Renamed the EA, instead of EMA, changed to MA. Actually initial version was named incorrectly as it was using Simple MA and not Exponential.
Anyway now you can try 4 different MA types and I already noticed quite an $$ improvement.
length option and shift options were added.
So now it is much more tweakable,
Post your tests, comments, observations
EA does actually trade, places orders, stops, targets. Because this is a very simple EA, it should forward test nearly identical to back tests. This is my hope. It's too early to tell. After running on demo for a week, I can back test it on the same data to compare back/forward tests. If they match well, we might have a very predictable and easy to optimize EA.
Optimizations showed that MA periods provided by Elliot 13,26,233 are very close to optimal indeed.
Optimization also showed that by using an exotic mix of MA Type, versus period, versus offset - produces a bit more $$, at least 25% more.
Last edited by Metatrader7; 08-18-2009 at 19:08.
So actual method is multi time frame and actually uses 3 MA per time frame and more in total. Of cause we don't have to go all the way to weekly. Realistically we should use 2-3 time frames:
M15,M5,M1 or H1,M15,M5
We don't have to program Multi Time Frame, that can only complicate the system and the accuracy of testing.
MAs can be easily scaled by multiplying length. For example 13 period MA on M15 = 39 period MA on M5. Well, almost, at least it is very very close and tests with close results. Remaining slight difference can be compensated through optimization.
So multi time frame MA periods will be:
for 1 min:
M1 M5 M15
13 65 195
26 130 390
233 1165 3495
for 5 min:
M5 M15 H1
13 39 156
26 78 312
233 699 2796
So this is a proposal for multi layer 3X3 version, so you can take your existing version and simply add extra conditions from 2 more sets of higher period MAs.
So insted of 3_MA_inverse, we should have 9_MA_inverse EA. May be you can come up with 6_MA_inverse version and we can compare the results for all of them.
You can continue with your version, I can add more tweaks later, the same way I already did with your initial EA.
We need more versions for testing
Let me know on your progress.
Last edited by Metatrader7; 08-18-2009 at 20:15.
The real point is overall logic(maximizing profit/lowering drawdown), MA combinations. The rest is to be tailored to market, broker, time frame...
I am running optimisation for M5 for all possible periods,types,shifts MA combinations. It takes about 24h, but results I am getting boosting $$ by 50% already and I wouldn't be surprised if it doubles that by optimization end.
Coming with multi time frame version
Last edited by Metatrader7; 08-18-2009 at 21:13.
Gitaniko you posted the EA, could you post your .eab settings for 3EMAinverse from eatbuilder.com
That will help,
Dependency on others is a bad habit
Because when the project runs its course the only thing left you can rely upon is your own skill
All I do is "thinking aloud". There is some magic in freeing your mind by sharing ideas, you get more and faster that way. That helps my own creative process and hope few people around