# [EM] Condorcet cyclic drop rule

MIKE OSSIPOFF nkklrp at hotmail.com
Mon Apr 2 18:24:28 PDT 2001

```>From what Blake said, maybe Condorcet meant margins, since margins
seem more what one would use for judging which propositions are more
likely to be objectively true, something that Condorcet expressed interest
in estimating.

I emphasize that I don't believe it's feasible to try to find out
those probabilities from the voting system. If we use a voting system
with avoidable strategy problems, then people won't vote sincerely,
and the margins mean _nothing_ about the probability that the propositions
are true. Voting is a strategy situation. We can and should
minimize strategy need as much as we can. Using a method with more
strategy need because it's defeat-measure estimates the probability
that the defeats are "true" pairwise comparisons when people vote
sincerely doesn't make sense, because why should they vote sincerely
with a method that has more strategy need?

In any case, I'm more interested in voters' rights than in trying to
judge the probability of propositions being right. Note that
doing Ranked Pairs based on the probability that propositions are right
isn't the same thing as electing the candidate most likely to be the best.

But if we're interested in estimating objective truth, why only
judge the pairwise propositions on the basis of their probability of
being true--why not just look for the candidate with the best probability of
being the objective best?

If people are going to vote sincerely, then use CR to estimate the
best candidate. If they're not going to vote sincerely, then
Ranked-Pairs(margins)'s probability estimates are going to be wrong,
based on false voting.

But say you want to use rankings, and estimate the best candidate:

I've suggested previously that it seems better to count the instances
that someone ranks X over someone else, to get an estimate of how good
X is, requiring complete rankings. That's a wording for Borda.

Elect the candidate who gets the most betterness votes.

The pairwise betterness propositions obviously aren't transitive, and
that seems to make their meaning questionable. You're saying:
"A is better than B, B is better than C, and C is better than A. Which
of those betternesses is weakest; we'll ignore it because it's less
likely to be true. The public said that C is better than A, but (since
it has least margin), you disregard that public vote result. Something
wrong there.  The fact is that the public has made 3 mutually incompatible
statements about betterness, and you want to fudge it
by judging the probabilities of propositions being right. Better to
just count how many betterness votes each candidate has--Borda.

I don't advocate Borda, and I don't think the objective betterness
or bestness estimates are the way to go, since elections are subjecdt
to strategy.

Mike Ossipoff
]
_________________________________________________________________