Goldfish (single-winner method)

[Blake Cretney]

I just thought of a way to make my Goldfish method get closer to GMC.
The Goldfish algorithm I originally gave I now call Goldfish 0.1.
Here is Goldfish 0.2.

1.  Make a victory matrix.  This is a matrix where every cell contains
the votes for that row, against that column if the row beats the column.
Otherwise zero.

2.  Successively find the largest value in the matrix.  The row is the
pair-wise winner, the column is the pair-wise loser.  Copy all the 
values from the loser's row to the winner's row, when they are
greater.  Eliminate the loser's row and column.

3.  The last candidate left wins.

The change from Goldfish 0.1 to Goldfish 0.2 is the removal of
the column copy rule.  So if B is eliminated in favor of A, and
C beats A, B beats C, the B beats C score and the C beats A score
will coexist in the matrix until the higher one is picked and wins
out.

This brings Goldfish closer to GMC and Tideman.

I believe it satisfies the criterion that
If Y has a majority against X equal to q, X will not win unless there
are C1, C2, ... Cn such that
X >> C1 >> C2 ... >> Cn >> Y
where >> stands for has a majority greater than or equal to q.
Once again, my motivation is to find an algorithm that
can be easily implemented.  An easy implementation of
Sequential dropping could make this all moot.


-----== Sent via Deja News, The Discussion Network ==-----
http://www.dejanews.com/  Easy access to 50,000+ discussion forums