[EM] Challenging the "instant runoff" name (and "instant playoff voting")

raphfrk at netscape.net raphfrk at netscape.net
Tue Jun 20 06:11:31 PDT 2006

> From: Rob Lanphier <robla at robla.net>
> It occurs to me that a pretty cool voting system could be devised 
> this path (and probably has been -- please let me know who I'm ripping
> off here).

That's pretty cool.  It also uses less comparisons than
STV-BTR.  It requires log2(N) comparisons, while BTR
requires N-1.

> Seeding could be done by number of first place votes (i.e.
> the plurality voting ranking).  So, for example:
> Ballots (M=Memphis, N=Nashville, C=Chattanooga, K=Knoxville):
> 42: M>N>C>K
> 26: N>C>K>M
> 15: C>K>N>M
> 17: K>C>N>M
> Seeding:
> #1 Memphis (42 first place votes)
> #2 Nashville (26 first place votes)
> #3 Chattanooga (17 first place votes)
> #4 Knoxville (15 first place votes)

You could also use the seedings to give a by to some of
candidates.  This would allow a non power of 2 number
of candidates.  (in effect, adding "blank" candidates
who have the lowest seedings)

> This system has a very similar appeal to Copeland: it's Condorcet 
> compliant and it's easy to explain to sports fans.  It has the added
> advantage that ties are much harder than in Copeland, since not all
> pairwise comparisons are considered.

Ties are not possible at all unless there is a pairwise tie.  In
Copeland, ties occur unless there is a clear condorcet winner.

> I wouldn't be surprised if there are grave problems with this system 
> the event that there's no single Condorcet winner, but I haven't 
> out what those problems would be.

It would still have problems with tactical voting and condorcet ties.
For example, assuming 3 choices and true preference as:

2: A>B>C
1: B>A>C
1: B>C>A
3: C>B>A

This gives the following result:
A>B: 2-5 => B wins
A>C: 3-4 => C wins
B>C: 4-3 => B wins

B is the clear condorcet winner and will win under your system.

The bottom 3 voters can force a circular tie by changing their vote to:

3: C>A>B

A>B: 5-2 => A wins
A>C: 3-4 => C wins
B>C: 4-3 => B wins

Under your system:

First choices:

A: 2
B: 2
C: 3

C is first seed.

First round

A v B => A wins

C     => C progresses to 2nd round

Second round

A v C => C wins

However, in the 3 candidate case, your system collapses to
IRV-BTR anyway.  Maybe with more choices it will not be as
easy to manipulate.

More information about the Election-Methods mailing list