[EM] Truncation criticized by IRVists

[Blake Cretney]

Michael Rouse wrote:

 >Here is an idea to prevent strategic truncation and ties (while still
 >allowing people to truncate or tie if they truly don't know/don't care/are
 >too lazy to choose).
 >
--snip--

 >This can be done by looking at each pair of candidates. If more people
 >prefer Candidate A to Candidate B in a pairwise comparison, then A 
would be
 >placed ahead of B on the ballots of everyone who showed no preference.
 >Continue comparing pairs of candidates until the entire preference list is
 >complete for all voters, then use the complete list in whichever voting
 >method you prefer.
 >
Once you do this, each pairwise comparison will have full participation
(not really, but for tabulation).  If we know, for example, that 5 out
of 100 people voted X>Y, but more voted Y>X, then we score 100-5=95 for
Y>X.  So you see, once we know the votes on the losing side, we can
calculate the votes on the winning side.  So, if we want to compare two
victories, we only need to compare the votes on the losing side for
each, whichever is lower is considered stronger.

--snip--

 >One possible problem would be if a voter left three or more candidates
 >unranked and these candidates formed a circular tie with voters who had a
 >preference. In that case, you might run an initial pass and order all the
 >ballots possible, then use the new ballot rankings and see if it will
 >resolve circular ties. If it doesn't, use your favorite circular 
tie-breaker
 >and then finish the ballot rankings. Then you can use the full ballot
 >rankings and run it through your favorite election method.
 >
Above I assumed that the desired ranking was possible.  If it isn't, and
you proceed as you suggest, then things get more complicated.

---
Blake Cretney ( http://condorcet.org)