From: John Conover <john@email.johncon.com>
Subject: Re: Bush/Kerry optimal wagers
Date: 2 Nov 2004 01:18:55 -0000
The movements in polls have high entropy short term dynamics, and if
one candidate is ahead the chances of the candidate remaining ahead
for at least one more day is erf (1 / sqrt (t)), where t is the number
of days the candidate has already been ahead. Using the Zogby poll,
for the last 8 days:
Bush / Kerry Difference
48 / 47 1
50 / 45 5
47 / 46 1
47 / 48 -1
46 / 50 -4
47 / 49 -2
47 / 48 -1
45 / 49 -4
Or, the chances of Bush's short term lead continuing through tomorrow,
November 2, is erf (1 / sqrt (4)) = 0.521143, which is very close-it
is very unusual that short term dynamics have to be used to forecast
the outcome of an election for optimal wagers.
John
BTW, one can not make money wagering on the erf (1 / sqrt (t))
distribution of run lengths; the tail of the distribution starts out
with a very significant slope, and gets very shallow after about 4
days-just enough, that in the long run, negates what one makes on the
steep slope.
--
John Conover, john@email.johncon.com, http://www.johncon.com/