|
PureBytes Links
Trading Reference Links
|
Ian Copsey wrote:
>In the original query there was no explanation of how the maximum
>DD was derived. If it is the one that is given in the system
>performance summary after optimization, then it is most likely
>totally fictitious. One method of establishing the MaxDD is to go
>through walk forward testing and study the drawdowns implied by
>this, if necessary calculating it in excel to overcome the breaks
>in between out-of-sample returns. Once you have done this over a
>long enough period - then add 30%-50%.
Another way to do this is through a Monte Carlo simulation, which
scrambles the order of trades and calculates new max drawdown values
for each new ordering, repeating several hundred (or thousand)
times, to generate statistics on the range of max drawdowns that can
occur from the trading system.
The Monte Carlo simulation can either use all the trades in each
scrambling (known as sampling without replacement), or it can
sample the list of trades randomly to generate a new list (known
as sampling with replacement). The "with replacement" sampling is
generally more robust (and generally results in a higher mean max
drawdown).
There are a number of products that do this, including
my own. There's a list of these, with links, at
http://unicorn.us.com/trading/competition.html
--
,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
// +__> Director of Research and Development
// \
// __) Unicorn Research Corporation -- http://unicorn.us.com
|