[EM] IRV vs Condorcet & Score?

[Dave Ketchum]

I wrote this to an audience considering IRV and talked of Condorcet  
for having almost the same ballot rules, but being better in other  
ways.  There are other methods also using ranked ballots, which I do  
not discuss here.

Range/score does ratings.  In its own way it lets voters vote for more  
than
one, and use ratings to easily indicate more and lesser preferences.

A quick comparison for voting for GOOD, SOSO, and BAD:  GOOD and BAD  
properly need high and low rank or rating.  SOSO requires different  
thinking:
      For Condorcet any rank above BAD and below GOOD will say all I  
can say.
      For Range/score if I rate it barely above BAD I risk it losing  
to BAD, depending on other voter ratings; if I rate it barely below  
GOOD I risk it beating GOOD, depending on other voter ratings.

Back to IRV:

I see IRV as an important topic, but have to argue against using it -
or for even promoting it and then suffering when those who learn that
it is less than perfect have nasty words for promoters.

What it offers voters is excellent - a voter can vote for more than
one, and can use ranking to easily indicate more and lesser
preferences.  These abilities are lacking in many election methods
such as Plurality, which is the method most of us have used most of
our lives.

Promoters can demonstrate the ease with which an election they choose
to demonstrate for IRV gets counted - discarding the candidates with
the least votes to correctly decide on the winner.

Those of us who have looked closer see what we call failures due to
IRV's way of making decisions with incomplete analysis, awarding wins
when complete analysis would declare a different winner - Burlington,
VT demonstrated this possibility recently in a real election.

I offer here an IRV demo - extreme, but showing what is too easy to
fail in a real election.  Here ALL like the Angel and Tom, Dick, and
Harry each have backers, so:
     34 T>A>D>H
     33 D>A>H>T
     32 H>A>T>D
     1   A

The 1 for A discards easily.  So do the 32 for H.  But this exposes 32
for A - which being low count gets discarded immediately.

So Tom wins with 66T>33D, though a closer look at the beginning would
have seen 68A>32H, 67A>33D, 66A>34T, 65H>34T, 66T>33D,  67D>32H.

I offer Condorcet, for which voters can vote exactly as for IRV but
have more complete analysis.  We can discuss more completely another
day.

Dave Ketchum