[EM] Reply to Rob regarding RCV

[Forest Simmons]

Elimination with take down reliably punishes the (most likely) cycle
instigating faction by electing their least favorite of the three members
of the artificial top cycle.

However we can go beyond that with a two stage manual runoff that asks (in
the first stage) if they would rather have a runoff between the other two
members of the cycle. If most of them say no, then the election result
stands. Otherwise, the two candidate runoff takes place.

It can be shown that optimal strategy by rational voters aware of the other
voters true preferences will always elect the sincere CW if there is one.

Otherwise (i.e. in the rare case of a sincere cycle) rational voters will
prefer the first offered candidate over a runoff between the other two.

To fully take advantage of this fact, instead of a choice between the RCV
winner and a runoff between the other two cycle members ... make the first
choice between the classical wv Condorcet winner and a runoff between the
other two top cycle members.

Rational voters will choose the sincere CW when there is one, otherwise the
cycle will be broken at the weakest wv link.

fws



On Sat, Sep 23, 2023, 10:22 PM Michael Ossipoff <email9648742 at gmail.com>
wrote:

>
>
> On Sat, Sep 23, 2023 at 14:52 Forest Simmons <forest.simmons21 at gmail.com
>
> Having returned home, I can resume what I was saying:
>
> As with other pairwise count methods, with Benham, the CW’s voters can, by
> refusing to rank anyone they don’t approve, deny support to inimical
> offensive order-reversers.
>
> But with wv, they need only weaken the defeat some. With Benham, they’d
> have to outright prevent the defeat.
>
> …& Benham doesn’t have wv’s truncation-proofness.
>
> But you (Forest) pointed out that the rankings in the examples of wv’s
> defense-successes are consistent with preferences other than those that I
> assumed, & therefore that maybe an exam of wv you defending against one
> kind of attack, wv could be interpreted as an example of wv allowing
> another kind of attack.
>
> Until I or someone else thoroughly examine that situation, of course I
> shouldn’t be proposing the wv methods or the other Condorcet completions
> that work similarly (such as some of the Approval completions of Pairwise
> Count).
>
> But in the meantime, & regardless of how the wv situation turns out,
> Sequential-Pairwise is proposable, because it prevents or deters offensive
> strategy in a different way, via unpredictability.
>
> But at least until the wv situation has been thoroughly examined &
> vindicated (it might not be), I shouldn’t claim anything ab wv.
>
> So, I found fault with Benham, but you may have shot-down wv.  :-)
>
> Michael
>
> voters might minimize support for the order reversal attack I described,
> by refusing to rank anyone whom they don’t approve. But in wv it’s only
> necessary to reduce the strength of the strategists’ defeat of whom they
> beat & vote over CW.
>
> But with Benjamin, those defensive voters would have to actually prevent
> the strategists’ candidate from beati
>
>>
>>
>> On Sat, Sep 23, 2023, 11:19 AM Michael Ossipoff <email9648742 at gmail.com>
>> wrote:
>>
>>> Hi Rob—
>>>
>>> …
>>>
>>> You wrote:
>>>
>>> …
>>>
>>> [quote]
>>>
>>> Oh Michael...where do I begin?  Your apparent move to the dark side
>>> makes me sad.  I realize that this intro may sound condescending, but I
>>> truly don't mean it that way.  I deeply respect your opinion. YOU were the
>>> one who taught me about "center squeeze" in 1995 or so, and made me rethink
>>> AV/PV/IRV/RCV (or whatever the name of the week is).
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> Yes, earlier I was quite critical of RCV (called IRV in those days,
>>> before San Francisco insisted on RCV, because “instant” was misleading
>>> people to expect an instantaneous count).
>>>
>>> …
>>>
>>> I was wrong.
>>>
>>> …
>>>
>>> You know that often the relative merits, advantages & disadvantages of
>>> single-winner methods aren’t what they at first seem. Never be prematurely
>>> sure that someone is wrong about such matters.
>>>
>>> …
>>>
>>> But I would like to remind you that I also continue to say that
>>> Condorcet, in its best versions, is my favorite, because, always electing
>>> the CW, it best gets rid of the lesser-of-2-evils problem (LO2E) for any &
>>> every  kind of voter, thereby accommodating even the most timid LO2E
>>> giveaway voter.
>>>
>>> …
>>>
>>> As I said, RCV only works & has merit if the voters aren’t timid
>>> overcomromising LO2E giveaway voters.
>>>
>>> …
>>>
>>> I don’t deny that Condorcet-failure is a disadvantage, but, with a good
>>> electorate, it doesn’t matter. I used to say that Approval’s equal-givaway,
>>> amounting to an abstention between Favorite & Lesser-Evil, is better than
>>> RCV’s favorite-burial…but not if you have an electorate that won’t do the
>>> burial !!!
>>>
>>> …
>>>
>>> Approval & STAR encourage Mr. Timid to do giveaway. RCV encourages
>>> everyone to be frank, honest, ambitious, hopeful !!!
>>>
>>> …
>>>
>>> …because, as I said, that won’t be a problem, because an electorate that
>>> has enacted RCV didn’t do so because they want & intend to vote a
>>> hold-you-nose lesser-evil over their favorite (They can & do do that now,
>>> in Plurality). If they enact RCV it’s because they want & intend to
>>> sincerely rank the candidates, expressing & fully-supporting their favorite.
>>>
>>>
>>> …
>>>
>>> …& so THEY WILL DO SO !!!
>>>
>>> …
>>>
>>> So don’t worry about LO2E strategy in RCV voting.
>>>
>>> …
>>>
>>> Anti-RCVists often say that RCV doesn’t really honor majority.
>>>
>>> …
>>>
>>> But, as I’ve been saying, though RCV doesn’t meet the Condorcet
>>> Criterion, it meets the Mutual-Majority Criterion (MMC).  Let me state
>>> an improved & expanded definition of MMC:
>>>
>>> …
>>>
>>> MMC:
>>>
>>> …
>>>
>>> If there are 1 or more sets of candidates such that a majority of the
>>> voters prefer the candidates of that set to everyone outside the set, then
>>> the winner should come from such a set.
>>>
>>> …
>>>
>>> (Then it goes without saying that, when there’s a Mutual-Majority
>>> (defined below) the winner will come from that Mutual-Majority.)
>>>
>>> …
>>>
>>>  [end of MMC definition]
>>>
>>> …
>>>
>>> A majority who all prefer some set of candidates to everyone outside
>>> that set, I call an Agreeing-Majority.
>>>
>>> …
>>>
>>> A majority who all prefer *the same* set of candidates to everyone
>>> outside that set, I call a  Mutual-Majority.
>>>
>>> …
>>>
>>> (Arguably a weaker definition of a Mutual-Majority would do: A majority
>>> who all prefer all of their favorites to everyone outside the set of all
>>> their favorites.)
>>>
>>> …
>>>
>>> When there are 1 or more Agreeing-Majorities, RCV always elects the
>>> candidate of the largest faction of  an Agreeing-Majority.
>>>
>>> …
>>>
>>> i.e. under those conditions, RCV always elects the favorite of an
>>> Agreeing-Majority.
>>>
>>> …
>>>
>>> RCV always elects the candidate of the largest faction of the
>>> Mutual-Majority.
>>>
>>> …
>>>
>>> i.e. RCV always elects the favorite of the Mutual-Majority.
>>>
>>> …
>>>
>>> That candidate isn’t an unpopular extremist, but instead has strong
>>> genuine majority coalition support, as defined above.
>>>
>>> …
>>>
>>> [quote]
>>>
>>> I just think you're incorrect about FairVote.
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> I didn’t say anything about FairVote.
>>>
>>> …
>>>
>>> Whether you like FairVote or not, that has no bearing on the merits of
>>> RCV.
>>>
>>> …
>>>
>>> FairVote, from the start, has insisted on offering the traditional RCV.
>>> We should respect that choice. RCV has about a century of precedent in
>>> Australia & Ireland. Proposing a traditional method with long precedent is
>>> a valid practical choice, & one that we should respect.
>>>
>>> …
>>>
>>> …& that proposal has been enormously successful, & is sweeping this
>>> country. Maybe its century of traditional precedent has something to do
>>> with its success.
>>>
>>> …
>>>
>>> When RCV was  initially being adopted, of course there were no
>>> computers, & so Condorcet’s complete pairwise-count would be infeasible in
>>> a large election. Sure, the Sequential-Pairwise (SP) pairwise-count only
>>> needs about twice as much vote-counting as RCV.
>>>
>>> …
>>>
>>> (Approval, Score, STAR & RCV all need roughly the same amount of
>>> vote-counting (they all vary greatly), & SP needs about twice as much.)
>>>
>>> …
>>>
>>> Maybe people didn’t want twice as much vote-counting. Or maybe they were
>>> afraid that SP would be rejected because of its Pareto violation (which I
>>> consider irrelevant, just like MinMax’s Condorcet-Loser violation).
>>>
>>> …
>>>
>>> Then there’s Coombs, which I guess would have about the same amount of
>>> vote-counting as RCV. But maybe they didn’t like Coombs because things
>>> could get ridiculous, like when I nominate Dracula in the primary, so that
>>> we can bury the Democrat under Dracula.
>>>
>>> …
>>>
>>> …but of course there are worse things than ridiculous.  Maybe we
>>> haven’t been fair to Coombs.
>>>
>>> …
>>>
>>> As I said, I prefer Condorcet, in its best versions, but it’s RCV that
>>> has the activist movement, big well-funded national organization,
>>> lobbyists, experienced & active campaign-managers, & big successes all
>>> around this country.
>>>
>>> …
>>>
>>> …with (I’ve read) on the order of 60 municipalities & 2 states having
>>> adopted RCV.
>>>
>>> …
>>>
>>> If RCV, & not Condorcet, is succeeding, we Condorcetists have nothing to
>>> complain about. When the RCVists were doing the work, we weren’t out there
>>> enacting anything.  Don’t blame the RCVists for that..
>>>
>>> …
>>>
>>> We should acknowledge, commend & appreciate what the RCV organization
>>> has accomplished.
>>>
>>> …
>>>
>>> [quote]
>>>
>>> RCV is already poorly understood.
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> RCV is enormously popular with progressives & progressive organizations
>>> & parties, such as the GPUSA, the U.S. Greens.   …because they
>>> understand that rank-balloting will allow them to express all of their
>>> preferences among the candidates.
>>>
>>> …
>>>
>>> …& because they’ve been correctly informed that RCV has genuine strong
>>> majority properties, when it coalesces the Mutual-Majority.  ..even if
>>> they haven’t heard about the details of those properties.
>>>
>>> …
>>>
>>> As for RCV’s definition, RCV can be defined very briefly, in one
>>> sentence:
>>>
>>> …
>>>
>>> Repeatedly eliminate the candidate who tops fewest rankings, till
>>> someone tops most of them.
>>>
>>> …
>>>
>>> [end of brief RCV definition]
>>>
>> Brief definition of Benham:
>>
>> Repeatedly eliminate the candidate who tops the fewest rankings until
>> someone tops most of them in every head-head matchup (between it and any
>> other uneliminated candidate).
>>
>> fws
>>
>>> …
>>>
>>> [quote]
>>>
>>> When I moved to San Francisco in 2011, I expected to grudgingly like
>>> voting in RCV elections, and I expected to enjoy ranking my choices   What
>>> I found instead was that very few people here understand how votes are
>>> counted
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> They would if they heard RCV’s brief definition.
>>>
>>> …
>>>
>>> [quote]
>>>
>>> , and many folks in my lefty political tribe here take great pride in
>>> their ignorance of math and the inner workings of their electoral system,
>>> trusting that the powers-that-be will count things correctly.
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> …& they’re right, when the method is RCV.  (…& likewise would be, with
>>> a good Condorcet version too.)
>>>
>>> …
>>>
>>> [quote]
>>>
>>> As "exhibit A", I will point to the recent clown show in Alameda County
>>> (i.e. just a few miles east of me, on the other side of a puddle known as
>>> the "San Francisco Bay"):
>>>
>>> …
>>>
>>>
>>> https://www.sfchronicle.com/bayarea/article/Alameda-County-admits-tallying-error-in-17682520.php
>>>
>>> …
>>>
>>>  It would seem that they had been counting RCV elections wrong for
>>> DECADES, and only noticed the problem in 2022.  Simplicity and precinct
>>> summability matters.
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> RCV’s brief definition is simple. RCV is simple, natural & obvious.
>>>
>>> …
>>>
>>> There’s a lot of mis-statement about “Precinct-Summability”, &
>>> questionable-ness about what “Precinct Summability” is supposed to mean..
>>>
>>> …
>>>
>>> Plurality, Approval & Score:
>>>
>>> …
>>>
>>> Candidates’ vote totals are summed in each precinct & sent to a central
>>> count-place, where there’s a central tabulation.
>>>
>>> …
>>>
>>> STAR:
>>>
>>> …
>>>
>>> Same, at first. Then the central counting place determines the two top
>>> scorers, & then presumably sends that information to the precincts, which
>>> still have the rankings, & the precincts each total the pairwise votes for
>>> each of the finalists over the other, & they all send that back to the
>>> central location, where the results are summed & the final winner reported.
>>>
>>> …
>>>
>>> Condorcet:
>>>
>>> …
>>>
>>> Each precinct counts the preference votes for A over B, & for B over A,
>>> for each of the N(N-1)/2 pairs {A,B}.  …& those totals are sent in to
>>> the central location, where the winner is determined according to the rules
>>> of whatever Condorcet version is being used.
>>>
>>> …
>>>
>>> RCV:
>>>
>>> …
>>>
>>> Each precinct counts the top-count score of each candidate, & sends that
>>> in to the central location.
>>>
>>> …
>>>
>>> The central location totals that count for each candidate, to determine
>>> which has lowest top-score, & sends that information back to the precincts.
>>>
>>> …
>>>
>>> Each precinct eliminates that candidate from its rankings, & repeats the
>>> first line above.
>>>
>>> …
>>>
>>> Repeat till the central location finds that one candidate tops most of
>>> the rankings.
>>>
>>> ---
>>>
>>> Notice something similar about those?  Every one of those procedures
>>> requires counts at the precincts, & also at the central location, &
>>> communication between the two.
>>>
>>> …
>>>
>>> How is RCV different?  It does such a procedure a number of times.
>>> That’s it. That’s the difference.
>>>
>>> …
>>>
>>> So it’s questionable regarding what is this “Precinct Summability” that
>>> Plurailty, Approval, Score, STAR & Condorcet all have, but RCV allegedly
>>> doesn’t have.
>>>
>>> …
>>>
>>> The same security measures, precautions & audits can be done with RCV as
>>> with any of the other methods whose procedures are described above.
>>>
>>> …
>>>
>>> [quote]
>>>
>>> [quote]
>>>
>>> Strategy-evaluation for Condorcet-complying pairwise-count methods has
>>> proven to be complicated & more difficult than one would expect.
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> This I will agree with. That is why I've hopped on the approval voting
>>> bandwagon for single-winner reform.
>>>
>>> [/quote]
>>>
>>> …
>>>
>>> That’s a bit hasty. Undeniably some of the pairwise-count
>>> Condorcet-compliant methods thwart &/or deter offensive strategy so well
>>> that it won’t be a problem, & the election of the sincere-CW will virtually
>>> always be elected.
>>>
>>> …
>>>
>>> There’s the argument that there are so many good Condorcet versions that
>>> choosing between them is prohibitively problematic, preventing the adoption
>>> of any of them.
>>>
>>> …
>>>
>>> No, several of the best versions can be offered to a proposal-committee,
>>> & it can discuss & evaluate them & then choose a proposal.
>>>
>>> …
>>>
>>> …& there are a few obviously simplest proposals, making the choice a lot
>>> less complicated & difficult than antii-Condorcetists claim:
>>>
>>> …
>>>
>>> MinMax:
>>>
>>> …
>>>
>>> Elect the candidate whose greatest defeat is the least.
>>>
>>> …
>>>
>>> (…implying the election of an unbeaten candidate when there is one ( as
>>> there nearly always is) ).
>>>
>>> …
>>>
>>> Condorcet//Approval:
>>>
>>> …
>>>
>>> Your ranking is counted as approving everyone you rank. If no one is
>>> pairwise-unbeaten, then elect the most approved candidate.
>>>
>>> …
>>>
>>> Majority-Defeat Disquaification//Approval (MDDA):
>>>
>>> …
>>>
>>> Your ranking is counted as approving everyone you rank. Elect any
>>> unbeaten candidate. If there are none, then every majority pairwise-beaten
>>> candidate is disqualified, & the un-disqualified candidate with most
>>> approvals is elected.
>>>
>>> …
>>>
>>> (I’ve added a Condorcet-winner electing clause, because I now feel that
>>> Condorcet’s failure-betrayal scenario is so rare & unpredictable as to be
>>> irrelevant to strategy.)
>>>
>>> …
>>>
>>> Sequential-Pairwise:
>>>
>>> …
>>>
>>> Order the candidates in a list such that the ones topping more rankings
>>> are listed below the ones topping fewer rankings.
>>>
>>> …
>>>
>>> Find the pairwise-winner among the top 2 candidates in the ordering.
>>> Then find the pairwise winner between that winner & the next candidate down
>>> the list.  Condtinue until you’ve found the winner of the then-current
>>> winner & the last candidate in the list. S/he wins.
>>>
>>> ----------
>>>
>>> I’d offer those 4 simple pairwise-count versions.
>>>
>>> …
>>>
>>> I’d also offer Approval, in case the jurisdiction either couldn’t
>>> afford, or didn’t want to spend for, rank-balloting equipment & software.
>>>
>>> …
>>>
>>> I’d also offer RCV & STAR, because some proposal-committee members might
>>> prefer them.
>>>
>>> …
>>>
>>> I’d personally propose & justify the choice of a Pairwise-Count, with
>>> RCV as next choice, & of course Approval if rank-balloting is infeasible or
>>> rejected.
>>>
>>> …
>>>
>>> If there were a ranked vote in the proposal committee, of course my
>>> ranking would be:
>>>
>>> …
>>>
>>> 1. Some ordering of the above-listed four Pairwise-Count methods
>>>
>>> …
>>>
>>> 2. RCV
>>>
>>> …
>>>
>>> 3. Approval if ranked-balloting is infeasible or rejected.
>>>
>>> -----
>>>
>>> Yes, I like Approval. Under different conditions, normal conditions, it
>>> would be my suggestion, though STAR would then be okay too.
>>>
>>> …
>>>
>>> But our voting-conditions are anything but normal.
>>>
>>> …
>>>
>>> We always have a sleazy & corrupt POS “lesser”-evil, & a
>>> dramatically-horrifying greater-evil.  (…& some better candidates who,
>>> our media insist, are unwinnable, minor, not-serious, candidates.)
>>>
>>> …
>>>
>>> We’re constantly told that the greater evil is the only bad result, & so
>>> we have to support the “lesser-evil” against the greater-evil (…& against
>>> our favorite).
>>>
>>> …
>>>
>>> Always.
>>>
>>> …
>>>
>>> It’s always this same discouraging, dismal & hopeless situation.
>>>
>>> …
>>>
>>> That isn’t normal.  It certainly isn’t natural.
>>>
>>> …
>>>
>>> So yes, I’d suggest Approval under normal conditions. But our highly,
>>> bizarrely, abnormal & unnatural conditions require special methods to deal
>>> with the (allegedly) hard choice between genuinely-wanted outcomes, &
>>> odious dismal regrettable & deplorable “lesser” evils.
>>>
>>> …
>>>
>>> For these conditions, we need something more powerfully-discriminating:
>>> a rank-method, to let everyone vote all their preferences among as many
>>> candidates as they want to.   …to elicit & count genuine favoriteness  immediately.
>>>
>>>
>>> …
>>>
>>> (…not after several election-cycles, because a lot of harm can be done
>>> in 4 years or 8 years.)
>>>
>>> …
>>>
>>> It’s like when, in _The Godfather_, Michael says to his attorney:
>>>
>>> …
>>>
>>> “No, you’re a peacetime consiglieri, but right now we need a wartime
>>> consiglieri.”
>>>
>> ----
>>> 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/20230923/e86c60be/attachment-0001.htm>