|
PureBytes Links
Trading Reference Links
|
Larry:
This is great. But how do you compute a Hurst component?
- Stuart
>>> Lawrence Lewis <lel@xxxxxxxxxxxxxxxx> 06/03 1:25 PM >>>
The Hurst exponent is named after J.M. Hurst, a hydrologist trying to
design reservoirs along the Nile delta in the early 1900's. Nice coffee
table trivia.
As far as traders are concerned, if you analyze a time series the way
Hurst did (called rescaled range analysis), and compute a Hurst exponent
of .5, then the time series is presumed to be random: the value at time
t+1 is not dependent on any previous values. Values other than .5
indicate some type of dependence on prior values.
The theory goes back to a random walk. If you move in a random direction
from a fixed point (each move is unrelated to any prior move), on average
(e.g. if you run this experiment a whole bunch of times) you will find
yourself the square root of the number of moves away from the starting
point. The square root is equivalent to an exponent of .5 If, on average,
you find yourself farther away - an exponent greater than .5, then
somehow, the direction you're going is dependent on the prior direction -
you're trending. If, on average, you find yourself closer to the starting
point - an exponent less than .5, you're anti-trending - if one move
takes you to go further away from the starting point, then the odds are
better than 50-50 that the next move will bring you closer.
Larry Lewis
----------
|