[extropy-chat] silent night

[Harvey Newstrom]

On Dec 30, 2004, at 10:00 PM, Robin Hanson wrote:
> I don't think you actually read the paper.  Try Table 3, "all years"
> column, first section.  It shows that our of 596 comparisons, their
> market beat polls 451 times.  This data is *not* just 24 hours ahead.

I read the paper.  It stated that "existing" evidence failed to 
demonstrate predictions more than 24 hours before the election.  By 
"existing" evidence, it refers to prior evidence.  It claimed that 
prior analyses had never been attempted and that prior evidence had 
never been achieved.  This paper claimed that it would be the first 
attempt at such analysis and present the first evidence for the first 
time.  My point was that they claimed there were no other sources of 
such evidence except for this paper.

> You also seem to have no idea how academic publishing works.  I'm sure
> they have been submitting the paper places and it has so far been
> rejected.  That's how things go in academia.  Most of us have perfectly
> good papers that have been rejected for several years.

I am sorry to hear that.  As a published technical author, I know it is 
hard.  But that doesn't change the fact that this is still an 
unpublished, unfinished, draft manuscript, that has not been 
peer-reviewed or reproduced by others.  Even if this is no one's fault, 
it still has not yet reached the level of "proof" that you seem to 
ascribe to it.

> 596 is small?

No, but the table with four elections that I cited is.  I also 
explained how they used an average over a poll, but took a single 
last-date number for the market.  The fact that their later number 
beats earlier poll numbers didn't impress me that much.  Likewise, 
knowing the number of times they beat the market in this way, but not 
by how much doesn't impress me that much either.  We could see a count 
of their claim for how often they were right, but not the actual data 
for our own review of how right, how often, and how they defined right. 
  Their dates also seemed to be arbitrarily chosen, possibly to show 
good examples, instead of using regularly spaced days or all days for 
analysis.  So while you can give some examples that aren't small or 
flawed, that doesn't refute other examples that are small and are 
dubious.

> It is not my job to go look up every relevant item, and spoon feed you 
> a URL or a quote,
> and then explain them all via long emails.

It is when you want to win an argument based on proof that you can't 
seem to find at the moment.  I concede that maybe there is proof, maybe 
you have seen it, but maybe you can't remember exactly where it is 
right now, and you are too busy to look for it.  But frankly, so am I.  
If you can't quote it off the top of your head as an expert in the 
field, and the Iowa market claims no such proof outside their single 
paper exists, and I haven't been able to find it in my previous 
searches either, then I have little interest in going on another 
fishing expedition right now.

I fully concede that Iowa markets might predict elections.  I believe 
markets are a public opinion poll, as are elections.  I makes perfect 
sense that they should be correlated.  I am just surprised that nobody 
is using this tool to predict future elections for great financial gain 
if they really are working as expected.  i was l looking for a quick 
explanation why this isn't happening, and haven't found one yet.  No 
big deal, I guess.

--
Harvey Newstrom <HarveyNewstrom.com>
CISSP, ISSAP, ISSMP, CISA, CISM, IAM, IBMCP, GSEC