[EM] Condorcet strategy spreadsheet (was, ...maybe it should be about Condorcet...)

[Jameson Quinn]

2010/1/31 Juho <juho4880 at yahoo.co.uk>

> One more addition / clarification.
>
> I wrote:
> > The three candidates just happened to plan their campaigns so that they
> generate a cycle. That could well happen.
>
> One could say that it is not probable that those three candidates would
> form such a cycle for the same reasons that voters usually do not have such
> cyclic preferences. But for single candidates that is much easier. And there
> may well be some minor things that change the opinions. One candidate might
> e.g. have lost credibility on topic X due to some personal problems in that
> area, or due to external propaganda (if not planned so by the campaign
> office and the candidate). Cycles are thus maybe not a very common pattern
> but surely possible in typical real life situations.
>

Generally, there are "natural alliances" in the voter population. Issue A
and B are, for whatever reason, seen as closer to each other than they are
to C (think of the at-times-contradictory grab-bags that are classed as
"left" and "right"). Campaign managers would tend to council the A candidate
to have a subfocus on B and vice versa. This would make such cycles less
probable than the ~8-20% (depending on the model) they would happen by
chance. Furthermore, traditional polling would have a hard time detecting
the cycle.

Still, I admit its possible. Say all of the above factors cut it down by
only 50%, and the base probability is 20%. Under these pretty-generous
assumptions, a detectable cycle happens under 1 in 10 elections. Now, you
have to further assume that the polls are accurate enough to detect who is
the likely honest winner. At most, you're at 8%. Now, you cut down to just
the groups who have an available strategy; together, they are less than 1/3
of the voters (or else their candidate would be winning). So, as a voter,
you have a priori around 2% chance of having a rational strategy available.

Really, it's not much.

Jameson Quinn
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20100131/d1babcc3/attachment-0003.htm>