election-methods@mailman.electorama.com

Technical discussion of election methods

View all threads

Strategy in Condorcet methods

KM
Kristofer Munsterhjelm
Tue, May 9, 2017 11:05 AM

When I was exploring three-candidate linear Condorcet methods, I noticed
that the most strategy resistant methods did not pass both monotonicity
and reversal symmetry. The best result I got for methods that passed
both were susceptibility to strategy around the 70% range, where Minmax,
Ranked Pairs, Schulze etc is around 75%, and Smith,IRV is around 12%.

So perhaps there is an impossibility theorem hiding somewhere that says
that a Condorcet method can't pass monotonicity, reversal symmetry, and
some criterion that confers strong worst case strategy resistance. Of
course, it's hard to say anything at all without knowing what the latter
criterion is.

But this train of thought got me thinking about the Chicken Dilemma
criterion. Perhaps that is one such criterion. And if so, do any of you
know a method that passes monotonicity, reversal symmetry, and the CDC?
Or of a possible strategy towards an impossibility theorem showing that
one can't have all of them?

When I was exploring three-candidate linear Condorcet methods, I noticed that the most strategy resistant methods did not pass both monotonicity and reversal symmetry. The best result I got for methods that passed both were susceptibility to strategy around the 70% range, where Minmax, Ranked Pairs, Schulze etc is around 75%, and Smith,IRV is around 12%. So perhaps there is an impossibility theorem hiding somewhere that says that a Condorcet method can't pass monotonicity, reversal symmetry, and some criterion that confers strong worst case strategy resistance. Of course, it's hard to say anything at all without knowing what the latter criterion *is*. But this train of thought got me thinking about the Chicken Dilemma criterion. Perhaps that is one such criterion. And if so, do any of you know a method that passes monotonicity, reversal symmetry, and the CDC? Or of a possible strategy towards an impossibility theorem showing that one can't have all of them?