election-methods@mailman.electorama.com

Technical discussion of election methods

View all threads

Universal Domain Tie Breaking Agenda

FS
Forest Simmons
Wed, Aug 3, 2022 1:39 AM

Is there a really good, decisive, general purpose, tie breaking order that
depends on chance only is case of complete symmetry?

Here's one that's better than nothing:

Let Top( X) be the percent of ballots on which X is not outranked by a Y
candidate. Let Bot(X) be the percentage of ballots on which X outranks no
candidate.

The suggest tie breaking order is given by the formula ...

F(X)=(Top(X)-Bot(X))/(200%-Top(X)-Bot(X)),

primarily,

and secondarily by just the numerator of F(X).

The only case that this does not handle well is the case where F is zero
for the tied alternatives.

Is there a nice way to invoke the pairwise wins in that case?

If we had a good enough Universal Tie Resolution Order, we  could use it as
a standard, default agenda for all of our agenda based methods, including
Agenda Based Chain Climbing, which would open up the UD umbrella to Banks
efficient methods!

-Forest

Is there a really good, decisive, general purpose, tie breaking order that depends on chance only is case of complete symmetry? Here's one that's better than nothing: Let Top( X) be the percent of ballots on which X is not outranked by a Y candidate. Let Bot(X) be the percentage of ballots on which X outranks no candidate. The suggest tie breaking order is given by the formula ... F(X)=(Top(X)-Bot(X))/(200%-Top(X)-Bot(X)), primarily, and secondarily by just the numerator of F(X). The only case that this does not handle well is the case where F is zero for the tied alternatives. Is there a nice way to invoke the pairwise wins in that case? If we had a good enough Universal Tie Resolution Order, we could use it as a standard, default agenda for all of our agenda based methods, including Agenda Based Chain Climbing, which would open up the UD umbrella to Banks efficient methods! -Forest