[Election-Methods] RE : Is "sincere" voting in Range suboptimal?

[Abd ul-Rahman Lomax]

At 02:15 PM 7/25/2007, Chris Benham wrote:

>How can it be best strategy to vote approval-style in Range 99, but
>with the voters using a more
>restricted ratings ballot some other way of voting becomes better?
2
Isn't that what has been claimed already by others? I.e., by 
restricting the ratings to 2 instead of 99, it is being claimed that 
there is an improvement.

What I've discovered, the significance of which Benham does not yet 
get -- and I'm going to assume that I haven't simply made some 
mistake, I make many -- is that Range 2 is better than Range 1. And as to this:

>BTW, I find it annoying and stupid that you (Abd) have (apparently 
>unilaterally)  changed the name of Approval
>from "Range2" to "Range1" and the name of what was previously called 
>"Range3" to "Range2" etc.

Get over it. I was not asked to vote when the Election Methods 
International Free Association voted on the question, nor did I have 
a proxy assigned enabled to vote for me, so I'm not bound by its decision.

Oh. There never was such a vote or consensus decision.

I, like many others, simply started out thinking in terms of the 
number of ratings. Cardinal Ratings 2 makes total sense, there are 2 
ratings. But Range ... and Range 2 .... It's an opportunity to 
introduce a somewhat clearer terminonlogy, and I apologize for 
disturbing the hide-bound oldtimers with it. The list of 
Range  methods, as I propose it, uses N to refer the number of 
preference steps, and the list begins with 1. Which means that the 
basic Range method is Range 1, not Range 2. Isn't it nice to have a 
list that starts with 1 rather than two?

The fact is that when I wrote, in the past, Range (number of ratings) 
I was always aware that there was a problem, and that my terminology 
was not based on some standard. So let's create a standard! Want to 
keep Range-N equivalent to CR-N, that's certainly a possibility. But 
I want to have a name that refers to the number of steps. It is also 
the name, in 0- Range, of the max vote. In Range 100, the max vote is 
100 if the min is zero. Likewise Range 99 has a max vote of 99. But 
0-99, with the old somewhat common terminology, was Range 100, and 
this did indeed cause some confusion. Vote percentages are wonky, for 
example... Want to have an exact midrange vote? 50% is possible if 
you choose an even number for the Range. Hence Range 2 has a midrange 
vote of 1.

Simple. Like it arcane and complicated? Follow Benham, he'll show you around.

>>Proportionally, offhand, it looks to me as if the voter's vote is 
>>not moot in the same percentage of cases for Range or Approval.
>>
>
>One of the approval strategies (mean-based thresholding) is best in 
>Range because it increases the chance that
>voters employing it will have some (positive from their perspective) 
>impact on the result.  Range ballots give voters
>a greater choice of  inferior strategies so I would expect that on 
>average in Range more votes are "moot".

That's an inference that is too complicated for me to address. 
Doesn't sound right to me, though.

The number of moot votes, as I'm defining it, does not depend at all 
on strategy. I'm considering the votes where the voter can possibly 
affect the outcome, and these are all equally likely. It's simple. 
Apparently too simple for Benham. There will always be inferior 
strategies. Voting against your favorite is an inferior strategy, isn't it?

I think Benham should look at the results before speculating about 
things that may depend on them. He's already noted that there are 
some surprises in there. Indeed there are.

I really wonder that nobody has bothered to look at election strategy 
in this way before. My guess is that someone has. But where? Is it 
published or available on the web?

>>My own political strategy is to promote Approval as being a great 
>>improvement at very small cost.


>Good luck with that. I agree that adding one more voting-slot is 
>attractive, just for candidates that the voter is
>unsure whether to approve or not. But doing so allows voters to 
>express a lot more pairwise preferences, and so
>raises questions about which is really the best algorithm. For 
>example, how do we justify not meeting the "3-slot Condorcet"
>criterion?

I don't give a fig about "Criteria," unless they measure something I 
truly want to maximize, such as *my* satisfaction, or "overall voter 
satisfaction," I want to do both. Given a forced choice, which I 
choose depends on the cost to me!

>Approval is my favourite way of counting 2-slot ballots, but  CR (or 
>Range) isn't my favourite way of counting 3-slot
>ballots.

Show that another way is better, my a measure that counts, I'll 
certainly look at it!