[EM] Condorcet question - why not bullet vote

[Kristofer Munsterhjelm]

robert bristow-johnson wrote:

> but if cycles are *not* involved (at all, going into a cycle from a 
> sincere CW or coming out of one), i would be interested in you showing 
> us how a strategy, such as Bullet Voting, can change the CW from someone 
> you didn't support to someone you do support.

To my knowledge, when there is a CW and your vote doesn't induce a
cycle, Condorcet methods pass LNHarm and LNHelp. The only way a
candidate can directly become a CW is if you rank him above the old CW
and vice versa.

Thus, truncation doesn't help your candidate. Either he wins by being
ranked above the old CW, in which case he would have won with a full
vote, or he doesn't, in which case he wouldn't. The only thing a full
ranking does is increase the possibility that your vote will pull the
winner away from the old CW towards someone you like better than him but
not as well as the favorite.

However, Condorcet methods pay for this criterion compliance within the
acyclical domain by introducing inevitable discontinuities between it
and the cyclical domain. The Condorcet criterion itself is incompatible
with the LNH* criteria: always electing the Condorcet winner means there
will always be transition cases (not the same for every Condorcet
method) where submitting a sincere ranking will pull the method outcome
across the divide and lead to an undesired result.