unexpected election results, in which the election winner lacks majority support.
The definition of a pairwise losing candidate is easy for voters to understand. It's a candidate who loses every one-on-one contest against every other continuing (not-yet-eliminated) candidate.
Eliminating pairwise losing candidates will eliminate the center squeeze effect, and will almost always eliminate Condorcet failures.
The following webpage further explains why this optional counting refinement is needed, and how it works:
https://votefair.org/rctab_add_pairwise_counting.html
I have already written lots of the code for this feature. It's in a branch at: https://github.com/cpsolver/RCTabPlus After submitting this issue I'll change the name of that branch to include the assigned issue number.
(In that same repository, the branch named feature_issue_972_count_overvote_when_single_continuing will continue to be kept separate because a pull request for it has been submitted here.)
Of course this refinement is unlikely to be merged into RCTab anytime soon. So when the software is ready I don't intend to submit a pull request for this feature.
Yet eventually a jurisdiction will wisely choose to require majority support for the winning candidate. When that happens the RCTabPlus version of RCTab will serve to exactly specify how an election-system vendor can implement this improvement that ensures the election winner is virtually always supported by a majority of voters.