[EM] The Equal Vote Coalition and robla

[Richard]

On 4/18/25 19:32, Rob Lanphier via Election-Methods wrote:
 > Having voted in many San Francisco elections (home of the name "Ranked
 > Choice Voting" to refer to IRV), I was surprised at how awful I've found
 > the actual task of voting here.  I expected to have nerdy fun with it,
 > but voting in the RCV elections has become a chore that I kinda dread.
 > I wouldn't dislike RCV as much if I didn't have to worry about
 > accidentally spoiling my ballot by assigning two candidates the same
 > rank.  ...


Fortunately there are two things we can do to eventually eliminate this 
frustration, which I too experienced in November in Portland's mayoral 
election.  (It was not an issue in the STV city-council election.)

Before listing my two suggested actions, here are some important 
clarifications:

* I have come to realize the source of this limitation is the "RCTab" 
software at the Ranked Choice Voting Resource Center (RCVRC).  If their 
RCTab software had allowed the option to correctly count so-called 
"overvotes," I and other local election-method advocates would have 
requested that option for Portland's new election system.

* Currently the RCTab software only offers two ways to handle two 
candidates at the same rank: skip that column of ovals (and resume 
counting at the next column that has just one mark), or dismiss all the 
remaining marks on that ballot when the counting reaches this marking 
pattern.

* When "equal ranks" are allowed, and correctly counted, ballot marking 
will become much easier.

* The promoters of STAR voting claim that ranked choice ballots are 
difficult to use.  Yet this difficulty occurs mostly because U.S. 
implementations of IRV do not yet count such "overvote" marks correctly. 
  (The other significant disadvantage can be overcome by eliminating 
pairwise losing candidates when they occur.)

* When software developers write IRV code to correctly count overvotes, 
usually they use decimal numbers, such as giving one ballot half 
strength for one equal-ranked candidate and half strength for the 
voter's other equal-ranked candidate.  This shortcut yields the correct 
result, in the sense that the correct candidate gets elected.  However 
this decimal-number shortcut is not acceptable for governmental elections.

* Here's how to correctly count overvotes in governmental elections, 
without departing from the principle of one voter supporting only one 
candidate in each round of voting.  First, imagine IRV is being used in 
a huge convention hall filled with all the voters and all the 
candidates.  Each voter lines up behind their first-choice candidate. 
The candidate with the shortest line is eliminated.  The voters in that 
shortest line then move to stand in line behind their second choice. 
(All other voters stay where they are.)  Now imagine there were two 
voters in that shortest line who have an equal preference for candidates 
B and D as their second choice.  These two voters agree that one of them 
will stand in line behind candidate B, and the other voter will stand in 
line behind candidate D.  This "pairing" approach easily extends to five 
candidates being marked in the same "rank" column.  (More than five 
"equal ranks" can be calculated, but not as easily, and not as fast.)

* Mathew Ruberg is the person in charge of the RCTab software at RCVRC. 
I've sent to him a detailed description for how to correctly count 
overvotes.  He says my request has been added to an informal list of 
requests, but it can't progress to their official list of requests.  He 
says that RCVRC needs to receive requests from election officials before 
they will begin to research this request.  Then, depending on the 
research results, they might add this request to their official to-do list.

* I suggested to Portland's election officials that they contact Mathew 
Ruberg with this request.  I don't think they did so (because I didn't 
get any reply from anyone).  That's understandable.  The election 
officials were overwhelmed with lots of other tasks after their first 
ranked-choice-voting election.

* Oregon's recent Measure 117, which would have adopted ranked choice 
voting for some Oregon elections, dealt with this issue by not 
mentioning how "overvotes" should be counted.  (I'm the person who 
requested in verbal testimony on an earlier bill that "voters must be 
allowed to rank two or more candidates at the same preference level.") 
If this measure had passed, this flexibility would have allowed election 
officials, or the Secretary of State, to choose how overvotes are 
counted, without needing to change any election rules.

* The promoters of STAR voting claim that ranked choice voting cannot 
correctly count overvotes, and that IRV cannot correctly count 
overvotes.  Regardless of whether "ranked choice voting" means IRV or 
any method that uses ranked choice ballots, this criticism is intended 
to block all uses of ranked choice ballots, regardless of how they are 
counted, even if the counting method is a Condorcet method.  This is a 
serious misrepresentation about "ranked choice voting" (regardless of 
which definition is implied).

With these clarifications in mind, here are the two action items I suggest:

* If anyone knows how to contact election officials in San Francisco, or 
anywhere else where the IRV version of ranked choice voting is used, 
please ask them to contact Mathew Ruberg to express interest in refining 
the RCTab software to correctly count overvotes so that this option can 
be adopted sometime after it becomes available.  Here's the webpage that 
encodes the email address for Mathew Ruberg:

https://www.rcvresources.org/mathew-ruberg

* Rob, or anyone else who interacts with the Equal Vote Coalition, 
please inform them that IRV counting rules can correctly count so-called 
overvotes.  Based on past actions it's likely some of them are still 
under the influence of STAR advocates who have falsely claimed that IRV 
does not allow a voter to assign two candidates at the same rank.  It's 
crazy that an organization named "Equal Vote" does not seem to realize 
that ranked choice voting, including pairwise-counted ranked choice 
voting, does not have all the disadvantages claimed by STAR voting 
promoters.

Yes, IRV doesn't always elect the correct candidate.  However, 
continuing to take the same counting shortcut that Australian election 
officials adopted more than 100 years ago is giving a bad reputation to 
all ranked choice voting methods, including Condorcet methods.

Richard Fobes
The VoteFair guy


On 4/18/25 19:32, Rob Lanphier via Election-Methods wrote:
> Hi Robert,
> 
> Due to a delivery glitch, I didn't get some of the replies to my EM 
> emails right away, including this one.  However, that's not the only 
> reason for the delay.  I've been thinking about how to reply for a 
> while.  I've even been having ChatGPT help me out.  More inline....
> 
>   On Thu Apr 3 12:58:59 PDT 2025, robert bristow-johnson wrote:
> 
>     On 04/03/2025 2:36 AM EDT Rob Lanphier <roblan at gmail.com <http://
>     gmail.com>> wrote: 
> 
>      > I joined the board of the Equal Vote Coalition in January 2025:
> 
>  > > https://www.equal.vote/about <https://www.equal.vote/about>
> 
>     [...]
>     I have mixed feelings about Equal Vote Coalition.  I am jealous of
>     the name and the domain name.  I think that name belongs strictly to
>     Condorcet advocacy.
> 
> 
> That's nice.  You should have registered it  :-P
> 
> You continued...
> 
>     I still consider EVC to be STAR advocates.  I don't consider STAR
>     (or any cardinal method) to be "equal votes".
> 
> 
> Admittedly, EVC is pretty heavily biased toward STAR advocates.  I 
> realize you're not interested in working with anyone who prefers 
> cardinal methods to ordinal methods, but some of us are ready to 
> compromise.  During the vetting process when I was being brought on 
> board, it was clear EVC wanted me because I might help provide more 
> depth in Condorcet advocacy, and I suspect they'd accept others with 
> even more depth of knowledge and Condorcet activist chops than me.  When 
> I was in my "job interview" for the board, I made no secret to them or 
> others about the fact that Condorcet methods were my first love, and I 
> still trot out my article from The Perl Journal to show that I've really 
> liked Condorcet methods for a while:
> https://robla.net/1996/TPJ <https://robla.net/1996/TPJ>
> 
> I still consider the Condorcet winner criterion very important.  But, as 
> I've pointed out many times on this list and elsewhere, I've found the 
> simulations published by folks like Ka-Ping Yee persuasive 
> demonstrations that approval voting elections and Condorcet elections 
> are very likely to result in the same candidate winning.  Approval 
> voting is the only pure cardinal system that I can get behind (so, for 
> example, I still find pure score/range voting a very poor choice for 
> high-stakes public elections).  Approval's strategy (and most cardinal 
> method strategy) is more complicated than most cardinal advocates would 
> have you believe, but it's not THAT bad.  At least, approval isn't that bad.
> 
> Having voted in many San Francisco elections (home of the name "Ranked 
> Choice Voting" to refer to IRV), I was surprised at how awful I've found 
> the actual task of voting here.  I expected to have nerdy fun with it, 
> but voting in the RCV elections has become a chore that I kinda dread.  
> I wouldn't dislike RCV as much if I didn't have to worry about 
> accidentally spoiling my ballot by assigning two candidates the same 
> rank.  I'm assuming/hoping BTR-IRV doesn't suffer from that same 
> problem, but don't assume that folks who live in an RCV/IRV jurisdiction 
> love it.  I've found many folks here who find it annoying, but I haven't 
> (yet) gotten most of them annoyed enough to make that the focal point of 
> their activism.
> 
> I'm not pushing STAR here in SF, but I'm not going to badmouth it the 
> way you do, Robert. I do badmouth RCV/IRV sometimes, and I've probably 
> alienated a person or two. The reason why I'm at peace with 
> STAR, though: it's not a pure cardinal system.  That final pairwise 
> comparison makes it difficult to come up with plausible scenarios where 
> STAR and Condorcet disagree.  I know you disagree, but I'm guessing you 
> haven't tried as hard as I have to come up with plausible scenarios 
> where STAR really fails.  The funny thing is that most of the times that 
> STAR and Condorcet disagree, it's because of a Condorcet cycle (which 
> anecdotal evidence suggests could happen as frequently as IRV<- 
>  >Condorcet disagreements, though I'm only vaguely aware of two or three 
> cycles in real-world elections).
> 
> The more I think about STAR vs strictly CWC-compliant systems, and the 
> more I try to get ChatGPT and other LLMs to make the best case they can 
> for you, I can't get a convincing case that the distinction is a reason 
> to divide into warring factions over.  Then when I ask ChatGPT to "find 
> Robert Bristow-Johnson's best case against STAR", I remain unconvinced 
> by what I've read from you (e.g. your post and comments here):
> https://www.reddit.com/r/EndFPTP/comments/1bbq6hu/ 
> heres_a_good_hypothetical_for_how_star_fails/ <https://www.reddit.com/r/ 
> EndFPTP/comments/1bbq6hu/heres_a_good_hypothetical_for_how_star_fails/>
> 
> You seem to be pushing voter behavior that's too uniform.  As /u/ 
> arendpeter <https://reddit.com/u/arendpeter> pointed out in their 
> findings, with the online polls they've posted, folks were a lot more 
> generous with 3 and 4 star ratings than you indicate they will be:
> https://www.reddit.com/r/EndFPTP/comments/1bbq6hu/ 
> heres_a_good_hypothetical_for_how_star_fails/kvbsc4j/ <https:// 
> www.reddit.com/r/EndFPTP/comments/1bbq6hu/ 
> heres_a_good_hypothetical_for_how_star_fails/kvbsc4j/>
> 
> As I speculate on my own behavior under STAR, I think of how I would 
> have voted here in SF in the 2024 mayoral race, and I'm pretty sure I 
> would have been more generous with my stars, even as a very partisan 
> voter.   I would have given a couple mayoral candidates 5 stars (because 
> I knew they wouldn't win, but so I could say "I voted for you", since 
> meeting any of them F2F isn't out of the question for me), I would have 
> given my favorite viable candidate either 5 stars or 4 stars, I would 
> have given my next favorite 3 stars.  I doubt I would have given any of 
> the others any stars.  This maps pretty closely to my RCV/IRV ballot.
> 
> As it turns out, my vote in 2024 didn't seem to matter much (in most 
> elections).  I think my bottom ranked candidate for mayor won, and he's 
> turning out to be a much better mayor than I feared he would be.  Maybe 
> San Francisco needs a political outsider after all.
> 
>     I also consider STAR or Score to have absolutely *zero* chance of
>     ever being adopted for governmental elections.  And while Score is
>     simple in concept, STAR, being a hybrid, is a little bit contrived.
> 
> 
> Saying "absolutely *zero* chance"  seems like ragey hyperbole.  And 
> honestly, given how hard you’re tilting at the BTR-IRV and Condorcet-TTR 
> windmills, it seems ironic.  I think STAR is a clever compromise which 
> seems like it will produce reasonable outcomes, and I've found it seems 
> to attract more enthusiasm than Condorcet RCV.  In fact, I've never seen 
> activists as excited about any strict Condorcet system as I have about 
> STAR.  I'm assuming you've heard of Arrow's theorem and Gibbard's 
> theorem; so I'm sure you're aware we're not going to arrive at 
> perfection.  Neither STAR nor Condorcet systems are as easy to explain 
> as approval is in my experience.
> 
>     We really should concentrate on reforming RCV to be Condorcet
>     consistent.  That's what EVC should be doing.  Personally, I don't
>     understand how or where EVC is getting their money to operate and
>     support as many employees as they do.
> 
> 
> I'm not about to tell EVC to drop advocacy for STAR and approval, and 
> It's not like EVC is rolling in money.  Sass isn't at EVC anymore, and 
> in general, you may be overestimating the number of people that get a 
> paycheck from them.  The reason they have such a high profile is that 
> Sara and crew are hard workers, and they managed to recruit a lot of 
> volunteers that also do amazing work for them.  Folks are willing to get 
> behind the legal entity and associated leadership, and they recruit 
> volunteers and they've taken the risks necessary to get significant 
> local news coverage.  They haven't had a lot of success at the ballot 
> box yet, but it seems easier to take potshots at them than it is to 
> actually do the work they've done.
> 
>     About "Ranked Robin" that's seems to me to be a Sass invention and I
>     don't understand why they tepidly promote that instead of just
>     Condorcet RCV in general.
> 
> 
> I don't either, to be honest.  I think part of the problem is that we 
> keep having fights to the death over naming.  "Ranked Robin" is a clever 
> name, but it's over 200 years too late to rename the Condorcet winner 
> criterion, and wastes political capital with the academic community.  
> They'll have an easier time revising their website than the academic 
> world will have revising all of the papers that have been written over 
> the years about Condorcet.  You're right to point out that "Ranked 
> Robin" may get mixed up with Sass's specific method outlined on 
> electowiki rather than the wider universe of pairwise-comparison methods 
> (like Tideman's "Ranked Pairs").  I'm not saying EVC should completely 
> abandon the name "Ranked Robin" and it's not clear that EVC needs to 
> even deprecate the name, but rather, should establish it as a 
> complementary name for describing Condorcet methods (and maybe, just let 
> Sass have the name for their specific Condorcet variant).
> 
> It would seem that some folks believe that "Condorcet" is a bad brand, 
> which it isn't.  Condorcet was a good guy, and though some folks think 
> that systems that are strictly CWC-complaint are too complicated, many 
> of those same people underestimate the complexity of the alternative 
> they promote.  We can have honorable disagreements about which election 
> method criterion we believe is the most important, but thanks to Arrow's 
> and Gibbard's theorems (not to mention non-mathematical considerations), 
> we're going to have to compromise on SOMETHING.
> 
>     And I am completely in disagreement with Sara Wolk about conceding
>     the label "Ranked-Choice Voting" to only the Hare method.  Borda,
>     Bucklin, and Condorcet are also RCV.  We must not allow FairVote to
>     appropriate the term "RCV" to mean *only* IRV.
> 
> 
> Living in SF, I've learned to assume that anyone who I know doesn't 
> understand voting systems is referring to IRV when they say RCV.  In 
> fact, almost everywhere on the Internet, that holds true too.  Maybe 
> that won't hold true a decade from now, but I'm not about to fight that 
> battle right now.  I think it's totally reasonable from a search-engine 
> optimization perspective for someone building a website to use "RCV" to 
> mean "IRV".  Note that SF is the home of the term "ranked-choice voting" 
> to mean "Australia-style preferential voting".  Also note that many 
> people, when pressed, think we're using Borda, but I'm pretty sure that 
> Burlington voters (and activists) aren't any smarter than SF voters and 
> activists.
> 
>  > > Most folks with the organization feel that STAR voting is superior 
> to Condorcet methods.
>  >
>  > For no good reason.  I guess it's their child, as ugly as it is.
> 
> STAR is fine.  I find it VERY difficult to come up with differences 
> between STAR and Condorcet that are likely to actually show up in the 
> real world.
> 
>  > > I feel like score/range voting requires too much work from voters 
> to figure out the best strategy (which, as near as I can tell, is either 
> vote the maximum score or the minimum; hence transforming it into 
> approval voting in practice).
> 
>     I had shown before that *with* the assumption that if a voter's
>     preference is A>B>C, they're gonna mark their STAR ballot A:5, B:1,
>     C:0 because that will best accomplish their political goal in nearly
>     all circumstances.
> 
> 
> Your 5-1-0 strategy assumes only three candidates.  I don't like score/ 
> range, because sophisticated voters will vote either the full score 
> (max) or nothing (min).  Yes, I suppose there may be sophisticated 
> strategies that can be used with STAR, but ONLY with incredibly good 
> polling, and in particular, only if one knows the number of viable 
> candidates.  As I've seen here in SF, that changes over the course of 
> the election season.
> 
>     They want A elected and for that to happen, A has to get into the
>     Automatic Runoff.  If A gets into the Runoff, then A is already
>     ranked higher than anyone else and their vote in the final runoff is
>     for A.  But if A can't get into the final runoff, then it's a choice
>     between B and C and then scoring B just 1 level above C is
>     sufficient for their entire vote to be for B.  Scoring B any higher
>     only makes it more difficult to get their favorite, A, into the
>     final runoff, so *ostensibly* there is no reason to score B higher
>     than 1.
> 
> 
> ...except if there's a candidate "D", and the voter has opinions on all 
> four candidates.  Your model doesn't scale above 3 candidates.  Almost 
> all alternative voting systems assume there will not only be viable 
> third parties, but viable fourth parties, and viable fifth parties.
> 
>     But I also shown that with that scoring regime, STAR would make the
>     same mistake that Hare IRV makes regarding the Center Squeeze...
> 
> 
> STAR doesn't make "the same mistake".  One has to have a comically 
> simplified voter model in order to contrive scenarios where STAR 
> uniformly performs identically to IRV.  Condorcet methods and the STAR 
> method are good at selecting compromise candidates because both make it 
> exceptionally likely that the compromise candidate makes it to the 
> "final round", where a critical pairwise comparison is performed.  Even 
> BTR-IRV needs to pick a "final round" by staging a series of 
> eliminations (like a lineal championship in boxing) where there's a 
> chain of matches until a championship round (the "final round") occurs.  
> We've spoken for years about making Condorcet more understandable by 
> figuring out how to seed tournament brackets, which lead to a "final 
> round".  STAR makes it reasonably easy to understand HOW the candidates 
> were selected for the final round, and I'm pretty sure it's effective.  
> STAR's not my favorite if I could wave a wand, but maybe I'd just make 
> myself dictator for life if I had access to a magic wand.  :-)
> 
> One thing I think we agree on: Hare IRV doesn't eliminate the spoiler 
> effect; it multiplies it.  Every round of an IRV election has the 
> spoiler effect.  One huge problem with IRV is how the second/third/ 
> fourth/etc preferences are ignored until the first preference is 
> eliminated.  Those unlucky enough to rank a spoiler victim too high will 
> have their subsequent rankings ignored.  STAR may not use a perfect 
> mechanism to select the final-round pair of candidates, but it's 
> defensible.  The Equal Vote folks were able to convince enough eager 
> volunteers that the mechanism was good enough to get it on the ballot, 
> and 30% of voters in Eugene thought it was good enough.  30% is not nothing.
> 
>     You would have to convince [voters] that they need to harm the
>     likelihood for A to get into the final runoff...
> 
> 
> Are you really trying to trot out "later-no-harm"?  How is Condorcet-TTR 
> or BTR-IRV better than STAR in passing later-no-harm?
> 
> I hope you're successful in palling up to Vermont legislators.  I truly 
> don't want to rain on your parade.  Just stop it with the bad-faith 
> raining on the STAR parade, okay?  You don't KNOW how voters will behave 
> if we start having high-stakes elections with the system.  Your 5-1-0 
> strategy assumes there will only be three viable candidates, and doesn't 
> take into account the incumbent vs non-incumbent dynamic (which causes 
> Greens and Republicans and Vermont Progressives and everyone else to 
> join forces and badmouth a Democratic-party incumbent).
> 
> Even the three-way race between Kiss, Wright, and Montroll in 2009 also 
> had Smith and Simpson in it.  I'm just going to assume that even Simpson 
> had folks knocking on doors, and I'll bet he was door-knocking himself.  
> All of Kiss's challengers were saying "the incumbent is awful", but I'm 
> guessing they all emphasized different reasons why he was awful.  When 
> one layers the "anti-incumbency" effect on the left-center-right 
> political model, we can see at least one more dimension to voters' 
> choices.  If one adds "quality of campaign", that turns it into a 3D 
> model.  If one adds "money" as a dimension...well, now we're in 4D 
> space, and that breaks my brain.  I'm guessing it breaks your brain as 
> well, Robert, even if you won't admit it.
> 
> Look, we should be allies, but it's kinda tough with your online 
> demeanor and clear overconfidence that STAR is unacceptable.  You 
> haven't proven to me that STAR is unacceptable, and I haven't found many 
> people who take your side who engage frequently online.  Please, correct 
> me if I'm wrong.
> 
> Rob
> p.s. I fear that a lot of people who THINK they understand RCV/IRV, but 
> when pressed to explain it, describe Borda.
> p.p.s. I'm annoyed that Marcus Ogren calls it "Ranked Robin" rather than 
> "Condorcet" in this Medium post, but it seems really interesting and 
> relevant to this conversation:
> https://voting-in-the-abstract.medium.com/voter-satisfaction-efficiency- 
> many-many-results-ad66ffa87c9e <https://voting-in-the- 
> abstract.medium.com/voter-satisfaction-efficiency-many-many-results- 
> ad66ffa87c9e>
> p.p.p.s. Some of you may have noticed I've turned on "Reply-all munging" 
> for the mailing list.  I'm pretty sure the increased use of DKIM is the 
> reason why I've missed out on many messages from this list.  Munging 
> helps with DKIM compliance and thus should make delivery more robust, 
> but it also means simple replies are more likely to go to the mailing 
> list rather than just the sender, even if you mean to just send a 
> private reply to the sender.  Be careful out there!
> 
> 
> ----
> Election-Methods mailing list - see https://electorama.com/em for list info