[EM] didn't check pairvotes both directions

[MIKE OSSIPOFF]

EM list--

Sorry, I made another error. In some of the situations that
I checked, for 0-info pairwise strategy, I forgot to consider
pair-votes in both directions.

You must be tired of my errors, but I'd checked for small
omissions, and I didn't realize that I'd made a big one, and
I was fairly sure that I'd done it right. What I must do, when
I make a determination like that, especially something
controversial, is wait a few days, during which I check on it
several times, persistently looking for a way it could be wrong.

Anyway, when the Margins determination is done right, I suppose
the terms affected by truncation will cancel eachother, and,
in keeping with common sense, Margins won't have 0-info incentive
to do other than rank sincerely, as Blake said.
If I considered that important, I'd be a Margins fan myself.
I won't mention Margins' 0-info strategy again unless I'm _sure_
it has incentive for doing other than ranking sincerely.

Since I missed considering some of the pair-votes in both directions,
which matters in Condorcet in the consideration of near BeatsAll
winners, my figure for the Ub needed to make it desirable to
rank B equal to A is probably off too. Maybe it will even turn
out to be .5, as Blake suggested. When I find out, I'll briefly
tell what the figure is here, but only when I'm really sure.

I wanted to study Condorcet's 0-info strategy because others have
brought it up, and Blake was using it as a criticism of Condorcet.
Of course my checking of my results before posting them hasn't
been thorough enough, and I'm aware of that.

Mike Ossipoff


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