On 06/12/2019 13.48, fdpk69p6uq@snkmail.com wrote: > Has BTR-STV been used in the real world or analyzed academically?  There > are only two sentences about it on Electowiki. Warren has written a bit about it here: https://rangevoting.org/BtrIrv.html (He states that every Condorcet method fails mono-add-top, which I don't think is true, but his example does show that BTR-IRV fails mono-add-top.) On a funny note, a dance contest program on TV here uses a judging method that could be seen as a variant of BTR-IRV (if you're willing to be very generous). It works like this: At the end of each program, the judges provide their rating of the contestants, and then the callers provide their rating by number of calls and web votes. Those are combined to produce an overall rating, and then the two bottom contestants do a repetition of their dances, during which the callers vote on who gets to continue on to the next program and who is eliminated. It's "like" BTR-IRV since there's a ranking, and the bottom two are subject to a runoff, and the loser of the runoff is eliminated. But it's a real runoff (not instant), and the initial "round" (program)'s ranking is based on ratings, not on first preferences. I don't think BTR-IRV has been used in public elections anywhere, unlike Borda-elimination (as you point out below). > I know Nanson/Baldwin methods have both going for them, and they're > Condorcet and relatively simple to explain to someone familiar with IRV: > "Instead of eliminating the candidate with the least number of > first-preference votes (which results in vote-splitting and discards > some voters' preferences), you eliminate the candidate with the worst > average ranking." > > > "In contrast to the Borda rule, our experiments with Baldwin’s and > Nanson’s rules demonstrate that both of them are often more difficult to > manipulate in practice. These results suggest that elimination style > voting rules deserve further > study." http://dx.doi.org/10.1016/j.artint.2014.07.005 James Green-Armytage's paper on strategic resistance seems to put Nanson/Baldwin around the same manipulability level as Minmax. http://jamesgreenarmytage.com/strategy-utility.pdf page 18.

> Warren has written a bit about it here: https://rangevoting.org/BtrIrv.html > > (He states that every Condorcet method fails mono-add-top, which I don't > think is true, but his example does show that BTR-IRV fails mono-add-top.) All Smith methods fail mono-add-top, but  MinMax Margins meets both Condorcet and mono-add-top. (One of it's alternative algorithms is just elect the candidate who needs the fewest extra bullet votes to be the CW.) But looking at that link I saw something different: > Like IRV, it suffers from "add-top failure > <https://rangevoting.org/AddTopFail.html>." Of course IRV meets mono-add-top.  A candidate can only be eliminated by having a shortage of top-rankings. No ballot's lower rankings can have any bearing on the result until the higher-ranked candidates have been eliminated. So if X is winning and then we add some ballots with X voted top, X cannot possibly be harmed if the method is IRV. Chris Benham On 7/12/2019 9:52 am, Kristofer Munsterhjelm wrote: > On 06/12/2019 13.48, fdpk69p6uq@snkmail.com wrote: > >> Has BTR-STV been used in the real world or analyzed academically?  There >> are only two sentences about it on Electowiki. > Warren has written a bit about it here: https://rangevoting.org/BtrIrv.html > > (He states that every Condorcet method fails mono-add-top, which I don't > think is true, but his example does show that BTR-IRV fails mono-add-top.) > > On a funny note, a dance contest program on TV here uses a judging > method that could be seen as a variant of BTR-IRV (if you're willing to > be very generous). It works like this: > > At the end of each program, the judges provide their rating of the > contestants, and then the callers provide their rating by number of > calls and web votes. Those are combined to produce an overall rating, > and then the two bottom contestants do a repetition of their dances, > during which the callers vote on who gets to continue on to the next > program and who is eliminated. > > It's "like" BTR-IRV since there's a ranking, and the bottom two are > subject to a runoff, and the loser of the runoff is eliminated. But it's > a real runoff (not instant), and the initial "round" (program)'s ranking > is based on ratings, not on first preferences. > > I don't think BTR-IRV has been used in public elections anywhere, unlike > Borda-elimination (as you point out below). > >> I know Nanson/Baldwin methods have both going for them, and they're >> Condorcet and relatively simple to explain to someone familiar with IRV: >> "Instead of eliminating the candidate with the least number of >> first-preference votes (which results in vote-splitting and discards >> some voters' preferences), you eliminate the candidate with the worst >> average ranking." >> >> >> "In contrast to the Borda rule, our experiments with Baldwin’s and >> Nanson’s rules demonstrate that both of them are often more difficult to >> manipulate in practice. These results suggest that elimination style >> voting rules deserve further >> study." http://dx.doi.org/10.1016/j.artint.2014.07.005 > James Green-Armytage's paper on strategic resistance seems to put > Nanson/Baldwin around the same manipulability level as Minmax. > http://jamesgreenarmytage.com/strategy-utility.pdf page 18. > ---- > Election-Methods mailing list - see https://electorama.com/em for list info

>> For what it's worth, there have been about 250 IRV elections in the US >> since San Francisco started using it in 2004, many of which were highly >> contested between three or more candidates, and that Burlington 2009 >> race remains the one and only one where the Condorcet candidate was not >> elected. It seems to me that if you care about the Condorcet candidate >> winning, IRV is a big step forward to that end. > A Condorcetist would probably say that this only supports his point. If > the chance that IRV is repealed is lower whenever the CW is elected, > then it's important to elect the CW, so why not ensure that the CW is > elected whenever he exists? > > (There are two not very complex ways of doing so, as this thread has > shown: Benham and BTR-IRV. So I don't really see the reason not to. Is > this about later-no-harm?) Obviously if we are only  (or perhaps just mainly) interested in electing the CW, then a Condorcet method is a bigger advance over FPP than  IRV. Later-no-harm is a big part of pro-IRV propaganda, but I don't put a big value on it in single-winner elections.  In single-winner elections I put a much bigger value on Later-no-Help, beacuse meeting that means that the method isn't vulnerable to Burial strategy. Compliance with Mono-add-Top is also nice. And I repeat that I think making sure that voters can (at least strictly) rank however many candidates they wish is more important than IRV versus Benham. Using some sort of ratings (or grading) ballot that limits the number of ranking/rating levels (to maybe as few as four) may be acceptable for some Condorcet methods, but never limiting the absolute number of candidates the voter is allowed to vote above bottom (or equal-bottom). Some people here don't like the Australian practice of declaring ballots that include any above-bottom equal-ranking to be "invalid" and not counting them towards the result. I repeat this is my recommended way of handling above-bottom equal-rankings in IRV or Benham: In any given round first we determine the fractional totals of each candidate. So ballots that equal-top rank (among remaining candidates) A and B  contribute half a vote to each of A and B. Those that vote equal-top A and B and C conrtibute a third of a vote to each of A and B and C (and so on). Then the ballots that equal top-rank more than one candidate are assigned in full to the one of those with the highest fractional total. The candidate with the lowest thus-derived total is eliminated. The point of this is to make Pushover strategising more difficult, and also it is consistent  with the "sincere STV" idea that the voter is anxious to maximise the chance that one of his/her favourites (among remaining candidates) should make it into the next round. Chris Benham On 6/12/2019 8:12 pm, Kristofer Munsterhjelm wrote: > On 06/12/2019 03.50, Greg Dennis wrote: >> Agreed. >> >> It seems like a bit of revisionist history to portray the cause of the >> repeal to be the failure to elect the Condorcet candidate. As has been >> noted, the repeal effort was led by the Republican Wright and his allies >> who felt that he should have won because he had the most first choices. >> Among the three leading candidates Wright was the Condorcet _loser_; his >> supporters wanted a plurality result, a result even further removed from >> any sense of majority. By electing Kiss, at least IRV chose a candidate >> in the smallest mutual majority set. >> >> Wright supporters didn't like the idea of a winner who in their mind >> "came in second," and you wouldn't have placated them by electing >> someone who (again, in their mind) "came in third." If anything, you >> probably would have been in a more precarious political situation. > On the other hand, when the method fails to elect the CW, it gives > opponents a group of allies almost immediately: namely, the majority who > preferred the CW to whoever *was* elected. If the CW is elected, there > is no such majority (by definition), and anti-repeal forces would have > to compose multiple majorities together. > > The point that the recall was a "hit job" kind of reinforces the point. > If Montroll was a better liked candidate in the Condorcet sense, then it > would presumably have been harder to arrange a hit against him. > >> For what it's worth, there have been about 250 IRV elections in the US >> since San Francisco started using it in 2004, many of which were highly >> contested between three or more candidates, and that Burlington 2009 >> race remains the one and only one where the Condorcet candidate was not >> elected. It seems to me that if you care about the Condorcet candidate >> winning, IRV is a big step forward to that end. > A Condorcetist would probably say that this only supports his point. If > the chance that IRV is repealed is lower whenever the CW is elected, > then it's important to elect the CW, so why not ensure that the CW is > elected whenever he exists? > > (There are two not very complex ways of doing so, as this thread has > shown: Benham and BTR-IRV. So I don't really see the reason not to. Is > this about later-no-harm?)

On 08/12/2019 03.27, C.Benham wrote: >> Warren has written a bit about it here: https://rangevoting.org/BtrIrv.html >> >> (He states that every Condorcet method fails mono-add-top, which I don't >> think is true, but his example does show that BTR-IRV fails mono-add-top.) > > All Smith methods fail mono-add-top, but  MinMax Margins meets both > Condorcet and mono-add-top. IIRC, it's an open question whether Smith is compatible with mono-add-top. Smith plus Plurality is incompatible with mono-add-top, though. (https://electowiki.org/wiki/Mono-add-top_criterion)