Kevin,
That is interesting, but you should have specified ER-Bucklin(Whole).
ER-Bucklin(Fractional) works fine in your example.
You wrote:
Consider these ballots:
40 A>B>D
35 D>B
25 C>D
B wins, with 75 votes in the second round.
Now let's raise B on some ballots:
40 A=B>D
35 D>B
25 C>D
Now candidate D wins, with 100 votes in the second round.
The "fractional" version would give the same result as if half the
A=B>D voters had voted A>B>D and the other half B>A>D. In the first
round, the A=B>D ballots each give half
a vote to A and B and in the second round they each give another half a
vote to A and B (and nothing to D) so B wins as before with 75 votes in
the second round.
Chris Benham
----
Election-methods mailing list - see http://electorama.com/em for list info
