[maemo-community] STV voting strategy

[From: Benson Mitchell]

On Fri, Mar 20, 2009 at 7:26 AM, Dave Neary <dneary at maemo.org> wrote:
> There is another way to count, which only makes sense with electronic
> counting (and is thus harder to verify), and that's fractional transfers.
>
Not precisely true -- it only makes sense with _logical_, rather than
physical, ballots, but the actual counting can be done with pen and
paper, especially for an election this size.

But the issue here, to me, is that  fractional transfer may be _hard_
to verify, but it's always verifiable. In random transfer, some
elections will always turn out the same way, and these are easy to
verify. But some elections will involve some close call that depends
on the random choice. In these elections, the results with fractional
transfer are _impossible_ to verify. There's always the possibility
that the "random" selection was cherry-picked to get the desired
outcome. For this reason, I greatly prefer the fractional transfer.


> We could use fractional transfer if you want - and it could even be
> retroactive to this election.
>
> To ensure we weren't running the risk of controversy, I ran the election
> under all variants of STV available, with different methods of
> calculating thresholds and transfers. I can confirm that under all
> scenarios I tried, the results remained the same.
>

For this election, I rather expected (from what I saw of the results,
or even just from the number of candidates) that it wouldn't matter.
But to safeguard future elections against corruption or any appearance
thereof, I'd strongly support a switch to fractional-ballot transfer.

>> I didn't mean you should vote for other than your #1. I just think that
>> with ranking method the next positions have bigger effect so it really
>> doesn't matter much if you put someone on first or second place (or rank
>> them as 100 or 90 on 0-100 scale) when council has 5 members. With STV
>> it is very significant since your one vote is counted really only for
>> one person (and most likely the first one on the list if the candidate
>> is popular).
>
> Yes, this is true. In STV, you are voting primarily to elect your
> preferred candidate and ordering is very important. We do not run the
> election with a pairwise run-off system like Condorcet (which isn't
> suitable to multi-seat constituencies, and which is very hard to
> understand).
>
As a curiosity, you may be interested to know that under a oneliner
pairwise system (simply counting who has the most pairwise winning
ballots against all other candidates), this election's top 3 were
predicted correctly, and 1 of the other two predicted winners was also
an actual winner. (I published these results on IRC last night,
labelling the top 3 as my predictions, and the others as bogus...)
Algorithm in sh, grep, wc, and dc:
$ for i in 1 2 3 4 5 6 7 8 9; do for j in 1 2 3 4 5 6 7 8 9; do grep
$i.\*$j.\*0$ ./election.blt|wc -l; done; echo ++++++++p;done|dc


Benson Mitchell