[EM] Replying to Richard Moore:

[Former IRV Supporter]

Richard Moore wrote:
> 
> Former IRV Supporter wrote:
> 
> > I suppose #1 and #5 are still technically frontrunners, but not what I
> > had pictured in my previous message -- probably because my first
> > reaction is still to think in terms of our current two-party system.
> 
> In fact it seems reasonable to suppose that under Approval, dominance
> by two front-runner parties would be weakened to the point where
> voters are more likely to simply vote their consciences, without
> the need for anything more complex than an above-the-mean utility
> strategy.

That may be.  Or more likely you could just vote above-the-mean
considering only the more viable candidates, and then add votes for any
obscure candidates you happen to like better than your more plausible
choices.


Some more thoughts regarding what is perceived as a front-runner:

[Richard's original example:]
> >   [...]   Suppose we have five
> > candidates, and my ratings for those candidates are
> > 10, 9, 8, 0, and 0. Let's say the first and last candidates
> > are the front runners. I adjust for this fact and find that
> > the strategic values for the candidates are 10, 5, 4, -9,
> > and -10.

To account for these strategic values for all *five* candidates, given
the individual's ratings, the odds of a tie for all five must have been
something like
 25.5% / 13.5% / 12.5% / 23% / 25.5%.  But in order for five candidates
to have these tie probabilities, the predicted vote counts must have
been fairly close -- say within one standard deviation of each another.

So if the probabilities were for a plurality election, the predicted
vote totals must have been something like
 20.8% / 19.0% / 18.8% / 20.6% / 20.8%

If the tie probabilities were for an approval election, the predicted
approval totals must have been something like
 50.0% / 47.2% / 46.8% / 49.5% / 50.0%

So I don't know whether this would have been characterized as a two
front-runner election, or a toss-up.