[EM] Further comments on Richard's criteria

[MIKE OSSIPOFF]

Let me add a few additional comments about Richard's message:

Richard said:

Any of the strategic criteria on the electionmethods.org site can be
defined without reference to "sincere preferences" or "favorites".

I reply:

In my message entitled "3 Ways of Writing Certain Criteria", I said that 
SFC, GSFC, WDSC, & SDSC can be written without reference to sincere 
preferences or sincere voting.

I told how that can be done, and I demonstrated it by posting a votes-only 
CC that's equivalent to my CC.


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.

I reply:

I've already pointed out that Plurality meets that criterion. But I'd like 
to add that every pairwise-count method, including Copleand, Dodgson, Black, 
etc. meets it too. Because in all those methods a candidate wins if s/he 
beats all the other candidates in pairwise comparisons.

Copeland, Dodgeson, Black, and Plurality don't meet SFC.

Richard continued:

GSFC: If X is in the Smith set and Y is not, and a majority of ballots
rank X over Y, then Y does not win.

I reply:

Not only Plurality, but every method that meets the Smith Criterion meets 
this criterion. For instance Copeland meets this one too. Richard's "SFC" is 
not equivalent to SFC. Richard's "GSFC" is not equivalent to GSFC.

Richard continued:

SDSC: For any combination of N ballots that all rank candidate X over
candidate Y, and any other combination of M ballots, if N is greater
than M, then Y does not win the election when the two groups of
ballots are combined.

I reply:

Not only does Plurality meet this criterion, but Ranked-Pairs fails it. 
Example:

40: ACB
25: BAC
35: CBA

60 voters rank B over A. 40 do not. 60 is greater than 40. But A wins in 
Ranked-Pairs.

Richard continued:

Note: The statement in the electionmethods.org commentary that
"Compliance with SDSC means that a majority never needs any more than
truncation strategy to defeat a particular candidate" is puzzling
(truncation strategy is "falsely voting two candidates equal", is it
not?).

I reply:

No, Richard is making a sloppy inference here. If you don't mention Joe, and 
you don't mention Moe, you aren't saying anything about them. For instance, 
you aren't saying that Joe & Moe are equal.   Not ranking them merely says 
that they aren't as good as the candidates whom you have ranked. It says 
nothing about the merit of Joe & Moe in relation to eachother.

By the way, as I pointed out, my definition of voted pairwise comparisons 
says that not voting X over Y or Y over X is one of the 3 possible voted 
pairwise comparisons with respect to X & Y.

But I didn't say that that means that voter is voting X & Y equal. I defined 
voted pairwise comparisons broadly & generally, because the fact that 
someone didn't vote between X & Y is worth recording too. But not voting a 
comparison between X & Y is very different from saying that X & Y are equal, 
or voting X & Y equal.

So the statement that not ranking X & Y doesn't vote them equal in no way 
contradicts my definition of voted pairwise preferences.

Richard continued:

Also, the statements in the commentary about "countering
offensive order reversal" are completely irrelevant, since the need
for defensive strategy is based on the opposing votes cast regardless
of whether those votes are the result of offensive order reversal.

I reply:

I've already answered that one thoroughly in my previous reply to this 
message.

Richard continued:

WDSC: For any combination of N ballots that all rank candidate X equal
to or higher than candidate Y, and any other combination of M ballots,
if N is greater than M, then Y does not win the election when the two
groups of ballots are combined.

I reply:

Not only does Plurality pass this criterion, but Ranked-Pairs fails it, in 
the same example in which RP failed SDSC.

In fact, Richard's WDSC & SDSC make a ridiculous requirement for rank 
methods: The require that if the candidates are in a cycle, no one must win.

Richard continued:

FBC: For some set S of ballots, if R is the set of results that can
occur when ballot B is added to S if candidate X is given the highest
ranking on B, and R' is the set of results that can occur if ballot B
is added to S when candidate X is not given the highest ranking on B,
then either X is a member of R, or R' is a subset of R.

I reply:

Based on Richard's accuracy-score with his other criteria claimed to be 
equivalent to SFC, GSFC, WDSC, & SDSC, there's no reason to believe that 
Richard's "FBC" is equivalent to FBC.

For that reason the time required to wade into his FBC definition above 
isn't justified.

In any case, even if Richard's "FBC" were equivalent to FBC, it wouldn't 
matter, because his criterion wording isn't usable. My wording of FBC 
directly and obviously speaks to a concern of voters and electoral 
reformers. A criterion that doesn't do that isn't useful. A rewording of FBC 
that doesn't do that isn't useful.

Mike Ossipoff

_________________________________________________________________
Keep up with high-tech trends here at "Hook'd on Technology." 
http://special.msn.com/msnbc/hookedontech.armx