[EM] MCA, IRNR, IEVS, retraction

[Warren Smith]

>MCA is Forest's method, unless you've made some change to it...

--oh.  Sorry, I thought it was from Benham.  Anyhow, it is this:
canddt with most-2approvals wins if gets >50%, else regular approval-winner wins

where there are 3 kinds of votes:  disapprove, approve, and 2approve  (a higher
level of approval).

--Olson's IRNR system rather surprised me.  I was not expecting it to do well because
it fails just about every property-satisfaction-theorem in sight (as I proved in
some RV posts).   But it does very well in IEVS testing.  Presumably because the
property failures are rare and/or not very severe.
IRNR's success is a good example of the fact that properties are merely a mental crutch
for use in judging voting methods, a crutch adopted by those of us without the
compute power to actually *measure* how good methods are.

> Approval Voting using "zero info strategy" is found in net in current
> IEVS scenarios to have better SU than
> every honest-voter-Condorcet method tried (and IEVS has about 20 of
> them).
> Venzke: This claim seems new. Is it?

--Sorry, I retract this claim as premature and misleading.

IEVS has about 15000 probabilistic+utility+behavior models of voters.
It does lots of  elections with all systems in all models.
It then has "summarizers" which try to summarize which systems did the best
over all the models (summarizing a vast amount of data).
All 3 summarizers agree approval is better than every Condorcet method.
However, some of those scenarios involved strategic voters.

I need to make better summarizer-controls so you can ask it to summarize only
SUBSETS of the data (such as, honest voters only, large #s of candidates only, etc)...

But anyway, to return to the honest-Condorcet vs zero-info-strat-Approval
question, there is a summary of that based on my old sims from 1999-2000 here
http://rangevoting.org/RVstrat4.html
(large table at bottom)
and the revised claim (sorry!) is that zeroinfo-Approval beats honest-Condorcet
usually with 3 candidates
but usually Condorcet beats ziA with 5 candidates.

Warren D Smith
http://rangevoting.org