|
PureBytes Links
Trading Reference Links
|
> Monte Carlo sims can be extrememly misleading with respect to
> drawdowns estimations. ... The random order of
> trades has nothing to do with what occurred in reality. In
> reality, the system may go thru a long phase where it is out of
> step with the market thus creating a drawdown. An MC sim would
> completely destroy this string of trades and randomly assemble a
> string that has no basis in reality.
True. However, the MC sim also takes all those trades and
reassembles them in a gazillion different random permutations. The
chances are that some of them will at least match the drawdown caused
by the out-of-step condition, and very possibly exceed it.
Personally I agree 100% with your comment. It seems to me that an MC
sim assumes that all the events (trades) are independent and
unrelated to each other. If that's the case, then randomizing their
order should be just as valid as the original order. But the market
*does* get into modes for long periods, and the trading system will
react to those modes, so you don't have independent events. It seems
like scrambling non-independent events will result in invalid results.
But people who understand the statistics better than I do claim MC
sims are valid for trading systems. So I tend to believe them but I
take the MC results with a bit of a grain of salt.
Gary
|