|
PureBytes Links
Trading Reference Links
|
The big advantage of a mechanical system is you can backtest it (out of
sample) and know what your results going forward should look like. If,
using statistical tests, you find that recent results don't fit the
historical pattern, then something has changed.
Just having a one month longer drawdown than you've ever seen before
probably isn't statistically signficant. But if the average of your last
100 trades is 3 standard deviations less than expected, then that is very
significant. See any undergrad statistics text for an introduction to
significance testing.
Aaron Schindler
----- Original Message -----
From: "David Folster" <mr_bond@xxxxxxxxx>
To: <omega-list@xxxxxxxxxx>
Sent: Saturday, July 13, 2002 11:02 AM
Subject: Limited life span of mechanical systems?
> I have heard from several of the most experienced traders on this list,
both
> privately and publicly, about how "mechanical systems don't work forever"
> and how you have to re-evaluate them as time goes on, etc. I am
definitely
> not an expert on this subject, but I would just like to question this
point
> a bit further.
>
> How do you know when your system has stopped working?
>
> This just doesn't make any sense to me. What criteria do you use?
Wouldn't
> it be too late by the time you figured it out?
>
> Once Mark Brown posted to this list that his best ever bond system in the
> world had longest historical DD of 9 months. So, what then, your system
has
> stopped working if your DD goes for 10 months? A year? What if you
decide
> to stop trading it and then the system turns around to have a gigantic
> runup? To me, this whole idea seems analagous to trying to trade a
> mechanical system while at the same time second guessing it and trying to
> cherry pick only the best trades. You end up taking the losers and
missing
> the winners, no?
>
> Perhaps the more experienced traders on the list can speak to this
question.
> Not trying to criticize, only to learn.
>
> Thanks,
>
> David
>
|