<html>
  <head>
    <meta content="text/html; charset=ISO-8859-1"
      http-equiv="Content-Type">
  </head>
  <body bgcolor="#FFFFFF" text="#000000">
    A generalised STV  (which is what is my invention of Binomial STV
    [BTV]) does not have a residual FPTP in the last round.<br>
    By including a rational exclusion count, BTV  [Binomial Transferable
    Vote] avoids the problem of excluding candidates, which seems to be
    what much of the discussion on this forum is about.<br>
    from<br>
    Richard Lung.<br>
    <br>
    On 26/07/2017 16:12, Andy Jennings wrote:
    <blockquote
cite="mid:CA+Df2ty34PE3rw-HY-bQA2phrsCtqp3gbERu1-pnEq+ZN16ezQ@mail.gmail.com"
      type="cite">
      <div dir="ltr">
        <div>
          <div>
            <div>I think you're right that this matches BTV in the major
              details.  Is <a moz-do-not-send="true"
                href="http://rangevoting.org/BucklinTV.html">http://rangevoting.org/BucklinTV.html</a>
              the best reference for BTV?  It doesn't have a page on
              Electowiki, yet, right?  We should add one.<br>
            </div>
            <div><br>
            </div>
            <div>There aren't many good, simple, rated-ballot,
              multiwinner systems, so it deserves to get talked about
              more, no?<br>
            </div>
            <br>
          </div>
          <div>I wasn't suggesting to change the name, but if you are...<br>
          </div>
          <div><br>
            <br>
          </div>
          Regarding the quota:  I see what you're saying about the
          crumbs.  I just have reservations about a quota that's not
          even going to try to represent 1/(s+1) of the population.  I
          realize that STV does it and STV has a track record.  But STV
          does it because it might come down to a one-on-one at the end
          and you want to say that a majority in the final round is the
          same as the quota for all the other rounds.  A rated system
          doesn't have that restriction.  I kind of like approval voting
          in the last round, even if the winner only gets 30-40%.<br>
          <br>
        </div>
        <div>If we choose one quota for the default, I'd hope we could
          add a footnote that the other one was a possible alternative.<br>
        </div>
        <div><br>
          <br>
        </div>
        <div>Tiebreaker and deweighting:  I don't feel strongly about
          these, but it's good that we're considering different options,
          looking for simplicity but also looking for corners that cause
          adverse incentives.  I think it's better to recommend good
          defaults than just including a bunch of options.<br>
          <br>
        </div>
        <div>Tiebreaker: If we were to use Hare quota (and "approval
          voting" in the last round), then the critical grade for the
          last round is going to be 0(F), so a GMJ tiebreaker or "most
          votes at or above critical" are not going to help break ties. 
          We'd either have to have a different rule for the last round,
          or a second tiebreaker.  "Most votes strictly above critical"
          would do it.<br>
          <br>
        </div>
        <div>Deweighting: If we wanted to, we could "assign" voters to
          the representatives for whom they were deweighted.  In that
          case, it would be advantageous to deweight in chunks as large
          as possible.  So subtractive deweighting would be better than
          multiplicative and rules that "deweight completely" are good. 
          But multiplicative is probably simpler.<br>
          <br>
        </div>
        <div>~ Andy<br>
        </div>
      </div>
      <div class="gmail_extra"><br>
        <div class="gmail_quote">On Tue, Jul 25, 2017 at 8:58 AM,
          Jameson Quinn <span dir="ltr"><<a moz-do-not-send="true"
              href="mailto:jameson.quinn@gmail.com" target="_blank">jameson.quinn@gmail.com</a>></span>
          wrote:<br>
          <blockquote class="gmail_quote" style="margin:0 0 0
            .8ex;border-left:1px #ccc solid;padding-left:1ex">
            <div dir="ltr">
              <div class="gmail_extra">This is a good idea.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">But on thinking about it further,
                I'm not sure whether it's not the same as BTV.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">BTV, like Bucklin, works by
                gradually lowering a "pseudo-approval threshold", and
                electing and deweighting candidates as they reach a
                quota of "pseudo-approvals". Andy's proposal, like MJ,
                works by directly looking at the "quota-th" highest
                rating, and electing and deweighting the candidate who's
                highest by that measure.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">But of course, we know that,
                aside from tiebreakers, MJ and Bucklin are the same
                thing. So the more I think about it, the more I think
                that (aside from quota choice, tiebreaker, and
                deweighting scheme; none of which are really specified
                by the label "BTV") Andy's proposal and BTV are the same
                thing.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">I could be wrong about this...
                can anybody else check my logic here?</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">Still. Even if this is just a new
                name for BTV, it's a good excuse to discuss that system.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">We could talk about how good it
                is. Pretty excellent! I like that it avoids the horrible
                center-squeeze breakage of STV. Even though the problems
                with center squeeze are much less in a multiwinner
                setting than in IRV, it's still ugly.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">When designing <a
                  moz-do-not-send="true"
href="http://wiki.electorama.com/wiki/Geographic_Open_List/Delegated_%28GOLD%29_voting"
                  target="_blank">GOLD</a>, I chose STV rather than BTV
                as a substrate. That wasn't because I prefer STV
                theoretically; it's just because of its longer track
                record.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">Also, we could talk about the
                ancillary design decisions: quota choice, tiebreaker,
                and deweighting scheme.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">Quota choice: I tend to prefer
                Droop, or a compromise V/(S+.5), over Hare. Basically,
                when you're assigning the last seat, you're left with
                the voters who are most atypical; the "crumbs" of the
                party system. If you use a Hare quota, then at best
                you'll find a candidate with some appeal to a full
                quota; but realistically, you might just find the
                biggest of a group of crumbs, who could easily have
                support from just 35-40% of a quota (based on 1/e, my
                SWAG for this kind of situation).  If you go with a
                Droop quota, on the other hand, the entire pool is 2
                quotas; and 2/e is 70-80% of a quota, much closer to
                fair.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">Andy's suggested deweighting
                scheme might help encourage bigger crumbs, but I'm not
                sure about that.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">Tiebreaker: I don't have a lot to
                say about this. GMJ-style seems like a good choice.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">Deweighting: This is where things
                get interesting. You don't want to have too much of a
                free-riding incentive, but you do want to deweight the
                votes which are "more satisfied" with the winners and
                not-deweight those which are "less satisfied" with the
                future potential winners.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">I like Andy's concept of
                subtractive, rather than multiplicative, deweighting. It
                makes things a little bit harder to describe, but it
                does mean that somebody who is "halfway decisive" twice
                will be fully deweighted, rather than keeping 1/4 of
                their voting weight; that seems fair to me.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">I think that Andy's rejected idea
                of "for those who only gave the new winner the threshold
                rating, deweight them last" was doing it wrong, so I'm
                not surprised that he decided it led to too big of a
                free rider incentive. If you're doing a GMJ tiebreaker
                anyway, then from a BTV point of view, those voters are
                essentially giving a fraction of an approval to the new
                winner. I think that only that fraction of their ballot
                should be at risk for deweighting; so their subtractive
                deweighting should be the minimum of their GMJ fraction
                and the overall deweighting.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">The other way to do things is to
                try to avoid deweighting voters insofar as they still
                have useful opinions about the remaining candidates.
                That's what Andy's proposed "completely deweight those
                who rate all remaining candidates at 0" rule would do.
                But this could still leave a very "crumbly" remainder at
                the end; imagine if the 100 candidates for the last seat
                each had 1% of the remainder giving them a top-rating.</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">So I can imagine more complicated
                schemes to do this. For instance:</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">
                <ol>
                  <li>Find the R candidates with the highest quota-th
                    ratings, where R is the remaining number of seats.
                    In other words, the prospective winners if you
                    proceeded from here on without any deweighting.<br>
                  </li>
                  <li>Of the deweight-able votes (counting only the GMJ
                    subtractive portion ot threshold votes), find the Q
                    which have the lowest max rating for those R
                    candidates. Deweight these completely.</li>
                </ol>
                <div>Note that the incentive of the above is not so much
                  to downvote early winners, as with traditional free
                  riding (though of course that is still possible if you
                  downvote them below their winning threshold), but
                  rather to up-vote late winners. That creates a
                  couter-free-riding incentive; a possibility I'd never
                  considered before.</div>
                <div>....</div>
                <div><br>
                </div>
                <div>But all-in-all, I think that Andy's suggested
                  deweighting scheme is pretty good, and I'd rather go
                  for "simple" than "theoretically awesome" here.</div>
              </div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra">Jameson</div>
              <div class="gmail_extra"><br>
              </div>
              <div class="gmail_extra"><br>
                <div class="gmail_quote">
                  <div>
                    <div class="h5">2017-07-24 21:58 GMT-07:00 Andy
                      Jennings <span dir="ltr"><<a
                          moz-do-not-send="true"
                          href="mailto:elections@jenningsstory.com"
                          target="_blank">elections@jenningsstory.com</a>></span>:<br>
                    </div>
                  </div>
                  <blockquote class="gmail_quote" style="margin:0 0 0
                    .8ex;border-left:1px #ccc solid;padding-left:1ex">
                    <div>
                      <div class="h5">
                        <div dir="ltr">Here's a multiwinner system
                          that's so simple that it should have a name,
                          but I don't think it does.  Let me know if it
                          does.<br>
                          <br>
                          It uses rated ballots.  The goal is to
                          repeatedly find the candidate whose top
                          quota's-worth of grades are highest and elect
                          that candidate, then de-weight a quota's-worth
                          of voters.  Some names worth considering:<br>
                          <br>
                          Sequential Best Assignment<br>
                          Sequential Constituent Matching<br>
                          Sequential Quota Allocation<br>
                          <br>
                          The method:<br>
                          <br>
                          N = Number of voters<br>
                          S = Number of seats<br>
                          <br>
                          1. Every voter grades every candidate.  (I'd
                          say 4 or 6 grades.)<br>
                          <br>
                          2. Each voter starts with weight 1.<br>
                          <br>
                          3. Choose quota Q = N / S. (*)<br>
                          <br>
                          4. For each candidate, calculate the minimum
                          of their top Q grades.  Let G be the highest
                          minimum.  Elect the candidate with that
                          minimum.  (Break ties as in GMJ: calculate for
                          each candidate what fraction of their G grades
                          are in their top Q grades, and elect the
                          candidate with the smallest such fraction. 
                          Break further ties by choosing the candidate
                          with the least number of G grades in their top
                          Q grades.)<br>
                          <br>
                          5. Deweight some voters to decrease the total
                          voter weight by Q, in this manner:<br>
                            a) any voter who gave the minimum grade to
                          all remaining candidates is deweighted to 0.<br>
                            b) for the voters not deweighted in (a) who
                          gave this candidate a grade of G or above,
                          find the deweighting D such that when the
                          deweighting formula:<br>
                          <br>
                            W_new = max(W_old - D, 0)<br>
                          <br>
                          is applied, the total voter weight in this
                          round is decreased by Q. (**)<br>
                          <br>
                          6. Repeat steps 4 and 5, applying voter
                          weights when calculating the top Q grades,
                          until S seats are filled.<br>
                          <br>
                          <br>
                          (*) With this quota, when you are filling say,
                          4 seats, then 25% of the voting weight gets
                          used up with each seat filled.  25% of the
                          voting weight will remain when choosing the
                          last seat.  That last seat will be determined
                          by the tie-breaker rule, so it is essentially
                          equivalent to approval voting, with any
                          above-bottom grade counting as approval.<br>
                          <br>
                          The other common choice of quota, Q = N / (S +
                          1), could also be considered.  When filling 4
                          seats, then, 20% of the voting weight gets
                          used up with each seat filled.  40% of the
                          voting weight remains to choose the last seat,
                          so the last seat is essentially filled with a
                          median-based method (GMJ).  20% of the voters'
                          opinions are, by design, left without a
                          representative.<br>
                          <br>
                          (**) I thought about another step (a') where
                          anyone who gave a grade strictly above G was
                          deweighted completely, but I think it gives
                          the voters too much incentive to down-weight
                          candidates who they think can get elected
                          without their help.<br>
                          <br>
                          I also considered another step (a'') where
                          anyone who graded the chosen candidates
                          strictly above all other candidates was
                          deweighted completely, but I don't think
                          there's much benefit for the added complexity.<br>
                          <br>
                          <br>
                          Any thoughts on which quota is better or on
                          the right name?<span
                            class="m_-2399388108750202778HOEnZb"><font
                              color="#888888"><br>
                              <br>
                              ~ Andy Jennings<br>
                              <br>
                            </font></span></div>
                        <br>
                      </div>
                    </div>
                    ----<br>
                    Election-Methods mailing list - see <a
                      moz-do-not-send="true"
                      href="http://electorama.com/em" rel="noreferrer"
                      target="_blank">http://electorama.com/em</a> for
                    list info<br>
                    <br>
                  </blockquote>
                </div>
                <br>
              </div>
            </div>
          </blockquote>
        </div>
        <br>
      </div>
      <br>
      <fieldset class="mimeAttachmentHeader"></fieldset>
      <br>
      <pre wrap="">----
Election-Methods mailing list - see <a class="moz-txt-link-freetext" href="http://electorama.com/em">http://electorama.com/em</a> for list info
</pre>
    </blockquote>
    <br>
    <br>
    <pre class="moz-signature" cols="72">-- 
Richard Lung.
<a class="moz-txt-link-freetext" href="http://www.voting.ukscientists.com">http://www.voting.ukscientists.com</a>
Democracy Science series 3 free e-books in pdf:
<a class="moz-txt-link-freetext" href="https://plus.google.com/106191200795605365085">https://plus.google.com/106191200795605365085</a>
E-books in epub format:
<a class="moz-txt-link-freetext" href="https://www.smashwords.com/profile/view/democracyscience">https://www.smashwords.com/profile/view/democracyscience</a>

</pre>
  </body>
</html>