[EM] Range strategy pathological example (was Re: IRV vs Plurality)

[Abd ul-Rahman Lomax]

At 03:07 PM 1/26/2010, Jameson Quinn wrote:
>You can also rework the numbers so that a strategic centrist beats 
>an unstrategic majority-top candidate:
>20% voters: 10 Ader, 7 Bore, 0 Cush
>20% voters: 0 Ader, 10 Bore, 8 Cush
>60% voters: 0 Ader, 6 Bore, 10 Cush
>
>Now the Bore voters can win, if they use strategy and Cush doesn't. 
>But Cush has 60% first-choice support, and wins an honest Range vote!

To summarize this, there is no advice from anyone that voting the 
normalized utilities shown, without any regard for election 
probabilities, is a sensible way to vote Range. What Mr. Quinn has 
done is to suppose that one whole faction of voters votes sensibly 
(with regard for who the frontrunners are, basically) and another 
faction votes ignorantly, assigning most of their vote strength to an 
irrelevant race.

Is this a partisan election? Mr. Quinn hints that it is. Where are 
the approval cutoffs? A vote of 0, 10, 8 does not mean that the voter 
is approving Bore and Cush! It merely indicates dislike of Ader. 
Remove Ader from the race, and what are the "normalized utilities"?

And Ader isn't really in the real race!

In the version above, we somehow have the Cush voters also approving Bore.

Get this: in most elections, highly likely that most voters will only 
vote for one candidate, this is very common even with preferential 
voting systems, after they stabilize. Lewis Carroll knew this over a 
hundred years ago. Range *allows* voters to vote with more 
flexibility, but it does not require it, nor would I necessarily even 
encourage it for major-party voters.

(But you can, with Range, indicate support for a minor candidate, 
encouraging that candidate to run again, but be careful. Make sure 
that if the candidate wins, you wouldn't be upset!)