[EM] Beatpath versus Ranked Pairs

[Michael Ossipoff]

On Sun, Oct 27, 2013 at 3:38 PM, Chris Benham <cbenhamau at yahoo.com.au> wrote:


> I've abandoned the idea that it is possible to have a very good method that
> just weighs
> the pairwise defeats by some measure and juggles only that information to
> decide the
> winner.

It depends on the conditions under which the method is to be used. I
speak of three different kinds of conditions, and I'll just briefly
define them:

(For this purpose, we can disregard the fact that our vote-countings
are unverifiable, and that our elections are therfore illegitimate.)

1. Current conditions:

Disinformational media. Gullible electorate. The media say that there
are only two viable candidates, and have a blackout of candidates and
policies other than those of two media-permitted parties. The
electorate nearly all accept the premise that they only have "the two
choices", and vote accordingly.

Strategy situation:

Perceived need for favorite-burial. FBC is needed.

Suggested voting systems:

Approval. Maybe Score (Range, Cardinal Ratings). ICT. Symmetrical ICT.

If we, in this country, could get a better voting system for current
conditions, we'd need one that complies with FBC.  Approval or Score
would do. CD would be desirable too. To have that, along with FBC, ICT
or Symmetrical ICT would be good.

Though I've heard of the lesser-of-2-evils problem existing in other
countries too, I can't speak for other countries, and I speak of
"currernt conditions" only as a problem here in the U.S.



2. Green scenario:

Open, participatory, and agenda-free media. An electorate who, in any
case, have shown, by their election of a progressive government via
Plurality, that they'd no longer be deceived by disinformtional media
anyway.

Strategy situation:

Perceived favorite-burial need no longer a problem. FBC no longer
needed. Chicken dilemma, however, still can sometimes exist. When it
does, it would take away any benefit of compliance with the Mutual
Majority Criterion or Condorcet's Criterion.

Suggested methods:

Benham or Woodall would be best.

IRV would be good enough, because it meets the Mutual Majority
Criterion (MMC) and has no chicken dilemma. Benham or Woodall would be
better, due to compliance with Smith.

3. Ideal majoritarian:

Electorate no longer petty, jealous and rivalrous enough to have
chicken dilemma.

Strategy situation:

No need for FBC, no need for CD.

Suggested methods:

Ranked-Pairs (RP), River, and Beatpath (CSSD).

--------------------------------------------

I suggest that current conditions are irrelevant to the choice of
voting systems, because voting-system reform can't happen without,
first, the election of a better government.

Green scenario conditions have relevance, because, if voting system
reform can happen, it will be under those conditions.

A number of political parties offer IRV in their plaforms. No doubt,
soon after the election of a progressive party to office, there would
soon be a referendum or initiative in which the public could choose a
voting system. Therefore, discussion of methods for the Green scenario
has more relevance than discussion of methods for current conditions.

Ideal majoritatian conditions are farther off, but not too
preposterous to discuss. For example, if we were using the CD-failing
Beatpath, RP, or River, I'd initially rank all of the progressive
candidates. If other progressive voters did the same, then I wouldn't
start defection.

It isn't inconceivable that, someday, ideal majoritarian conditions
could exist, probably arising from Green scenario condions.

But methods for ideal majoritarian conditions have little practical
relevance right now, because, for this country, the Green scenaro is
the best that could happen.

Nontheless, the ideal ranked voting systems, for ideal majoritarian
conditions, are of aesthetic interest, and are interesting and
worthwhile as a matter of curiosity. It's worthwhile to ask what the
ideal rank methods would be. Aside from practicality, the question of
ideal majoritarian rank methods is of interest for its own sake, just
as a theoretical or hypothetical question.

I think they'd be RP, River and Beatpath.

(Maybe "Green scenario" is a little mis-named, because I'm not sure
that the Greens would really be the most winnable progressive party)


>
> If we want a Condorcet method the meets Mike's Chicken Dilemma criterion
> then I
> like the "Benham" method.

Agreed.

>
> If we want a Condorcet method that meets things like Minimal Defense and
> Mono-raise

But Beatpath, Ranked-Pairs(wv) and River meet those three criteria, in
addition to MMC and SFC.

> then I like a method of Forest's, the name of which I forget:
>
> We infer above-bottom ranking as Approval, equal ranking and truncation are
> allowed.
> We make a "chain", beginning with the most approved candidate A. If the
> second-most
> approved candidate B covers A, then we add B.  We continue considering each
> candidate in turn in order of approval (from higher to lower), adding a
> candidate to the chain only if it covers all the candidates already in it.
> The winner is the last candidate added to the chain.
>
> This method won't fail to elect a positionally dominant uncovered candidate,
> and doesn't
> have any random-fill strategy.

I realize that there's room for different suggestions about voting
systems for ideal conditions, as I've described them, or even for some
specified conditions that are even more ideal.

I personally feel that the best pairwise-only methods are best for
ideal majoritarian conditions.

> BTW, I recall forming the opinion that Beatpath (aka Shulze) is slightly
> better than Ranked
> Pairs (and probably also River).

I don't claim to really know their relative advantages and
disadvantages, but RP builds a complete structure of the strongest
mutually consistent defeats. Maybe that's why, in Steve's simulations,
the electorate preferred RP's winner to that of Beatpath a lot more
often than vice-versa.

As for River vs RP, both meet Condorcet, MMC, SDSC and SFC.

When River doesn't keep a defeat of an already-defeated candidate,the
interests of the people whose preference s  represented in that
not-kept defeat are still well-protected by that candidate's even
stronger kept defeat.

Looking for practical considerations for the value of a voter's right
to have his vote for X>Y counted, we could look at what that vote can
guarantee for hir, given a method's strategy-guarantees, due to its
critrion compliances. You vote X>Y because you don't want Y to win.
The voters of that stronger, kept defeat don't want that
already-defeated candidate to win, any more than do the voters that
candidate's unnecessary 2n defeat. And they, too, benefit from the
guarantees of CC, MMC, SDSC, and SFC. Therefore, they well-protect the
interests of the voters of the not-kept unnecessary 2nd defeat of that
already-defeated candidate.

...And, by not keeping that 2nd defeat, sometimes some other defeat,
some public pairwise decision can be honored. Surely we don't want to
avoidably fail to honor public pairwise decisions (pairwise defeats).

So it seems to me that River is better than RP, just on the basis of
what I've said here. But I don't claim to know much about those
methods' properties--except that it seems to me that CSSD, RP, and
River all meet CC, MMC, SDSC, and SFC.

Michael Ossipoff