[EM] More about the simulation that I suggest

[Michael Ossipoff]

 I know that people have been doing simulations of ideal majoritarian
methods, and of course Eppley has done one that answers the question
of RP vs Beatpath.

But I don't suppose anyone has done simulations to compare the various
RP versions, as regards public preference between their winners.

Public preference between methods' winners seems the most important
question for a simulation to answer. Of course that won't always be
transistive.
It won't always point to a public-preferences CW, among the methods' winners.

When it doesn't, then of course the best thing is to compare the most
popular methods, the ones that are expected to be the best. Which of
_their_ winners is preferred to that/those of the other(s)?

Also, with those pairwise preferences between the various methods'
winners, one could do a count by an ideal majoritarian method, or by
various ideal majoritarian methods.

I suggest this simulation because I don't know if there's been a
simulation comparison of the various RP versions, in terms of public
preference betwen teir winners. HDRP/MMV wasn't even defined at the
time that previous simulations were done anyway.

I expect that the winner, the method that would beat the others, in
terms of public preference between winners, would be MAM-River. I
expect that MAM-River would beat MAM.

I exptect MAM to beat CIVS-RP and HDRP/MMV, because the later depart
more from what RP would do if there were no equal defeats. But I can't
say for sure,of course, and it's worth finding out via a simulation.

In general, I expect the River versions to do better than the ordinary
RP versions of MAM, CIVS-RP, and HDRP/MMV.

It would be nice to find that CIVS-RP does better than HDRP/MMV,
because of CIVS-RP's much, much briefer definition. That's another
thing that would make the simulation interesting.

The purpose wouldn't be to try to closely simulate actual U.S.
conditions. Of course in our actual elections there are very many
voters, and the issue-dimensions are strongly correlated--so much so
that it's nearly a 1-dimensional left-right political spectrum.

But, in order to make the methods often give different results, we
want lots of cycles, and lots of equal defeats.

We want lots of dimensions, no correlation, a fair number of
candidates, and very few voters.

I suggest that, if anyone heres interested in voting systems for ideal
majoritarian conditions (and that's the conditions that people here
are mostly interested in), then someone might want todo the simulation
that I suggest.

If no one else does it, then I might, if I want to take on such a
committment to a project. I probably won't. Even if I do, it will be
some time before I start. So, you-lot will have plenty of time to do
it first.

Michael Ossipoff