[EM] Markus: You still haven't defined "prefer". So your criteria remain undefined.

Craig Carey research at ijs.co.nz
Thu Feb 12 21:58:03 PST 2004


I rejoin the list to criticise Mr Schulze. Mr Ossipoff is
shooting up his theory but it looks like years may be needed.
Anyway, when Mr Schulze makes a mistake, the mistake is not
identified.

Mr Shulze just posted up rules that actually need "voters".

So wee know that there are n "kinds of papers", and maybe m "voters".
Mr Schulze is unable to do much with voters while he lacks the
m x n matrix of real numbers. No theorist designing fair
preferential voting methods is irrational enough to copy Mr Schulze
and imagine that a missing m x n matrix is worth having. Actually
explaining it will be a problem. However Mr Schulze has so far
not admitted to its existence. Perhaps he wants to say that
the 2 candidate First Past the Post method can not receive a
Real number that is not a Rational number (a root of a cubic
polynomial). Such special cases have no importance in the
consideration of the expert since using Real numbers.

Clearly Mr Schulze does not say that there is a 1 to 1 correspondence
between papers and voters. Certainly there tends to be just that
in public government elections. But governments would be more
interested in fair methods. Each time Mr Schulze writes of
pairwise comparing methods like the Schulze method (and it actually
might NOT Be a Pairwise Comparing method since Mr Schulze withheld
the definition of "Strictly Prefer"), he is lowering the chance of
a 1 to 1 correspondence between voters and papers.

Plain English that lacks secret meanings would be an improvement.


At 2004-02-12 14:16 +0100 Thursday, Markus Schulze wrote:
>Dear Mike,
>
>an election method is a mapping from a given input to a
>given output. Where this input comes from or what this
>input actually represents is of no concern as long
>as it has the properties required by this election
>method.
...

Do have a look at your rule and correct the wording.

You worded your rule in a way that makes it even less able
to do something if voters lack an existence.

A better wording than "given input and ..." isn't really going
to have a lot of success against the deviousness appearing
here.

A better idea is that a rule can ported into REDLOG, and proven
to be useful there, and then converted back into English, and
found to have exactly the same wording. I.e, the rule a single
meaning.


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

A worsening problem at this mailing list is this:

Mr Schulze is quoting text demonstrating that he is wrong, and then
writing as if he believes he is in the right. Unless Mr Schulze
can develop new improved ways of advancing his credibility, it
is not going to be the case that people believe what he writes.

Previously Mr Schulze used to say he 'did not understand'.
I suppose it was disbelieved that such defences were a way to
seem to be winner.

=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=
[1]:

As readers could perhaps recall from December, Mr Schulze withheld the Schulze
Condorcet variant algorithm from me. 

Later in January 2004 he privately promised to answer questions from me that
asked about his algorithm which had  appeared in Voting Matters edition 17.

(At the moment I still do not have the definition of the algorithm).

So I asked about his algorithm. In this next quoted public mailing list
message he seemed to welch on the commitments (pull out of gambling
promises). Since a public reply to a privately sent request, I sent another
request. But he reverted back to not replying.

So I now ask Mr Schulze to make available ALL of the Shulze algorithm,
and especially the part is described (with unacceptable inadequacy) with
the meaningless words "[voters] strictly prefer".

The article violated the Schulze rule on avoiding the word "voters".

In so far as I can spare a few words to comment on the article, it seemed
that it even lied in its title since there it said that the algorithm is
new and monotonic.

ONLY the creator of the algorithm prevents it from be attacked and shot
apart with Monte Carlo testing. Mr Schulze seemed to want to say that it
has got to be good because otherwise the procedure of copying (and
patching in) the Floyd algorithm would have been bungled. Neither numbers
(standing in for points and parties) nor symbols (central to fair methods
and fair rules) seem to get used.

http://lists.electorama.com/pipermail/election-methods-electorama.com/2004-February/011984.html

>[EM] Ossipoff's lost 8 proofs now supplied by Adam Tarr Re: [EM] Clarifying the definitions 
>Markus Schulze markus.schulze at alumni.tu-berlin.de 
>Sun Feb 1 03:32:03 2004 
>
>Dear Craig,
>
>you wrote (1 Feb 2004):
>> About 15 hours ago I asked Mr Schulze for a definition of his
>> Schulze method (published in October 2003) and even now I do
>> not have a statement of what the Floyd algorithm is actually
>> inserted into.
>
>The Schulze method is defined in Section 2 of my paper.
>Well-definedness is proven in Section 3 of my paper:
>http://groups.yahoo.com/group/election-methods-list/files/nmciswem.pdf
>
>The Floyd algorithm is neither a part of the definition
>of the Schulze method nor of the proof that this method
>is well defined.
>
>******
>
>You wrote (1 Feb 2004):
>> Code filling the "d" matrix is perfectly missing from the
>> Schulze submission to Voting Matters.
>
>In Section 1 of my paper, I write: "It is presumed that each
>voter casts at least a partial ranking of all candidates."
>In Section 2, I write: "Suppose that d[A,B] is the number of
>voters who strictly prefer candidate A to candidate B."
>
>Markus Schulze
---------------------------------------------


No variant can be trusted if part of it has to be alwyas withheld many
many times.

That pretends to inform on what the algorithm is. He refers me and
perhaps the readers to the PDF. There is no complete-enough definition
inside of the PDF.

That has been very obvious for a long time now.

There is no information on how to get from ballot paper counts over to
"d" matrix elements.

Note that Mr Schulze's "well definedness" proof makes no attempt to
find out if a method is well defined or not.

I still hope for the Schulze algorithm. Obviously it is one of the
world's most secret methods; a method claimed to be monotonic yet that
almost certainly is NOT monotonic. Unfortunately Mr Shulze believes it
is his duty to keep the algorithm secret. He never gave any reasoning
for that hard to miss agenda.

It could be to bathe in praise from the Election Methods List members
at having created a new Floyd algorithm. A thing so bad that crucial
parts of the algorithm pre-processing data before it reaches the Floyd
algorithm, are permanently secret with the reasoning also never making
its way out to this mailing list or to me privately.

Schulze recently said here that some components of the "d" matrix are
zeros. Is Schulze Condorcet variant "the "X over Y" number not actually
the negative of the "Y over X" number ?.

A completely suspect claim of monotonicity in a GPL-ed PDF file of an
algorithm that is not available. Mr Schulze got on with the editor and
both do not like having their claims checked using computer programs
running in a computer.

Maybe Mr Schulze could try a lot harder to get out, the reasoning for
withholding the full algorithm.

=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=
[2-a]:

http://lists.electorama.com/pipermail/election-methods-electorama.com/2004-February/012076.html

>[EM] 3 ways of writing certain criteria 
>Markus Schulze markus.schulze at alumni.tu-berlin.de 
>Tue Feb 10 01:34:03 2004 
>
>Hallo,
>
>I propose the following criterion as an alternative to FBC:
>
>   Each voter must be allowed to vote as many alternatives 
>   tied for top as he wishes. Suppose Q1 is the number of
>   voters who strictly prefer at least one candidate to
>   candidate A. Suppose Q2 is the number of voters who
>   strictly prefer candidate A to candidate B. Suppose
>   Q1 < Q2. Then candidate B must be elected with zero
>   probability.
>
>Markus Schulze

=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=
[2-b]:

http://lists.electorama.com/pipermail/election-methods-electorama.com/2004-February/012077.html

>[EM] 3 ways of writing certain criteria 
>Markus Schulze markus.schulze at alumni.tu-berlin.de 
>Tue Feb 10 06:54:02 2004 
>
>Hallo,
>
>I propose the following defensive strategy criteria.
>
>Criterion 1:
>
>   Suppose Q1 is the number of voters who strictly
>   prefer at least one candidate to candidate A.
>
>   Suppose Q2 is the number of voters who strictly
>   prefer candidate A to candidate B.
>
>   Suppose Q1 < Q2.
>
>   Then candidate B must be elected with zero probability.
>
>Criterion 2:
>
>   Suppose Q1 is the number of voters who strictly
>   prefer candidate B to at least one candidate.
>
>   Suppose Q2 is the number of voters who strictly
>   prefer candidate A to candidate B.
>
>   Suppose Q1 < Q2.
>
>   Then candidate B must be elected with zero probability.
>
>Markus Schulze


Shulze has a very secret undisclosed summing formula:

   What else could the words "the number of voters who
   strictly prefer ..." mean ?.

It seems to be exactly the same secret summing formula that
I was asking for: the one he said was inside a PDF when in
fact it was not. (I quoted the text of Mr Schulze above).

=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=
[3]:

I much agree with Mr Ossipoff in this way: Mr Schulze was, in
the 3 e-mails quoted above, using the word "voter[s]" in such a
way that the definitions said nothing if voters ceased to exist.

Voters are naturally irrelevant to rules. Societies are not
irrelevant to decisions on which rules to use.

We all know what "voters" means. It tends to exclude computer
subroutines. Particularly here the elections need not always
be public elections. They could elections in meeting halls.
Persons are obviously initially free to cast multiple votes
and fractional votes. Simply a mere use of the word "voters"
by Markus Schulze will fully fail to rule out the possibility
of one ballot paper being associated with a non-integral
number of voters.

The mere fact that Mr Schulze uses the word voter suggests to
me that ever since 1998 (about when the list started)
Mr Schulze has not taken an interest in the theory of
good preferential voting methods. That would involve a use of
algebra. To add the idea of voters seems to be doing something
like this: adding a matrix that associates these two:

 * each voter (discret I suppose) (a vector with m components)
 * the Real-valued vector with n components, and n is typically
   not equal to m, unless yet more text is added.

None of that extra text was actually added by Mr Schulze.

The competent expert never bothers saying what the relationship
between papers and voters are. 

---

http://lists.electorama.com/pipermail/election-methods-electorama.com/2004-February/012088.html

>[EM] Markus, those aren't the kind of criteria that you said you like. 
>MIKE OSSIPOFF nkklrp at hotmail.com 
>Tue Feb 10 22:15:01 2004 
>
>Markus,
>
>You said:
>
>I propose the following criterion as an alternative to FBC:
>
>   Each voter must be allowed to vote as many alternatives
>   tied for top as he wishes. Suppose Q1 is the number of
>   voters who strictly prefer at least one candidate to
>   candidate A. Suppose Q2 is the number of voters who
>   strictly prefer candidate A to candidate B. Suppose
>   Q1 < Q2. Then candidate B must be elected with zero
>   probability.
>
>I reply:
>
>Markus, referring to the above, and also to your subsequently posted 
>criteria: You said that you don't like criteria that mention voters' 
>preferences, and that you only like criteria that only mention voted 
>ballots.
>
>Mike Ossipoff


He did say that and now he tries to introduce some irrelevancies to 
confuse readers.

Shulze blundered in 2000AD and implied he understood one of
Mr Ossipoff's definitions. I commented on it and he actually
figured out that to make minor criticisms of Ossipoff's definitions
created amuch bigger criticism against the critic for purporting
to understand a definition that held no meaning.

Here, three to four years, later, he simply creates definitions
that are just as bad. He puts the words "strictly prefer" into
them, so if after enough private demands for the meaning, he
presumably would betray a private promise to disclose the
equation of the idea using a public message to EML where he
seeks to get people to understand his latest thinking.

=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=
[4]:

http://lists.electorama.com/pipermail/election-methods-electorama.com/2004-February/012086.html

>[EM] Markus, those aren't the kind of criteria that you said you like. 
>Markus Schulze markus.schulze at alumni.tu-berlin.de 
>Wed Feb 11 02:55:01 2004 
>
>Dear Mike,
>
>I wrote (10 Feb 2004):
>
...
>You wrote (11 Feb 2004):
>
>> Markus, referring to the above, and also to your subsequently
>> posted criteria: You said that you don't like criteria that
>> mention voters' preferences, and that you only like criteria
>> that only mention voted ballots.
>
>My criteria refer to a set of preferences without asking where
>this set of preferences comes from. When I give you a set of
-----

Oh, so he does not "ask" if "voters" are the source of the
preferences.

If voters do not exist then the rules do not exist too.

Ossipoff's objection said that Mr Schulze does not refer to
voters. He violated his principle by referring to voters.

By saying he does not additionally line them up against a
wall, would similarly be no defence.


-----
>preferences and I tell you who the winner is, then you can check
>whether in this example these criteria have been violated without
>having to know where this set of preferences comes from. This is
>exactly the kind of criteria that I said I like.
-----

The original wording in the definition must be CHANGED or else
the problem is not removed. Mr Schulze believes he wins but
clearly he lost.

The defence was to construct a whole lot of observers
("then you can check").

Mr Schulze says that the origin of the input "is of no concern
for the analysis of [the] election method".

If a method is being certified then the data might come from
voters, etc., and if the method is being tested, then the
rule knows the method. As far as I know, the method has to
be known algebraically to the rule or else there is a problem
with internal ties (unless using epsilon delta infinitesimal
probing etc. etc., none of which Markus Schulze has got around
to explaining clearly. Why is that Markus?).


--

>
>I wrote (31 Jan 2004): "Election methods are usually defined as
>a function from a given input (e.g. a set of partial rankings of
>the candidates) to a given output (e.g. a probability distribution
>on the set of candidates). Where this input comes from is of no
>concern for the analysis of this election method."


I thought that their polytope expressions were publicly
accessible. The polytope definition of the so called Schulze
method has not been available to everyone who asked for it in
the last decade. Isn't that right Mr Schulze.

>
>I wrote (2 Feb 2004): "In so far as election methods are defined
>only on the cast preferences and not on the sincere preferences,
>whether a given election method satisfies a given criterion must
>be reflected in the way this method uses the cast preferences.
>Therefore, there is no need to include the sincere preferences
>in the definition of a criterion."
>

What is the difference between an Ossipoff "sincere preference"
and a Schulze "sincere preference" in a zero winner election.
That is the "roll call" election over STV-style papers where
it does not matter if the Ossipoff "favorite" is the 1st
preference in the paper's list, or the last.

>   Suppose Q1 is the number of voters who strictly
>   prefer at least one candidate to candidate A.


---

Methods that are partly arbitrary could still get past real rules
(i.e. rules that are not certifying statements).

E.g. hair lengths could be allowed to have an infinitesimal effect on
a 2-candidate election's results.
The classic "anonymity" rule stops that sort of

But a method derived from axioms won't need an anonymity rule.


=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=--=


At 2004-02-12 11:57 +0000 Thursday, MIKE OSSIPOFF wrote:
...
>Richard continued:
>
>Here are my "votes-only" definitions:
>
>SFC: If X beats all other candidates in pairwise comparisons, and
>a majority of ballots rank X over Y, then Y does not win.
>


It won't be acceptable to say "pairwise comparisons" for that involves
real numbers that are not disclosed.

It seems to be the fashion over here: without more than 2 English
words, construct Eppley-Ossipoff-summation operators parameterized by
weights that are not actually mentioned. Then the method is really too
unfair to be used.




* end *

IFPP theory: http://groups.yahoo.com/group/politicians-and-polytopes



___________________________________________________________________________

        "As Una [Honeysuckle Creek, Australia] watched
Craig Carey
             walking on the surface of the Moon she spotted
       a Coke bottle that was kicked in the right hand side of
    the [TV] picture."
        http://www.ufos-aliens.co.uk/cosmicapollo.html
        http://www.aulis.com/




More information about the Election-Methods mailing list