Flaws in the roulette wheel by Curtis McKallip, Jr.
Few games of chance are perfectly random. To the extent they are NOT, profit may be made by betting
on those states which occur with greater than random frequency and against those which occur with less
than random frequency. In the late 19th century, William Jaggers, a British engineer, hired six men to
write down the winning numbers on a roulette wheel for a month of play. By identifying the numbers
which came up with greater-than-random frequency and then betting on them, he earned a profit of 1.5
million francs. The anomalous numbers were created by a roulette wheel which was poorly balanced.
What happens if Jaggers' method is applied to commodity trading?
Jaggers' source data was a roulette wheel, ours is the marketplace. Ideal sources should be "ergodic" and
"stationary." Briefly, an ergodic source produces a time series which, given an interval of sufficient
duration, will return to states which are closely similar to previous states. A stationary source produces a
time series whose statistical properties do not vary with the choice of time origin. All ergodic sources are
stationary but not all stationary sources (for example, one that gets "stuck" in a certain state) are ergodic.