1 |
On Thu, 2006-08-10 at 20:24 +0000, Ferris McCormick wrote: |
2 |
> On Thu, 2006-08-10 at 21:11 +0100, Ciaran McCreesh wrote: |
3 |
> > On Thu, 10 Aug 2006 20:03:26 +0000 Ferris McCormick <fmccor@g.o> |
4 |
> > wrote: |
5 |
> > | So the "glue" is rather easy; problem is the specific balloting |
6 |
> > | method. STV supports several protocols for selecting some number of |
7 |
> > | winners from a list of candidates, but Condorcet is not among them, |
8 |
> > | because Condorcet is really a "pick single winner" method. |
9 |
> > |
10 |
> > All you need to do is delete the single winner from the election and |
11 |
> > repeat the process. |
12 |
> > |
13 |
> |
14 |
> True. I was hoping no one would notice, however, because that gets |
15 |
> tedious (although once you have the ballots, it can be automated to a |
16 |
> large extent). At some point, we should re-examine policy and run some |
17 |
> controls to see if a voting method more closely designed for what we are |
18 |
> actually doing might be more appropriate. |
19 |
> |
20 |
|
21 |
Thanks to a pointer from Jan Kundrát, I have been able to build a master |
22 |
STV ballot from last year's council ballots. I had previously failed to |
23 |
note the "Eliminate candidates..." option to STV. But having noticing |
24 |
it, I am able to follow Ciaran's suggestion and run repeated Condorcet |
25 |
cycles over last year's ballot until 7 members are selected. Running |
26 |
STV in this manner selects the same council as we did elect. |
27 |
|
28 |
Total effort --- perhaps 15 minutes work for a new election (I have |
29 |
scripts which will build a master ballot from a collection of individual |
30 |
ballots, and from that the effort is minimal --- say, 60 seconds total |
31 |
if you are slow.) |
32 |
|
33 |
So, my "tedious" comment is someplace between misleading and wrong. |
34 |
|
35 |
I note in passing that NO voting method designed to select a set of |
36 |
several winners at once for this sort of election chooses quite the same |
37 |
winners as repeated application of Condorcet does. The deterministic one |
38 |
do agree with each other, however, on a slightly different set of |
39 |
winners. At some point after this year's election we should follow up |
40 |
on this. |
41 |
|
42 |
Regards, |
43 |
Ferris |
44 |
|
45 |
-- |
46 |
Ferris McCormick (P44646, MI) <fmccor@g.o> |
47 |
Developer, Gentoo Linux (Devrel, Sparc) |