Expert Advisor Development

[quant...@gmail.com]

Based on my observations and adventures in MT4 programming there
appears to be two basic contingents out there. One I would call the
quick strategy type and the other the realistic strategy type. And I
must admit that when I first started in this I was in the quick
strategy category. As things went along, after trying more than 1000
coding combinations and reviewing hundreds of expert advisors it
became rather clear that there was more to this business than meets
the eye.

I was pleased to see that most of the guys interviewed in the World
Trading Championship were of the same basic view. They however
classified the difference as between those who have figured out all of
the coding issues in MT4 and those who haven’t figured it out. And one
thing for sure is that there is a very large learning curve in
figuring this out.

If you get the chance to read the interviews I would highly recommend
it. You should be able to find a link on the MT4 Official Forum site.
The interview I liked best was with a programmer out of Russia who was
with a company that was devoted to creating successful Expert
Advisors. He said they had developed over 70 profitable experts using
the MT4 library indicators and the one he had entered used 17
different pattern detection codes in the M15 time frame. He was not
going to sell his program, opting to keep it for his customers – in
Russia.

Another thing he said that struck a cord with me is that the emphasis
in Expert Development in his opinion was going to be toward pattern
recognition on single pairs. In this direction it is about looking for
patterns, and programming the pattern which essentially can be done as
“fixed” parameters. They will then optimize only for Take Profits and
Stop Losses. He said that even then they had to do adjustments because
he believed that the optimizer actually finds “minimal” profit
combinations and not “maximal” combinations.

The development protocol they followed which I also believe is a very
necessary part of the process is to first do their optimizations, then
6 month and 1 year historic tests. This is followed by at least one
month of forward testing, then another month of mini or micro account
testing. Only if the Expert Advisor survives all of these tests do
they consider it suitable for live trading. Now the first thing to
know from this is that from start to finish it takes at least 3 months
to create a stable and profitable Expert Advisor. Secondly, if you are
going to spend this amount of time on this program you should take
some time to make sure your strategy is well thought out and has a
VERY good chance of working AND being programmed.

One of the reasons for the necessity of all of these tests is the fact
that the MT4 compiler, and all of the various test modes seem to have
different coding tolerance levels. With each round of tests you will
find that more tweaking to the program is required. Things the
optimizer doesn’t catch will appear in the historic tests. The forward
test will reveal more problems. It is the same for the live tests.
Only when an Expert Advisor demonstrated consistent results in
historic tests using Control Points, and Tick by Tick, and these
results match the performance in forward tests and in live tests is it
really suitable for trading. And this contestant stated that all of
their Experts would do this.

Those people in the “quick strategy” group will not believe you if you
tell them that it is possible to get the same results in all of these
different test modes. Obviously, at least one person has resolved the
coding issues with MT4 AND obviously, the FAP group is not of the
second category of MT4 programmers.