[EM] Realism of Tideman vs Schulze numerical testing

Craig Carey research at ijs.co.nz
Sat Nov 4 17:22:35 PST 2000

At 17:58 04.11.00 +0100 Saturday, Markus Schulze wrote:
 >Dear Craig,
 >there is a paper by Woodall in which he demonstrates that
 >no preferential single-winner election method can meet the
 >following four properties simultaneously [Douglas R. Woodall,
 >"An impossibility theorem for electoral systems," Discrete
 >Mathematics, vol. 66, p. 209-211, 1987]:
 >1) monotonicity;
 >2) later-no-help and later-no-harm;
 >3) if no second preferences are expressed and candidate A
 >   has a plurality of first preferences, then candidate A
 >   must be elected;
 >4) if more than half of the voters strictly prefer candidate
 >   A and candidate B to every other candidate, then either
 >   candidate A or candidate B must be elected.
 >Which of Woodall's properties would you be prepared to see
 >not satisfied by your preferred system?
 >Markus Schulze

Hi Markus.

This is a question that I can answer without pausing: property
4 is due to be rejected. In my opinion the odd people out are
those that uphold the Condorcet pairwise comparing idea, not those
that disagree with it. If the STV followers had clear ideas, they
would amongst those that reject pairwise comparing [despite the
bad [non-monotonic] start the Alternative Vote makes of agreeing
with Condorcet in the {(AB),(B),(C)} problem]

Property 3 ought be rejected too since it wrong for all elections
with 0 winners.

Suppose a method flops out in an ideal galactic election test.
For example, suppose the method is based on pairwise comparing
(premise "(4)" above) ?. That would lock in failure under large
scale testing. The question above seemed a bit pointed. The faith
fails because the whole religion is error. Markus: why does your
method fail to fail to properly respect and handle complex
vote trading [done by papers, not people] and the complex
coalitions that would tend to exist in extremely large elections?.

Dr Woodall just put a bad premise to the left of an implication
sign. To say that he liked the premises is going a bit far when the
data is just paper or a PDF file.

Is Mr Norman Petry's program coded correctly?. Have you checked it?.
If not, why not?.

Can you tell us Markus: which of these two is the better method, the
Schulze method?, or the Tideman method?. The criteria might as well
be some that can test all preferential voting methods.

You have the words "my preferred system", i.e. 'preferred method'.
It is possible to write on just rules and not state a preferred

This is like a nuclear bomb that eliminates pairwise comparing: it
will not cope with bribing and complex interacting coalitions in
extremely large elections. How can you seriously want to hint in
any way whatsoever that your Schulze method is worth even knowing?.

 From time to time, the Condorcet defender scrape off their shoes a
bit of a rebuttal saying something like: you need to have very small
expectations of our methods.  Why Markus Schulze?. Is you method
somehow hardened against attacks. It has a credibility that is
vapourised hasn't it?. Even the looked fight of Tideman vs Schulz
in the kidergarten sandbox doesn't look like it is about to start.
Lack of interest?.

Craig Carey

More information about the Election-Methods mailing list