[EM] minor wording correction

[MIKE OSSIPOFF]

When I replied to David, I said that if he told me his utilities
for the candidates, I could calculate those 2 probabilities that
Markus mentioned. Of course really I wouldn't need any utilities
to calculate those probabilities. The probabilities are calculated
from the assumption that we have no information about some situations
being more likely than others.

Markus's formula uses utilities to determine which course of
action would be better. Of course if one were doing mathematical
strategy, one would determine the expectation-improving value of
each pair-ordering that you could vote, and, from that,
determine what ranking is the best. Of course it could be a
strict ranking, or a ranking with equally-ranked candidates. It
could be a complete ranking, or a truncated one. Of course I'm
not saying that some particular count rule wouldn't preclude one
or more of those possibilities.

So the result of the strategy calculation would be a ranking--
your best ranking for that election.

***

By the way, if Blake's suggestion of wv with automated strategy
assistance was intended as a way of saying that wv strategy is
more complicated than that of Margins--it isn't. Margins, like
Condorcet, would require checking out all of those situations I
described. The difference is that Margins would require more
computation, because your ordering of A & B can change the
the margin of the defeat between them regardless of the direction
of the defeat.

Any method has a mathematical strategy, and rank methods have
very complicated strategies. Someone might criticize some method
by saying that there could be strategic incentive to do this or
that. Sure, those strategies are so complicated, with so many
possible situations, that no doubt they could give you all sorts
of incentives, depending on your utilities. But I caution against
assuming that that's what we should judge the methods by. The
real problems are the _gross_ problems, where a majority are
strategically forced to do something that often will give away
the election, in order to make a lesser-evil keep a greater evil
from winning.

The fact that it's horrendously complicated to calculate
mathematical strategy with rank methods, without a computer
program for it, shouldn't be counted against those methods in
general. So what? If you don't know the exptectation-optimizing
strategy, then you can rank sincerely with Condorcet. Is there
some meaningful sense in which someone who calculates his
mathematical strategy in the same election will have an unfair
advantage over you? Condorcet's unique basic guarantees say
that he won't be able to steal the election by thwarting
majority wishes. Sure, that other guy is doing better to achieve
his maximum possible expectation than you are. But it isn't
something that he can use against you to cheat a majority out
of a rightful win.

I've never heard of anyone willing to spend the time, or use up
the paper, to write a strategy for IRV. IRV, very unlike Condorcet,
is a method where strategy is very much needed. Good luck figuring
out what your strategy should be in IRV. Sure, it isn't easy to
calculate it in Condorcet (or Margins), but in Condorcet you
don't really need it to protect you, though you might like it for
that extra tweak to increase your utility expectation. I'd
conjecture that with Condorcet, there's only a small improvement
in using mathematical strategy vs sincere ranking anyway.

Mike Ossipoff

______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com