[EM] "spoiler effect" argument

[James Green-Armytage]

A question:
	What is the best way to demonstrate to IRV fans that Condorcet has less
of a "spoiler effect" than IRV?

Here's one idea:
[beginning of argument]
	Define a spoiler as a non-winning candidate whose deletion from the
ballot would change the winner.
	Case 1: There is a Condorcet winner (with respect to the preferences as
expressed on the ballots). A Condorcet-efficient method will certainly not
have a spoiler. However, if IRV gives a different result, then it will
certainly have a spoiler.
	Case 2: There is a majority rule cycle (with respect to the preferences
as expressed on the ballots). Both Condorcet-efficient methods and IRV
will certainly have a spoiler.
	Hence, the frequency of spoilers is necessarily greater in IRV than
Condorcet.
[end of argument]

	I'm nearly sure that the conclusions in case 1 are correct, but I'm not
entirely sure about the conclusions in case 2. Can anyone prove or
disprove the case 2 conclusions?
	Are there ways to make this argument stronger?
...
	One interesting thing to do would be to run computer simulations to see
how frequently IRV produces a spoiler candidate given n candidates and k
voters, compared to some method like beatpath or ranked pairs. Does anyone
have the resources to do this? I think it would make the argument seem
more concrete if you could estimate e.g. 2% chance of spoiler in beatpath
vs. 7% chance in IRV, for some particular number of candidates.

my best, 
James