Polling Ourselves

@MichaelOssipoff, what is the "wv" variant of Ranked Pairs?

“wv” stands for “winning-votes”.

The strength of a defeat is measured by how many people rank the defeater over the defeated.

wv brings powerful burial-deterrent…both in the form of Minimal-Defense compliance, & also in probabilistic autodeterence of burial.

wv Condorcet is invulnerable to offensive truncation strategy by a faction.

No other single-winner method is as strategy-free as wv Condorcet, e.g. RP(wv).

RP(wv) accepts equal-ranking & short rankings.

RP & Schulze are widely recognized as the best methods based on the number of criteria met.

RP is a lot more simply defined than Schulze.

Is strict ranking required, or is equal-ranking permitted?

@jack-waugh

Certainly. This is just a poll on single-winner methods.

A PR poll would be of interest, especially since I haven’t heard of one before. …so it’s due.

…for later, when there isn’t already a different poll in progress.

Jack Waugh

@michaelossipoff When I look up "Ranked-Pairs 'Winning Votes'", I'm seeing at least one assertion that there is an error in Wikipedia. Would you identify the exact definition of Ranked-pairs/Winning-votes that you want to nominate?

Do you know JavaScript? If not, what programming languages do you know?

Is Ranked-pairs/Winning-votes Frohnmayer-balanced?

Can it be tallied from the pair matrices alone from the precincts?

Do you know of a variant based on Score-type ballots (assigning a scalar to each voter-candidate pair), that would take the scalars into account under some circumstances? Would it be possible to tally such a variant based on totals from the precincts rather than gathering the ballots at a central location?

Here's an illustration for why scalar evaluation can signify more than mere ranking (even permitting equal ranking). Consider these two Score votes: Nader 1, Bush .99, Gore 0; vs. Nader 1, Bush .01, Gore 0. These seem quite different in their rating of Bush relative to the other candidates, but would reduce to the same ranking.

By the way, what voting systems are we going to be using when we vote on voting systems?

@jack-waugh said in Polling Ourselves:

@michaelossipoff When I look up "Ranked-
Pairs 'Winning Votes'", I'm seeing at least one assertion that there is an error in Wikipedia. Would you identify the exact definition of Ranked-pairs/Winning-votes that you want to nominate?

I’ve heard of various different tiebreakers for use in small committee votes, where there could be pairwise ties that could cause a tied result.

For public elections, of course pairwise ties would be vanishingly rare even in municipal elections, & nonexistent in state & national elections.

My tiebreaker rule for public elections is to elect the tie-member who tops the most rankings.

For this vote, I’d add that if two tie-members have the same number of Nth-place rankings, then compare their (N+1)th-place rankings.

But for our poll, I’d also report the tie. …in addition to applying the abovedefined tiebreaker.

Tiebreakers are the only RP(wv) different versions I’ve heard of.

But let me tell the count-rule:

One at a time, list the pairwise-defeats, strongest ones first.

But, if the next strongest one cycles with already-listed ones (makes a cycle with them, thereby contradicting them), then skip that one.

When all defeats have been either listed or skipped, elect the candidate who isn’t defeated by any listed defeat.

Do you know JavaScript? If not, what programming languages do you know?

My computer has two programming languages user-available. Either could do the exhaustive pairwise-count with lots of voters & candidates.

But if we get so many voters that it’s necessary to automate the Condorcet count, then I’d call that a very successful poll. Let’s hope we get that many participants !!!

Is Ranked-pairs/Winning-votes Frohnmayer-balanced?

I haven’t heard of that criterion.

Can it be tallied from the pair matrices alone from the precincts?

Yes.

Do you know of a variant based on Score-type ballots (assigning a scalar to each voter-candidate pair), that would take the scalars into account under some circumstances?

No. I don’t know of a reason to do so. I chose RP(wv) for its strategic properties, for which it excels. Pairwise count ordinal methods have the best strategy properties.

Would it be possible to tally such a variant based on totals from the precincts rather than gathering the ballots at a central location?

I don’t know I’ve never considered a cardinal hybrid, or heard of a reason for one.

By the way, what voting systems are we going >to be using when we vote on voting systems?

When someone proposes a voting system to use, that person is suggesting that people post
a vote by that method’s balloting…& is offering to count such ballots.

e.g. I invite people to vote an Approval set of poll-alternatives, & a ranking of them too.

(If they’d rank differently in RP(wv) & Hare, then write rankings for both. Speaking for myself, only the presence of completely unacceptable candidates in an election would give me reason to strategically vote a different ranking for Hare.)

Toby Pereira

I would nominate score and approval, but it seems they've already been nominated. Ranked Pairs with winning votes has already been nominated too, but I'd also want to nominate a Condorcet method that uses a cardinal ballot (which could still just be ranked pairs with winning votes).

I think cardinal ballots work better to prevent burial. If there are two main candidates, A and B and also a lesser known candidate C, people are likely to vote A>C>B or B>C>A. And this is not just because they are tactically burying. If there are polarising candidates, this is likely to be an honest vote. This is why burial-resistant Condorcet methods probably don't always work how people might hope. But with cardinal ballots, people are likely to just vote Max, 0, 0, so the opposition candidate is not buried.

For cardinal Condorcet, I'd want at least a 0-9 0r 0-10 scale. 5 would not be enough to make the distinctions people might want to make.

@jack-waugh

I like that better. Those Score-versions are simpler, briefer, & less removed from Approval.

I withdraw my nomination of Hare. We all oppose the “RCV” proposals. “RCV” proposals are disqualified by FairVote’s promotion-fraud. Not wanting “RCV” is something that we all agree on.

Jack Waugh

but I'd also want to nominate a Condorcet method that uses a cardinal ballot

So, are you going to? Time is running out.

Jack Waugh

@michaelossipoff said in Polling Ourselves:

I withdraw my nomination of Hare.

In that case, you might as well withdraw your nomination of Choose-one Plurality, if you nominated it as well, for just the same reasons.

Jack Waugh

I think Approval may well work so well as anything else in large public elections, but is too coarse for an election with fewer than 100 voters.

@jack-waugh said in Polling Ourselves:

@michaelossipoff said in Polling Ourselves:

I withdraw my nomination of Hare.

In that case, you might as well withdraw your nomination of Choose-one Plurality, if you nominated it as well, for just the same reasons.

I didn’t nominate that one. But, if I had, I’d withdraw it for sure.

Toby Pereira

@jack-waugh said in Polling Ourselves:

but I'd also want to nominate a Condorcet method that uses a cardinal ballot

So, are you going to? Time is running out.

OK, I do nominate the above. 0 to 9 scale, ranked pairs, winning votes.

Jack Waugh

@toby-pereira I am curious to know at what points in the tally the scores might figure in (other than via the rankings derived from them).

@jack-waugh said in Polling Ourselves:

@toby-pereira I am curious to know at what points in the tally the scores might figure in (other than via the rankings derived from them).

The results of the balloting s will be counted & reported at the end of the voting period.

About 2 days ago I suggested a 48-hour nomination-period. That 48 hours has elapsed, & so the nomination-period has concluded, & the voting-period has begun.

I don’t know how long a voting-period I suggested 2 days ago, but let’s say a 2-week voting-period. That’s plenty of time for anyone who intends to, to get around to it.

So shall we say that the voting-period ends at the midnight at the end of March 23rd…whenever that moment occurs in your time-zone? (By Standard-Time or Daylight-Savings-Time (Summer-Time), whichever your region is using).

That seems easier than a uniform time in GMT.

Everyone, at that time, should count the ballotings for & by all of the methods that he has proposed for the poll, & report the winners by each of those methods.

In the voting, I suggest that participants vote on each issue, posting a ballot for each issue by each proposed method.

.. or as many of those as desired. Maybe just an approval-set, a ranking, & a ballot by one’s preferred score-version (or all of the proposed ones if desired).

(A voting-system’s merit in use, because it includes easy administration & easy security-auditing, varies so closely with enactability & implementation, that there’s really not much reason to vote them separately, unless you feel that a more complicated method’s strategy-freeness for voters outweighs its less-easy administration & security-auditing.)

Of course tallying could begin as soon as any votes come in. Only the completion of the count & the reporting of the completed count need wait till the end of the voting-period.

@jack-waugh said in Polling Ourselves:

@toby-pereira I am curious to know at what points in the tally the scores might figure in (other than via the rankings derived from them).

@jack-waugh said in Polling Ourselves:

@toby-pereira I am curious to know at what points in the tally the scores might figure in (other than via the rankings derived from them).

Oh I see what you mean. I was wondering that too. …though it only really matters to the method’s proponent, who will count it at the conclusion of the voting-period.

@jack-waugh said in Polling Ourselves:

I looked at the RP Wikipedia article, & I see what you mean.

The article defines RP as using margins & not wv.

I feel that “RP” is a count rule that can use either wv or margins. (I like Eppley’s MAM, & it’s the count that I always select at CIVS, but I disagree with him on terminology-point.

I propose RP only with wv. Schulze uses wv in his method, & Eppley uses it in MAM (Maximize Affirmatived Majorities, his wv RP, specifying a particular tiebreaker for small committees).

As mentioned at the talk-page, wv is better for protecting against strategy.

Check out the talk-page—It mentions yours truly, in connection with criteria.

River is different from RP. River also skips a defeat if its defeated candidate is defeated by an already listed (kept) defeat.

I emphasize that RP is the rank method that I propose. Someone there said that River is better at burial-deterrence. I disagree. It seems to me that River undermines autodeterence.

Anyway, RP is much, much better-known, older, & more prestigious.

I’ve specified the simple tiebreaker that I propose for public RP…& the slight elaboration of it that I propose for this poll.

BTW, my tiebreaker proposal for Approval ( which could easily have a tie in a small poll) is:

If two candidates have the same number of approvals, then, for each of those two, sum the overall number of candidates that they’re approved with. Elect the one overall approved with the fewest.

@michaelossipoff When I look up "Ranked-Pairs 'Winning Votes'", I'm seeing at least one assertion that there is an error in Wikipedia. Would you identify the exact definition of Ranked-pairs/Winning-votes that you want to nominate?

Do you know JavaScript? If not, what programming languages do you know?

Is Ranked-pairs/Winning-votes Frohnmayer-balanced?

Can it be tallied from the pair matrices alone from the precincts?

Do you know of a variant based on Score-type ballots (assigning a scalar to each voter-candidate pair), that would take the scalars into account under some circumstances? Would it be possible to tally such a variant based on totals from the precincts rather than gathering the ballots at a central location?

Here's an illustration for why scalar evaluation can signify more than mere ranking (even permitting equal ranking). Consider these two Score votes: Nader 1, Bush .99, Gore 0; vs. Nader 1, Bush .01, Gore 0. These seem quite different in their rating of Bush relative to the other candidates, but would reduce to the same ranking.

By the way, what voting systems are we going to be using when we vote on voting systems?

@jack-waugh
@Toby-Pereira

My ballots:

(First a brief note. When a poll is discussed, & people in addition to me are interested in participating, then I proceed with that poll, regardless of how few people have indicated an intention to participate.

I proceed with the poll, however few participants it might have. A poll with more than one person interested is a poll of interest.)

Voting Systems:

Though I’m an Approvalist, & Approval is my single-winner proposal, I nonetheless approve
RP(wv), & rank & rate it above everything but Approval. That’s because RP(wv) is what I propose to people who insist on rankings.

…as many progressives do, because they’ve only heard about “RCV”.

So then I’d offer them RP(wv), as a better replacement for “RCV”.

In that sense, then, RP(wv) is one of my 2 main proposals. But I’d offer Approval 1st, & offer RP(wv) only if they insist on rankings.

So, voting-system ballots:

I combine both categories in one balloting, because I feel that, because of administration & security-auditing, the methods compare the same in both categories.

Approval set:

{Approval, RP(wv)}

Ranking:

Approval
RP(wv)
Score (100, 99, 50, 1, 0)
Score (5, 4, 3, 2, 1, 0)
Score (2, 1, 0)
6.STAR
RP/Score hybrid (because of complexity & newness, & because I don’t know its merit)

Score ballots:

(100, 99, 50, 1, 0):

Approval: 100
RP(wv): 99
STAR, & RP/Score hybrid: 50

In Cardinal methods we vote by our inclination. In Score, my inclination is to vote sincerely (estimated-merit-proportional), not strategically.

Not knowing RP/Score hybrid’ s merit, I rate it middle with STAR, because I rate its likely merit closer to 50 than to 1.

I rate & rank all Score below RP(wv) & Approval, because they aren’t as good as Approval in east brief definition,; easy explanation, proposing, implementation (Approval ca have literally zero implementation-cost), administration, & security auditing.

Additionally, Score complexly loses Approval’s absolute minimalness, & therefore also its unique unarbitraryness.

Among Score methods, I like the expressiveness of (100, 99, 50, 1, 0), though it didn’t give me the flexibility to distinguish RP/Score hybrid from STAR.

…with (5, 4, 3, 2, 1, 0) next in that regard.

There are advantages & disadvantages to STAR’s runoff. Most important to me are the resulting FBC-failure, & the strategy-complication resulting from the runoff.

Additionally, the runoff makes STAR depart from Approvalist by two removes, instead of Score’s one.

That doesn’t double the arbitrariness—It squares it.

(5, 4, 3, 2, 1, 0):

Approval: 5
RP(wv): 4
Score(100, 99, 50, 1, 0): 3
STAR, Score (5, 4, 3, 2, 1, 0), Score (2, 1, 0), &
RP/Score hybrid: 2

(…to rate three near middle)

RP/Score hybrid:

Approval: 10
RP(wv): 9
Score (100, 99, 50, 1, 0): 5
Other Score methods & the hybrid: 4

Presidential candidates:

Ranking:

Stein

Approval set:

{Stein}

All the other methods:

Stein max score.

Let’s say that presidential candidates can still be nominated. If anyone else gets nomination & vote, then my approval-set would become:

{Stein, Williamson}

…& my ranking would become:

Stein
Williamson

Toby Pereira

@jack-waugh said in Polling Ourselves:

@toby-pereira I am curious to know at what points in the tally the scores might figure in (other than via the rankings derived from them).

They wouldn't in that method. The scores would be used purely to determine the ranks. But I think scores are slightly easier for a large number of candidates, and I think scores act in a small way to prevent burial of candidates (as said above).

That said, I would also nominate Smith//Score now, but it appears that it might be too late! (0 to 9 ballot)

Jack Waugh

This post is deleted!

Jack Waugh

This is to gather in one place the list of nominees.

Approval (Ossipoff, Pereira, Waugh)
Ranked-Pairs(winning-votes) equal-ranking allowed (Ossipoff)
STAR (Ossipoff, Waugh)
Score{2, 1, 0} (Waugh)
Score{100, 99, 90, 50, 10, 1, 0} (Waugh)
Score{5, 4, 3, 2, 1, 0} (Waugh)
0 to 9 scale (only used for ranking), ranked pairs, winning votes (Pereira)
Smith//Score (0 to 9 ballot) (Pereira)
quantile-normalized score, with integer scores from 0 to 100 (Frankston)
0-9 score (Pereira)

Jack Waugh

For what to propose to the public:

  In Approval:

    1 Approval
    1 STAR
    1 Score{2, 1, 0}
    1 Score{100, 99, 90, 50, 10, 1, 0}
    1 Score{5, 4, 3, 2, 1, 0}
    1 Smith//Score (0 to 9 ballot)
    1 0-9 score
    0 quantile-normalized score, with integer
      scores from 0 to 100
    0 0 to 9 scale (only used for ranking), ranked
      pairs, winning votes.
    0 Ranked-Pairs(winning-votes) equal-ranking
      allowed

  In STAR:

    5 Approval
    5 Score{2, 1, 0}
    5 Score{100, 99, 90, 50, 10, 1, 0}
    5 Score{5, 4, 3, 2, 1, 0}
    5 0-9 score
    4 STAR
    4 Smith//Score (0 to 9 ballot)
    1 0 to 9 scale (only used for ranking), ranked
      pairs, winning votes.
    0 Ranked-Pairs(winning-votes) equal-ranking
      allowed
    0 quantile-normalized score, with integer
      scores from 0 to 100

  In Score{5, 4, 3, 2, 1, 0}:

    5 Approval
    5 Score{2, 1, 0}
    5 Score{100, 99, 90, 50, 10, 1, 0}
    5 Score{5, 4, 3, 2, 1, 0}
    5 0-9 score
    4 STAR
    4 Smith//Score (0 to 9 ballot)
    1 0 to 9 scale (only used for ranking), ranked
      pairs, winning votes.
    0 Ranked-Pairs(winning-votes) equal-ranking
      allowed
    0 quantile-normalized score, with integer
      scores from 0 to 100

  In Score{2, 1, 0}:

    2 Approval
    2 Score{2, 1, 0}
    2 Score{100, 99, 90, 50, 10, 1, 0}
    2 Score{5, 4, 3, 2, 1, 0}
    2 0-9 score
    1 STAR
    1 Smith//Score (0 to 9 ballot)
    0 0 to 9 scale (only used for ranking), ranked
      pairs, winning votes.
    0 quantile-normalized score, with integer
      scores from 0 to 100
    0 Ranked-Pairs(winning-votes) equal-ranking
      allowed

  In Score{100, 99, 90, 50, 10, 1, 0}:

    100 Approval
    100 Score{2, 1, 0}
    100 Score{5, 4, 3, 2, 1, 0}
    100 0-9 score
    099 STAR
    099 Score{100, 99, 90, 50, 10, 1, 0}
    090 Smith//Score (0 to 9 ballot)
    001 0 to 9 scale (only used for ranking),
        ranked pairs, winning votes.
    000 quantile-normalized score, with integer
        scores from 0 to 100
    000 Ranked-Pairs(winning-votes) equal-ranking
        allowed

  In 0 to 9 scale (only used for ranking), ranked
    pairs, winning votes:

    9 Approval
    9 Score{2, 1, 0}
    9 Score{100, 99, 90, 50, 10, 1, 0}
    9 Score{5, 4, 3, 2, 1, 0}
    9 0-9 score
    8 STAR
    8 Smith//Score (0 to 9 ballot)
    1 0 to 9 scale (only used for ranking),
      ranked pairs, winning votes.
    0 quantile-normalized score, with integer
      scores from 0 to 100
    0 Ranked-Pairs(winning-votes) equal-ranking
      allowed.

  In quantile-normalized score, with integer
    scores from 0 to 100:

    100 Approval
    100 Score{2, 1, 0}
    100 Score{5, 4, 3, 2, 1, 0}
    100 0-9 score
    099 STAR
    099 Score{100, 99, 90, 50, 10, 1, 0}
    090 Smith//Score (0 to 9 ballot)
    001 0 to 9 scale (only used for ranking),
        ranked pairs, winning votes.
    000 quantile-normalized score, with integer
        scores from 0 to 100
    000 Ranked-Pairs(winning-votes) equal-ranking
        allowed

  In Smith//Score (0 to 9 ballot):

    9 Approval
    9 Score{2, 1, 0}
    9 Score{100, 99, 90, 50, 10, 1, 0}
    9 Score{5, 4, 3, 2, 1, 0}
    9 0-9 score
    8 STAR
    8 Smith//Score (0 to 9 ballot)
    1 0 to 9 scale (only used for ranking),
      ranked pairs, winning votes.
    0 quantile-normalized score, with integer
      scores from 0 to 100
    0 Ranked-Pairs(winning-votes) equal-ranking
      allowed.

  In 0-9 score:

    9 Approval
    9 Score{2, 1, 0}
    9 Score{100, 99, 90, 50, 10, 1, 0}
    9 Score{5, 4, 3, 2, 1, 0}
    9 0-9 score
    8 STAR
    8 Smith//Score (0 to 9 ballot)
    1 0 to 9 scale (only used for ranking),
      ranked pairs, winning votes.
    0 quantile-normalized score, with integer
      scores from 0 to 100
    0 Ranked-Pairs(winning-votes) equal-ranking
      allowed.

  In Ranked-Pairs(winning-votes) equal-ranking
    allowed:

    Approval
  = Score (all ranges)
  > STAR
  > Smith//Score (0 to 9 ballot)
  > 0 to 9 scale (only used for ranking), ranked
    pairs, winning votes.