[FoRK] Fwd: Hey, big spender...
jbone at place.org
Thu Aug 28 19:11:38 PDT 2008
On Aug 28, 2008, at 8:46 PM, Lucas Gonze wrote:
> On Wed, Aug 27, 2008 at 10:28 AM, Jeff Bone <jbone at place.org> wrote:
>> That's the thing about Black Swan events. They may not be very
>> likely ---
>> indeed, you'll note that the high water mark for my prediction of a
>> GD given
>> Obama influence is "only" 1/3 --- but the impact is severe enough
>> that you
>> have to weight accordingly.
> The impact is also unlikely enough that you have to weight
> accordingly. Your 1/3 estimate is less likely than your 29% estimate,
> which is itself less likely than your 20% estimate.
Nonsense, get thee to a stats textbook. The magnitude of any estimate
(by any method) says nothing about its accuracy.
> I think there's a lot of meaty detail to look at here, and a mailing
> list isn't a context that will produce informative results. Something
> like a wiki spreadsheet would make more sense, and it would need
> fairly broad participation to become meaningful.
I've thought about that sort of thing a lot over the years. IMHO, the
best thing that could be done IMHO is to produce a Bayesian network
where everyone agrees on the topology, then use prediction markets to
weight the priors and conditionals. That would yield a result where
the prediction was entirely transparent. OTOH, simple prediction
markets of the top-level outcome, with enough participants, would
arguably be just as reliable... e.g., I suspect TradeSports is
probably a pretty good indicator of a lot of political propositions.
(In that case, accuracy is proportional to some function of the number
of contracts and the number of bettors.) The theory is that the
market factors in all the relevant and disjoint inputs and information
about the issue; the downside is that it's not transparent.
You can believe my estimate or not. I happen to think it's a pretty
good one, and I touch a lot of the critical factors involved on a day
to day basis professionally.
Time will tell. I myself am taking precautions in either case,
regardless of the outcome of this election.
And again --- extraordinary risks require extraordinary caution.
More information about the FoRK