Richard, > I think of RCIPE as being like a pyramid where the foundation layer is > solid because pairwise losing candidates deserve to be eliminated > (because they must not be allowed to win). You do know that IRV already meets Condorcet Loser, right? > As an alternative to Borda counting I prefer the counting method in > Instant Pairwise Elimination: > > https://electowiki.org/wiki/Instant_Pairwise_Elimination > > "... If an elimination round has no pairwise-losing candidate, then the > method eliminates the candidate with the largest pairwise opposition > count, which is determined by counting on each ballot the number of > not-yet-eliminated candidates who are ranked above that candidate, and > adding those numbers across all the ballots. ..." That is an odd definition of "largest pairwise opposition count" which usually means the greatest number of pairwise votes against in a pairwise comparison. 49 A 24 B 27 C>B B>A 51 >49     A>C 49-27   C>B 27-24 A has the maximum pairwise opposition (51) so if we eliminate A then C wins, a failure of the Plurality criterion (because A has more first-preference votes than C has any votes). (I gather your version just behaves like IRV in this case and eliminates B,) Nonetheless that version would meet Condorcet and Mutual Majority and Majority Loser. If the claims on the page about "mathematical criteria" are true, then IRV does much better. Not on the list is the Plurality criterion, which IRV meets. Of those on the list, it meets Condorcet Loser, Majority, Majority Loser, Mutual Majority, "Cloneproof" (aka Clone Independence), Later-no-Harm, Later-no-Help, "Burying". Of those this method only meets Condorcet loser.   IRV would also meet "Resolvable" and "Polytime" based on my guess as to what they mean. But hey, near enough is good enough and IRV (but presumably not this method) is "flawed". > This method always passes the following criteria. > > * Condorcet loser > <https://electowiki.org/wiki/Condorcet_loser_criterion>: pass > * Resolvable: pass > * Polytime: pass > > This method sometimes fails the following criteria. > > * Condorcet <https://electowiki.org/wiki/Condorcet_criterion>: fail > * Majority <https://electowiki.org/wiki/Majority_criterion>: fail > * Majority loser > <https://electowiki.org/wiki/Majority_loser_criterion>: fail > * Mutual majority: fail > * Smith <https://electowiki.org/wiki/Smith_criterion>/ISDA > <https://electowiki.org/wiki/ISDA>: fail > * LIIA: fail > * IIA: fail > * Cloneproof: fail > * Monotone: fail > * Consistency: fail > * Reversal symmetry: fail > * Later no harm: fail > * Later no help: fail > * Burying: fail > * Participation: fail > * No favorite betrayal: fail > Chris B. > > *Richard, the VoteFair guy*electionmethods at votefair.org > <mailto:election-methods%40lists.electorama.com?Subject=Re%3A%20%5BEM%5D%20Poll%2C%20preliminary%20ballots&In-Reply-To=%3C8b9b8ac4-72a6-46a9-ad34-680460de3f54%40votefair.org%3E> > /Mon Apr 22 08:42:18 PDT 2024/ > > > ------------------------------------------------------------------------ > On 4/21/2024 4:33 PM, Kristofer Munsterhjelm wrote: > > What do you think of BTR-IRV in that respect? Or Borda-elimination? > > Neither explicitly checks for a Condorcet winner. > > Regarding BTR-IRV, yes it's simpler and fully Condorcet compliant. > > However, some voters will distrust the idea that the Condorcet winner is > being protected from elimination in spite of possibly repeatedly getting > the fewest transferred votes. > > I suspect that BTR-IRV inherits some odd characteristics from IRV, such > as the ones that show up in a Yee diagram and what Star fans refer to as > the center-squeeze effect. > > Adding another layer on top of a flawed method doesn't make it better. > > I think of RCIPE as being like a pyramid where the foundation layer is > solid because pairwise losing candidates deserve to be eliminated > (because they must not be allowed to win). On top of that is the IRV > layer, which is not as strong, but it's only invoked when a counting > cycle does not include a pairwise losing candidate. > > Regarding Borda-elimination, I distrust any method that's based on Borda > counting because that method is vulnerable to strategic voting and, for > fair results, requires only one mark in each row and only one mark in > each column (on a paper ballot of the kind used here in the US). > > As an alternative to Borda counting I prefer the counting method in > Instant Pairwise Elimination: > > https://electowiki.org/wiki/Instant_Pairwise_Elimination > > "... If an elimination round has no pairwise-losing candidate, then the > method eliminates the candidate with the largest pairwise opposition > count, which is determined by counting on each ballot the number of > not-yet-eliminated candidates who are ranked above that candidate, and > adding those numbers across all the ballots. ..." > > Specifically, using IPE, when a voter buries a disliked candidate, the > counting is not affected by how deep that candidate is buried. > > KM, thanks for your wise questions. > > Richard Fobes > The VoteFair guy > > > On 4/21/2024 4:33 PM, Kristofer Munsterhjelm wrote: > >/On 2024-04-22 01:11, Kevin Venzke wrote: />>/Hello, />>//>>/ From a marketability standpoint I kind of like RCIPE. If your audience />>/understands IRV then they probably can understand the concept of a />>/candidacy />>/that has become futile during the count, who can be foretold to be a />>/loser />>/well in advance of the end. />>//>>/To me it is not that the pairwise losing candidate "deserves to be />>/eliminated" but that within the logic of IRV it's intuitive that that />>/candidate shouldn't need to play a role, affecting things, if they're />>/doomed />>/to lose. (Or perhaps it's just me who thinks that's intuitive.) />>//>>/In contrast the "beats all" winner concept would be a bridge too far, />>/because while there might be a candidate who can win every final > pairing, />>/IRV imposes additional requirements to get to that point, so nothing is />>/assured about that status. />>//>>/But if the audience doesn't know IRV then it would be harder for me to />>/find />>/an argument for RCIPE. />//>/What do you think of BTR-IRV in that respect? Or Borda-elimination? />/Neither explicitly checks for a Condorcet winner. />//>/-km / >