[EM] Revisiting Copeland
robla at robla.net
Thu Sep 8 14:41:15 PDT 2005
In thinking about proposals for public elections, it occurs to me that
we may have dismissed Copeland's method too soon.
A refresher on Copeland: Count up the pairwise victories. The one with
the best Win-Loss-Tie record is the winner (where a win is 1 point, a
tie is 1/2 point, and a loss is 0 points). This is incredibly compelling
and intuitive to sports fans, who are used to comparing teams by
Interestingly, the Copeland winner(s) is/are provably in the Smith Set,
and is very often equivalent to Smith set. You have to have two or more
cycles for the Copeland set to be different from the Smith set.
So, why not Copeland, with a simple tiebreaker (e.g. plurality, or at
most Minmax(wv))? That'd be very similar to how the NFL (U.S. football)
picks the divisional champions, which is first by win-loss-tie, then by
intradivisional win-loss-tie, then by total points scored.
( fyi, this is a continuation of a conversation that began here:
More information about the Election-Methods