A quick and free library for MT4, much to the delight of neuralnetworkers - page 18

 
VladislavVG писал(а) >>

Right - I remembered a "trick" that was used back in "prehistoric" times

But it doesn't work with your dll.

Only if delay is >= 1000 ms.

 
lasso писал(а) >>

This does not work with your dll.

Only if the delay is >= 1000 ms.

That dll I linked to a stopwatch - that's why such a delay is needed. Then I converted it to ticks. Since a message about analogous solution appeared in the branch, I didn't post it.

>> Good luck.

 
Hi all, I modified FANN-EA.mq4 a little (I am not a programmer, so there are some errors) to determine if the next bar will close on the plus or minus position. I've been watching it all evening and it often guesses the direction (I haven't gathered statistics). I have already registered some errors in it and it has not corrected them. Before the test, I ran optimization on a 5-month period, 5-minute chart, the euro-dollar pair.
Files:
fannrea1.mq4  10 kb
 
Graff писал(а) >>
Hello! I modified FANN-EA.mq4 a little (I'm not a programmer, so there must be some errors) to determine if the next bar will close with profit or loss. I have watched it during all evening, it often guesses the direction (I haven't gathered statistics). Help to test, refine, fix bugs. Before the test I have carried out optimization on a 5 month period, 5 min timeframe, EURUSD.

How do you prefer to get the result? A graphical one? Or may be simpler? ;-))

And if seriously. I've added delay to create grids (see above), uncommented them here and there. Optimization - 6 months. I have tested it for 1 month. SL=700pp (new) Period - H4 (let's at least try to forecast it, not M5. We are not pilots).

Optimization period. Beauty!!!!!!!

Note in this test, the number of deals is comparable, although the period is six times less. WHY?


The test +1 month

Absence of a good teacher.

Files:
 

I don't understand the original version, at least - can this EA do locks (because not all brokers accept them for execution)? For when opening a long position, it seems to check only if some long positions are opened, but it doesn't look exactly whether short positions are opened.

Also, I wonder why in the input vectors of the MACD the loop parameter i, which is changed in increments of 3, is also multiplied by 3 - is it some special idea to use only every ninth candle?

 
lasso, thank you for your reply! At the moment I am using the EA for advisory purposes. On Friday my variant was tested on 121 bars of 5 min TF, of which 59 bars were correctly guessed. The result should be better on longer timeframes. For convenience I would like to add to my EA automatic optimization function (like during optimization running in the Strategy Tester) during initialization of the EA. I think in my case it is possible. Please advise how to implement this function, preferably with code.
 
Graff писал(а) >>
lasso, thank you for your reply! At the moment I am using my Expert Advisor for the advisory purposes. On Friday my variant has been tested on 121 bars of 5 min timeframe and I have correctly guessed direction of 59 bars. The result should be better on longer timeframes. For convenience I would like to add to my EA automatic optimization function (like during optimization running in the Strategy Tester) during initialization of the EA. I think in my case it is possible. Please advise how to implement this function, preferably with code.

59 vs 121 - you get the idea ))

.......

I don't have auto-optimisation implemented.

In general, there are solutions. For example: Automatic optimization of trading robot in the process of real trading

Search rules....

 

Is the problem the period or should the date of study be set higher?

 

Yeah, I test one result and then I start testing another one ))

all clear)

 
Can you please explain to a newcomer? Why can't I use this EA for trading?
I've optimised-over-optimised it and ended up grail-grail.
I tested EUR/USD H4 from 2006 to 2008 (by shifting a month forward after each optimization) and then substituted values for 2009, month by month, 10 times for each month, since they change each time. As a result, in 7 out of 10 cases I got a positive result, and so I did it for 2009. Where is the hidden clue? Why cannot it be used? I don't understand half of your speeches, explain in simple terms.