|
PureBytes Links
Trading Reference Links
|
Christian writes:
>
> I am also constantly thinking about applying Monte-Carlo
> simulation on a trading model which is based on a
> portfolio of markets. This is IMO very hard because
> the various markets are or could be intercorrelated
>
I wonder whether you've read the whole article
all the way to the end? I think you may find
Section II. particularly relevant to the concern
above, especially Figures 4 through 7. Take a
look and see what you think.
Mark,
don't wonder, I have read it till the end - more than once ;) Your
made a clear conclusion in Section VII.
What I ask for is an opinion of doing that type of
monte-carlo-simulation one step earlier. Imagine an extreme example
where the trades are fitted (perfect example would be a over trained
neural net) perfectly to the past. So doing MC on that result-set brings
not much.
So the original question was and still is, if you or others see a value
in testing with artificial data, be it scrambled from real data or
derived with methods like Mandelbrot's. Test runs with such kind of data
can be done multiple times with the same trading model.
Purpose is the same as yours - having a statistical more meaningful
information about a trading model.
Christian
|