|
PureBytes Links
Trading Reference Links
|
Here's a comment from my friend Dave Chamness, who originally told me
about the Saitta technique:
====
Just as the normal distribution emerges from non-normally distributed
things, so the random walk can emerge from summations of non-random
walks. Dice have a flat distribution, 1-6 are equally likely, with a
mean of 3.5, and a standard deviation of 1.8708. If I add dice rolls
and subtract 3.5 from each, then I will see a random walk. If I add
10 rolls at a time, I will see a normal distribution in the sum, even
though each individual roll does not have a normal distribution.
Saitta added actual market changes, selecting days randomly. He
compared that to the actual order of days. He found that the changes
in the actual data were bigger than expected from a random order, so
up days do cluster together.
The argument about an individual day containing trends does not
invalidate Saitta's measurement of random walk. The argument that
days may not all have the same standard deviation also does not
invalidate Saitta's measurement of random walk.
What Saitta did is quite robust, and it did find some trendiness.
|