Re: Newbes, mechanical systems, expectancy, etc.

[Herman van den Bergen <psytek@xxxxxxxx>]

One more time on mechanical systems :-)

I think many will agree with me that for a newby the last year was a bad
time to start trading, especially with imperfect trading systems. I am in
the newby category and made several false starts that cost me money. My
advice: be patient, don't jump in untill the time is right, wait for a few
good signals and have an exit strategy when you enter. Starting a system is
the most difficult time because your are trading with your hard earned
money or savings, not with profits.

Be cautious when optimizing your system, Steve has been suggesting to keep
it simple, and I agree completely. If your system doesn't work without
optimization it probably never will. If your system doesn't work, go visit
the archives, some helpful traders have given away some real good starter
ideas and even formulae!

Some people argue that backtesting is useless, backtesting with heavy
optimizing: yes, I consider that useless. But if you test a "basic" system
over a ten year history and it produces an almost straight equity line
(allow for some drawdowns!) and gives you half-a-million% or more return
over that period, heh, I wouldn't throw out the formulae. If you really are
disgusted with it: send it to me :-)))

If you write a system that randomly generates trades within a certain
duration, say 2-9 days, and you run a few thousand tests, you'll easily
find one that makes 500%+ ann. profit. Same if you pick days of weeks, or
whatever other randomizing method. The point is that if you overlay a
thousand random trading histories on a price start there will alway be one
that looks good. And then to think that at times we think that the MS limit
of 32000 optimizations is not enough :-(

Finally, when your perfect mechanical system gets overly complicated (I sin
too) remember that all your formulae use the same basic data: OHLC. What
does that mean? Well, I think that if you had a formula that was two pages
long and you would use a super computer that could simplify the algebra,
remove all redundant components and sub formulae etc., it would probably
come up with something only a few lines long. Think about it...

Happy trading,
Herman.