From: John Conover <john@email.johncon.com>
Subject: Re: CNET.com - News - E-Business - Latest dot-com bomb: TheMan.com
Date: 4 Nov 2000 01:27:02 -0000
There are a lot of folks using tsinvest on the election tracking
polls, (everyone is saying its too close to call, so everyone is
hunting for alternative numerologies.)
You don't have to use tsinvest to do it. You can figure it out in your
head using the exact method that tsinvest does.
The current (Friday AM,) tracking, polls say Bush 49%, Gore 43%, with
a margin of error of +/- 3%.
The first thing that has to be done is to judge the accuracy of the
numbers-tracking polls are statistical samples. Note that it is very
close to a zero-sum game; what Bush gets, Gore loses, and
vice-versa. So, (the numbers work out quite nicely,) if the margin of
error is +/- 3% per standard deviation, (thats what a statistical
estimate, or margin means,) the numbers would be wrong, (Bush less
than 49 - 3, and Gore more than 43 + 3, or more than equal for Gore,)
one standard deviation of the time, or 16%, (or correct, 84%,) of the
time.
So, if the election were held today, (and depended only on the popular
vote,) Bush would have an 84% chance of winning.
But what about next Tuesday-4 days time. A trip to http://www.cnn.com
gets the tracking poll history since early September. Just eyeballing
the graph, it looks like the the fractal of the difference between
Bush's and Gore's tracking poll wandered around within a 10% range,
(i.e., each was +/- 5%, or between 40% and 50%, with a mean of 45%,)
and it took about 18 calendar days to do it. So, sqrt (18) * R = 0.05,
where R is the standard deviation of the daily increments. So, R would
be about 1%. And, in 4 days time, it would be about sqrt (4) * 0.01 =
0.02, or about 2% is all that Gore could hope for, with an 84%
probability.
What that means is that for Gore to take away 3% from Bush, and add it
to his own, would be 1.5 standard deviations, or he could pull it off
only 6.7% of the time, and 93.3% of the time he couldn't.
So, Bush has a 0.84 * 0.93 chance of winning, or about 78%,
(considering only the popular vote,) based on the accuracy of the
tracking polls, and the ability of Gore to move them.
Its interesting because the popular and electoral votes may differ for
the first time in history, (a very small chance, but not
insignificant.) But as far as the popular vote is concerned, it looks
like Bush has callable edge.
John
BTW, if you want to play with tsinvest, the -C and -c options control
the way the probabilities are computed for the accuracy of the value
of a stock. The method is almost exactly the same, (except that the -c
uses 1 / erf (sqrt (t)) instead of the approximation, 1 / sqrt (1)).
John Conover writes:
> BTW, the election is kind of different this year. For the first time,
> the Presidential candidates' media strategy is being centralized
> around the Internet, (as opposed, in recent history, to television
> and, before that, radio.) This includes the web sites, (which are
> huge, heavy hitters, doing streaming media, etc., in addition to
> slinging web pages-a million pages an hour is fairly common,) and
> e-mail, (what most folks would call spamming-johncon attack
> G. W. Bush's site for it yesterday; twice.)
>
> How do I know this? It seems that there is a battle of web sites
> raging, and everyone is watching everyone else's web site. And, how do
> they do that? See: http://www.johncon.com/nformatix/ for particulars.
> (The IT staff of all candidates are using it.)
>
> John
>
> BTW, there seems, so the story goes-I only talk to the candidates' IT
> techies-that there is an information war going on; it seems that
> information/disinformation strategies are being used on web sites
> since everyone knows that everyone is watching everyone else's site.
>
> What is the outcome of such a scenario? Well, (surprise, surprise,)
> the characteristics of the tracking polls will be a Brownian motion
> fractal, and the underlying game-theoretic structure will be
> zero-sum. If everyone's sites have equal power, (which they do,) the
> game will be fair.
>
> The tracking polls will just kind of wander around, one candidate
> gaining, at the expense of the other-for a while-and then the other
> way around-with the constituency all confused as what the numbers mean
> (sound familiar?); and like it or not, the poll numbers do determine
> votes.
>
> But one should not play fair zero-sum games, right? Ultimately, no
> player wins, right?
>
> Well, kind of. One solution is to force the informatics, (what the
> Linux community calls "social engineering",) to deteriorate into a
> prisoner's dilemma kind of scenario, i.e., tell the biggest lie about
> one's opponent, first, just before the election, so it will also be
> the last.
>
> The reason everyone is watching everyone else's site is to find out
> who is going to cave, and play the defection strategy first,
> (hopefully, one's opponent-so its a high stakes bluff waiting game,
> followed by a quick and concerted reaction.)
>
> Bottom line, expect mud slinging to intensify through the following
> few days; the candidate that is behind in the tracking polls on
> Thursday, or so, has nothing to loose by defecting first, (and
> everything to gain,) since the opponent knows it, and would never cave
> first. The idea is to sling just enough mud to be ahead on Thursday to
> be in the strategic position of being able to sling mud last, if one
> has to.
>
> John Conover writes:
> >
> > BTW, using the Internet this way is an important paradigm in itself.
> > The Internet is about what people want people to know about what they
> > know, (for good or bad, hate and sex is out there with the politics,
> > commerce, news, and tautologies of the 21'st century-an evolving
> > snapshot in time of what we know, or think we do, which is in about
> > the same proportions as the general populace; so a search engine can
> > tell us a lot about where we have been, who we are, and where we are
> > going. For biz applications, see http://www.johncon.com/nformatix/.)
> >
>
--
John Conover, john@email.johncon.com, http://www.johncon.com/