[EM] Tideman vs IBCM & DCD

MIKE OSSIPOFF nkklrp at hotmail.com
Sun Mar 19 20:07:00 PST 2000


EM list--

I'd like to make a few brief comments regarding the Tideman vs
Schulze discussion. First I'd like to discuss Tideman vs IBCM
& DCD, in terms of a standard suggested by Blake.

Blake suggested that if A beats B pairwise, then A should
be more likely to finish ahead of B in the output ranking than
vice-versa. (I'd just say that A should be more likely to win).

Another way of saying that is that if we're comparing 2 methods,
then a method is more appealing if its winner usually beats the
other method's winner.

The winner by the method known as IBCM or DCD usually beats
Tideman's winner when the 2 methods pick different winners.

DCD, like Tideman, is a Cycle Condorcet method, which (when
defeats are measured by how many support the defeat) meets BC
and all the defensive strategy criteria.

***

Blake & Markus were also discussing which method, Tideman or
Schulze does the best job of not rewarding truncation. I agree
with the person who said that it isn't desirable to punish
truncation, because truncation is a reasonable & natural way
to vote, and shouldn't be considered as being done as offensive
strategy. But it shouldn't be rewarded.

I don't have an opinion on whether Tideman or Schulze does
better in that regard. Because they're both Cycle Condorcet
methods, they both meet GSFC, and both avoid letting
supporters of a nonmember of the sincere Smith set from stealing
the election for their candidate by truncation when a majority
rank a sincere Smith set member over their candidate.
(if we measure defeats by votes-against--I'll abbreviate that
"wv", for "winning votes", because that term is used here).

But if defeats are measured by margin-of-defeat, then I suggest
that it's a meaningless question whether Tideman or Schulze
does better at not rewarding truncation. That's because
Schulze(margins) & Tideman(margins) both reward truncation in
the biggest way, letting it steal the election from a CW who
is ranked over the truncators' candidate by a majority,
when no one votes unfelt preferences.

No Condorcet version will do that. (Recent discussion has
shown that it's reasonable to limit "Condorcet" to wv methods).
In other words, Schulze(wv), Tideman(wv), SSD(wv), SD(wv)
and DCD(wv) wouldn't ever reward truncation in the blatant
way that every Margins method will, including Schulze(margins)
& Tideman(margins).

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



More information about the Election-Methods mailing list