> Alternatively, you may export your time series to a txt file
and
Sorry, forgot to say that you have to discretize the time series
(or
better the returns), e.g. using 8 values, then you generate
random
time series using the same alphabet of 8 values.
If you don't
discretize in such a way that all letters have the same
frequency, then it
is better to use shuffled versions of the original
(discretized) time
series.
Btw, on the NYSE100 stocks (EOD data), I found no correlation
at all
between predictability of the time series and profitability of TA
(if
anyone is interested by the paper, just drop me a mail), but I
might
not have tested the right TA rules
..
Regards,
Nicolas