Tideman Definition Different?

Mike Ositoff ntk at netcom.com
Mon Aug 3 13:36:41 PDT 1998



In the archives of this list, I ran across a letter which,
if I remember correctly, quoted an article written or co-written
by Tideman himself. The letter quoted a passage that defined
Tideman's method (probably not calling it that).

The point of the person who posted the quoted passage was that
the original definition didn't say to "skip" a defeat, and
leave it skipped and ignored for the rest of the count, which
is what the description in the _Journal of Economic Perspectives_
article (Winter '95) seems to suggest doing.

The crucial phrase in the quotation was "...while preserving
all pair orderings with greater majorities". As opposed to
all greater defeats that haven't been skipped. So previously-
skipped defeats _aren't_ subsequently ignored.

***

So then, defined that way, Tideman's method wouldn't have the
problem that I posted here a few days ago. Provided that
it uses votes-against, I don't know that it has any problem
that prevents it from being as good as EM's best. Admittedly,
I don't know what it's properties are, though it does seem
promising at first glance.

The person who posted the quoted passage also said that,
when defined that way, Tideman is equivalent to Schulze. 
Is that correct? I don't know which other method, if any,
it's equivalent to.

But, if it is the same as Schulze, that's good, because
it means that a method has been proposed in journal articles
which (if modified to use votes-against) is the same as the
best simple count rule we know of.

That's encouraging, just as it's encouraging that one of the
academic authors had a guest-editorial published in the Washington
Post, on June 21, '22, describing Plurality, IRO, Borda, &
Condorcet, and saying that the aurhor preferred Condorcet.
That's encouraging, because, with the use of votes-against,
Condorcet is one of the best methods--even if better ones have
now been found.

If it turns out that Tideman, by that better definition, 
_isn't_ equivalent to one of our known best methods, it
still appears promising. I don't want to appear to want to
bash new methods that might be better. I welcome better ones,
and if there's another one that's better than Smith//Condorcet(EM),
then good--the more excellent methods the better.

Mike Ossipoff


	



More information about the Election-Methods mailing list