[EM] Quinn, regarding Lomax reply

[MIKE OSSIPOFF]

Jameson:

You wrote:

Abd: You are right that C will probably win the chicken dilemma Mike 
stated, where C has 49%, under almost any system except SODA. 

[endquote]

No, A will win in Approval, if A all the A voters approve B, and all the B voters approve A.

Forest posted a solution for the C/D problem, for Approval and RV. I've discussed several other ways whereby the problem can
be dealt with in Approval. 

In AOC, MMT, or GMAT, A will win, because the B voters know that they'll get nothing from the A voters if they don't approve A, at least
conditionally. That's because they'll know that the A voters will probably only give to B a conditional approval.

MMPO2 and MDDTR automatically elect A.  People have said "What if the B voters are sincere about their A vs C indifference? 

Answer: Then A isn't the CW. But if the B voters prefer A to C, then A is the CW.

It's more important what happens when there's a CW, than when there isn't one. Therefore, it's best if A wins in that example.

MMPO2 and MDDTR are doing right when they elect A.

Their problem is their burial strategy. Because of that, I no longer like them as much as I did.

In ICT, A will win, for the same reason as A wins in AOC, MMT, and GMAT.

I've never claimed that Bucklin or ABucklin solves the C/D problem, the Approval bad-example problem. It doesn't.
But AOCBucklin, ACBucklin, GMATBucklin and MMTBucklin do solve it. Of course Approval with any or all of the
conditional options solves it.

So it's an exaggeration to say that only SODA solves it.





You continued:

In the end, I do not believe that AOC conditionality
 really fixes the problem any more than Bucklin does.

[endquote]

Bucklin doesn't solve it at all, nor ABucklin. AOCBucklin, etc. does.

AOC solves it for the reason described above.

You continued:

If C voters give 
conditional approvals to A and B, then A and B voters are again tempted 
to seek a leg up on the others by conditionally approving C, and again 
if both do so C wins. I doubt they'd be so shortsightedly partisan, 
though, just as I doubt it with Bucklin.

[endquote]

We've already discussed that. I acknowledged that possibility, and I said that I don't know of any FBC-complying method (at least one that
uses only ballots) that really completely avoids a co-operation/defection problem. With all the FBC methods I know of, a C/D problem
can come back, via burial strategy, as you describe above.

I've told why that isn't as bad as the C/D problem that AOC, ICT, etc., solve. The persistent C/D problem requires burial
strategy. It won't be as much of a problem.



You wrote:



There are only two ways I know of to truly fix this 
dilemma. One is as with IRV

[endquote]

IRV is a method that seems to completely avoid a C/D problem, and it fails FBC.  That's worse than having a completely unmitigated C/D problem,
and especially worse than the lesser C/D problem had by the defection-resistant methods such as ICT, AOC, etc.




You wrote:

Still,
 as I said above, half-solutions like AOC or Bucklin 

[endquote]
'
Bucklin isn't even a half solution, unless it's made conditional, like AOCBucklin.

AOC could be called a half-solution, as could ICT, MMPO and MDDTR, and AOCBucklin, GMATBucklin and MMTBucklin--because,
with those methods, the C/D problem can come back via burial strategy.

IRV is a complete solution to C/D, at the (unacceptable) cost of FBC failure.

Mike Ossipoff













may be enough if C 
has more like 40%; by merely making the slope down to a C win less 
slippery, such systems may avoid that outcome.


Still, since AOC is not really better than Bucklin 
here, while it is clearly more complex, I think there's no reason to 
waste our breath on it.
Jameson 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20120315/c7c1ad1f/attachment-0002.htm>