[EM] Comparing CW with "write-in pairwise" CW

[rbj at audioimagination.com]

>>> ... Condorcet is often indecisive as well.

>>

>> how often, really, is Condorcet indecisive in a real election?

>>

>

> I don't know about a real election, but if you just randomly generate scenarios, Condorcet is frequently indecisive.
well, we *know* that cycles *can* occur.
so using a ranked ballot doesn't pass the "monkey test" if your expecting or requiring "decisive"
results.  but voters aren't monkeys (some might find that premise debatable) and there is usually some political alignment on the spectrum.  in 2000, voters for Ralph Nader would not as likely choose W as their second choice as they would Gore.  similarly voters for W would not likely
choose Nader as their second choice.
 
> That's what I meant.
>
well, we *know* that cycles *can* occur.
so any decent Condorcet-compliant method should have a procedure for dealing with cycles.  because i am not convinced that in political reality, cycles would be "frequent" and i'm even less convinced that, if a cycle *did*
occur, it would involve more than 3 candidates (no more often than once in a blue moon), then i still think a simple method for dealing with cycles (like Ranked Pairs or MinMax) is as good as a complicated method (sorry Marcus).  and with only 3 candidates in the Smith set, i think that Ranked
Pairs or MinMax (using margins) will come up with the same winner as Schulze.
 
but i really think that in real elections with real voters (not a simulation) that we'll find that Condorcet is not frequently indecisive.  but it *could* happen, which is why we have to plan for it
(and i am still unconvinced that anything more complicated than Ranked Pairs need be in that plan).
 
L8r,
 
r b-j

 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20131030/75c93b5f/attachment-0003.htm>