On Feb 20, 2007, at 15:39 , Michael Ossipoff wrote: > Juho wrote: > > My
sympathies towards minmax(margins) come primarily from the way > it handles
sincere votes. > > I reply: > > But there wont be sincere votes for it to
handle, to the extent > that it doesnt allow sincere votes. Thats why the
defensive > strategy criteria, and the wv Condorcet methods were proposed.
Juho replies:
Do you mean that margins would be so "strategy inviting" that most voters
would turn to strategic voters (in practical real-life elections) if margins
are used?
I reply:
Yes, voters would be more likely to regret sincere ranking in margins than
in wv.
Juho continues:
And that WV would solve that?
I reply:
Yes. By meeting SFC. And by making it much easier to deter offensive
order-reversal, merely by not ranking the perpetrators candidate.
Juho continues:
(I'm under the impression that this kind of impacts are not very big and
that they may work in both directions.)
I reply:
WV is much more strategy-free. The difference is unidirectional.
Juho continues:
I think Dodgson counts the sum of defeats. I'm not talking about that, just
basic minmax(margins) (that actually implements "additional votes needed to
become the Condorcet winner").
I reply:
But Minmax only scores the candidate according to his worst defeat. That
doesnt tell what it would take to get rid of all of his defeats and make
him the CW. Dodgson does that. But Dodgson fails Clone Independence and all
of the defensive strategy criteria.
Minmax is used with more than one meaning, and so its best to call PC by
the name PC.
Id said:
> Minmax isnt a good method name, because its used with more than > one
meaning.
Juho replies:
Also with other meanings than minmax(margins), minmax(winning votes) etc?
I reply:
Yes, its often used to refer to Simpson-Kramer, a method defined in the
Winter 95 Journal of Economic Perspective in a way quite different from PC.
Juho continues:
The defensive strategy criteria and wv Condorcet were proposed for > a
reason. ... As I already mentioned I don't like counter strategies to much.
I reply:
Then youd like methods that meet SFC, because theyre much more free from
need for counterstrategies. The wv Condorcet methols meet SFC. Additionally,
SFC is met by MDDA. Also, MMPO meets SFC, but was shown to have some kind
of a problem that could cause serious embarrassment.
Ive proposed MDDB, which also meets SFC. MDDB replaces the approval count
with an SR count, for part two, to choose among the un-disqualified
candidates.
SR (Summed Ranks): Elect the candidate over whom (summed over all the
ballots) fewest candidates are ranked.
[end of SR definition]
MDDA & MDDB also meet FBC, as does SR. MDDA meets SDSC, but MDDB probably
fails that and WDSC. But MDDB could have a better chance of acceptance than
MDDA, if MDDA is rejected because of anti-Approval prejudice.
SR, meeting only FBC, is a substitute for Approval, if, as seems to be the
case, Approval cant get accepted by the public.
Id propose SR only if the better rank methods couldnt gain acceptance, and
if SRs simplicity made it the only accepted rank method. And, of course,
only if Approval cant get accepted, which, from my conversations with
members of the public, seems to be the case.
Anyway, a number of methods meet SFC: wv Condorcet, MDDA, MDDB, and MMPO.
Juho continues:
If real-life elections end up in media and parties proposing various counter
strategies to voters one day before the election (to strategies that some
groups are planning or might try) I'm sure that election method would
receive some criticism.
I reply:
Yes, but of course thats exactly what always happens with Plurality.
The best methods wont have that problem.
Juho continues:
I'm more interested in methods where strategic voting stays at levels where
no counter strategies need to be considered (and where strategies are not a
serious risk in the first place). > The wv Condorcet versions are much more
free of strategy-need, and > much more resistant to offensive strategies
(for instance, > offensive truncation isnt a problem in wv Condorcet).
There are also examples in the other direction. - Sincere votes: 49:A,
49:BC, 1:CB => B wins - Strategic votes: 49:A, 49:BC, 1:CA => C wins if
winning votes are used
I reply:
Its never been denied that offensive order-reversal can be a problem in
Condorcet. Its much more of a problem in margins than in wv. One of the
advantages of wv over margins is that, in wv, offensive order-reversal is
easily thwarted by simply not ranking the reversers candidate. In other
words, in wv, you can succeed with offensive order-reversal only if your
victims are trying to help you. Would that make you feel proud of yourself?
<smiley>
So, if you dont want a problem with offensive order-reversal, I recommend
wv to you.
Juho continues:
I however hope that we are discussing rather small differences in strategic
performance here
I reply:
No.
Juho continues:
(since my basic thinking is that Condorcet methods are at their best in
situations where strategic voting will not become a major issue due to the
natural strategy resistance of all Condorcet methods).
I reply:
You mean, due to the natural strategy resistance of wv Condorcet methods.
Juho continues:
Yes, good criteria are exact. But too often I see argumentation that refers
to scenarios that are possible in theory but that maybe never occur in
practice and/or whose impact is minor and/or requires lots coordination,
many strategic voters, has risk of failure etc
I reply:
Good point. Offensive order-reversal, the only thing that could cause a
strategy problem in wv (truncation causes a strategy problem in margins),
requires lots of co-ordination, many strategic voters and has great risk of
failure--especially in wv, where its so easily thwarted, merely by not
ranking the perpetrators candidate.
In margins, a CW could be defeated by truncation even if it is inadvertent,
lazy, hurried, or otherwise non-strategic. But of course the election could
be stolen from the CW by strategically-intended truncation too, in margins.
Mike Ossipoff
.
----
election-methods mailing list - see http://electorama.com/em for list info
