[EM]

[Blake Cretney]

On Wed, 14 Mar 2001 00:18:36 -0000
"MIKE OSSIPOFF" <nkklrp at hotmail.com> wrote:

> You probably know this already, but it's often said that, if
> the important output is an ordering of the candidates, then
Tideman's
> method is best. If considerations about the output ranking are
> more important than SSD's or BeatpathWinner's advantages.

I find it rather peculiar to imagine that the best ordering of the
candidates doesn't put the best candidate at the top.  That seems to
be what you're saying.

> A brief wording of Tideman:
> 
> Drop the strongest defeat that's the weakest defeat in a cycle.
> Repeat till there are no cycles. At that time, any candidate with
> no undropped defeats wins.

My preferred definition is as follows:

Ranked Pairs gives the ranking of the options that always reflects 
the majority preference between any two options, except in order to
reflect majority preferences with greater margins. 

I suspect Mike prefers his definition.  I suspect arguing the point
would be futile.

> [end of definition]
> 
> I've heard that Tideman can become a bit awkward when pairwise ties
> & equal defeats are likely, which is how it is when there aren't
> lots of voters.

I'd have to know exactly what is meant here in order to respond. 
Writing an implementation is made much more difficult by the
possibility of ties, but I've already written an implimentation.
 
> The preferred version of Tideman's method on this list is the
> version that measures the strength of a defeat by the support of
> that defeat--the number of people voting for that pairwise defeat.
> Defeat-support, as opposed to margins (defeat support minus defeat
> opposition). Tideman himself used margins. Defeat-support gives us
> compliance with important strategy criteria.

I prefer margins as well.  It's important to recognize that just
because Mike posts a lot, this doesn't make his views official.

I've created a web page that argues for margins:
http://www.fortunecity.com/meltingpot/harrow/124/path/inc.html

> I believe that, when considerations related to an output ordering
> aren't the most important, BeatpathWinner, and its equivalent,
> Cloneproof SSD, and SSD, are more preferred on this list. 

At least he's undecided on this point.  I'd actually be interested to
know how others feel.  My general impression is that most active
posters to the list aren't even interested in this debate, but I could
be wrong.

> Of course
> that's something that we could establish with a poll, if we
> choose voting systems for the demonstration poll, or for a later
> demonstration poll. I personally prefer SSD to Tideman. Cloneproof
> SSD when there aren't many voters and pairwise ties are possible.

Also, unless you're Mike Ossipoff, please call Tideman's method
"Ranked Pairs".  That's the name Tideman himself gave the method.

---
Blake Cretney