October 2004. In the interest of my friends
who are currently suffering from
political
frenzy, angst or high-spin-uppedness...here is the remarkable
JWC TradeSports-Based Presidential Election 2004
Oracle!
Right. Now the facts, please:
TradeSports is a betting site, run from Dublin. People buy
contracts on all sorts of wagers (usually of a sporting nature),
including political outcomes. Go check their website if you
want to know where I get the raw data that is the basis for this
trivial exercise.
It happens that TradeSports sells daily betting contracts on the
state-by-state outcomes of the Bush vs. Kerry event. It
is perhaps a not-unreasonable hypothesis that people around the world
who probably want to make a few shekels on a bet are less
biased /
engaged in wishful thinking about the outcome of the election; they may
care (or know) enough to study a state's political situation
sufficiently, and thus make a reasonable assessment of its propensity
toward
Bush or Kerry. Whether that hypothesis proves out is ... TBD on 2
November!
Anyhow, here is what leads to my daily
predicted outcome that you can see on this graph (remember that 270
electoral votes are needed for either candidate to win)...
The inner product of the 51 element constant vector of state
electoral
votes (50 states + DC) and the weight vector
derived from the time-varying contract bid values as
follows:
For a given state contract <=
$50 (i.e., Bush will lose there, by definition)
=> weight = 0
For a given state contract >
$50 (i.e., Bush will win there, by definition)
=> weight = 1
Note that, although the contract vector definitely changes from
day to day, the inner product might not change. In any case, what
you see is a plot of the "free market" electoral vote prediction as a
function
of day until the election. Forget those stupid
"opinion" polls!
Helpful Suggestions for people who are still Troubled by the
election
outcome...
- Contemplate the short, enigmatical life of E.Galois,
who conceived the basis of group theory.
Happy Voting!!!!
November 18, 2004:
Of course, we now know the results of the Election, with Mr. Bush
defeating Mr. Kerry and Mr. Nader quite decisively...
Bush: 59,834,866 votes, with 286
Electoral Votes
Kerry: 56,373,514 votes, with 252 Electoral
Votes
Nader: 406,940
votes, with 0 Electoral Votes.
See USA Today for a good
map. So the TradeSports-based model actually performed quite
well.
One final note. The dramatic dips seen on November 2 in the late
afternoon and early evening were due to the infamous "insider trading"
that arose when media people gained access to flawed exit poll
results! I wonder if they lost lots of $$$?
August 6, 2008:
Here is some historical
poll data, showing just "how wrong they can be"! Also see the
graph.