Machine learning for robots - page 6

 
Ivan Negreshniy:

So you can not, moderator will reject compiled with protection, because it must be the source, and the source will reject, because it is necessary to protect the rights of the seller - vicious circle works:))

But there's nothing surprising here, because the legal status of robots of different professions and in the object environment, so far, is poor...

So don't shine the EA on. Just a file with trades in the tester on the history as a source for the MO (from where - doesn't matter - a magpie on the tail brought it, and it makes no difference). And then your results on test and demo.

Sort of where we started - what we have come to. But all can be confirmed, checked in a short time. You do not have to wait till somebody marks something out and sends it to you.

As far as I understand, you only need deals.

 
I lost my mood ( I wonder why it happened so). I marked only entries, the system is not reversed, the number of purchases and sales is not even
 
mytarmailS:
Oh man, half an hour marking chart on a pattern, saved the template, loaded it and my signals have turned out to be mixed up with the standard arrows, with those originally in the EA, I do not feel like it( I wonder why it happened so. I marked only entries, the system is not reversed, the amount of buying and selling is not even.

As I wrote at the beginning of the branch, there is an auxiliary advisor makeSignals, which itself draws signal arrows and can also filter them according to different trading strategies - indicators.

The auxiliary EA has the following settings:

  • Signal bars count - number of bars by which the signal is calculated
  • Buy signal pips - estimated number of points of profit for the buy signal
  • Sell signal pips - calculated number of points of profit for the sell signal
  • Start date time - beginning of a period in which signals are calculated and applied
  • End date time - end date of the period in which signals are calculated and applied
  • Arrow drawind type - type of the graphical object - arrows used for drawing signals
  • Indicator used type - type of indicator used as a filter of signals
  • Clear all on exit - delete all graphical objects when the Expert Advisor is disabled

The Expert Advisor searches within a given period and plots on the chart all signals that match the calculated parameters (number of bars and number of pips) and filters them, if the indicator used, such as ZigZag, EMA, MACD, OsMA, CCI, RSI, RVI, etc.

Information about the signals is displayed in the comment line - an interval, the size in points, and the current number of BUY and SELL, respectively.

To form a pattern, the Expert Advisor runs on the chart with the selected time interval and the delete flag removed, perhaps several times, for example if the strategy is a combined one, then the signals can be edited, deleted or changed and the ready pattern is saved to a file (*.tpl).

If necessary, I can fine-tune the code of this EA, add new filters for strategies you described or you may do it yourself.





Files:
makeSignals.mq4  31 kb
 
Ivan Negreshniy:

As I wrote at the beginning of the branch, there is an auxiliary advisor makeSignals, which itself draws signal arrows and can also filter them according to different trading strategies - indicators....................

Well that's what I used thismakeSignals. I ran "make signals"

(Here's just an example)

with these parameters

deleted signals

saved the template.

I open the template again and get the same as it was in the beginning


 
mytarmailS:

Well, that's what I usedmakeSignals for. I ran "make signals"

(Here's just an example)

with these parameters

deleted unnecessary signals

saved the template.

I open the template again and get the same as it was in the beginning


Perhaps you have not removed the remove signals flag - Clear all on exit or the EA itself from the chart, before saving the template and it is now reloaded, every time you view the template.

If the reason for loading the EA, you can delete the link to the EA directly in the template text.

 
Yuriy Asaulenko:

So don't shine the EA on. Just a file with trades in the tester on the history as a source for the MO (from where - doesn't matter - a magpie on the tail brought it, and it makes no difference). And then your results on test and demo.

Sort of where we started - what we have come to. But all can be confirmed, checked in a short time. You do not have to wait till somebody marks something out and sends it to you.

As far as I understand, you only need deals.

Your proposal basically works, but it is more difficult to do clean experiments using cloning EAs than indicator signals because EAs often use all sorts of manipulations with pending orders, averaging, trailing stops, etc., and this obscures the work of the calculation model.

As an example, I generated two clones of the MACD Sample.mq4 Expert Advisor from the terminal standard delivery; I trained one of them on trades on all available EURUSD H1 MetaQuotes-Demo history and the other one on a period from the beginning of 2018, the model based on open prices.

Checking in the tester showed that both clones duplicate the trades of the original, somewhere between 50-70%.


However, the test reports of the original and the two cloned EAs are very different.

MACD Sample

MACD Clone All

MACD Clone 2018

As can be seen from the charts, the most profitable clone trained on the period of the current year 2018, despite the fact that in the sample formed by the visual test of the original MACD Sample, used for its training only a few dozen deals (see the template) and the fact that for him actually all history beyond 2018 is OOS.

If anyone is interested to take a closer look at how the clones work, I can send you EA files in my personal message.

Files:
 
Ivan Negreshniy:


Here you go, I think I got it done...

There aren't many trades because I set them up manually, but you can try


it's for a 5 min chart

Files:
 
Yuriy Asaulenko:

Why not do the following for experiment: download some super-duper strategy from Market, run it in the tester (we trust the tester)), and submit the results to NS, RF, SVM or something else. And we don't have to wait - we try it on the demo and see the results.

The signals asked for training are not needed from the Market. Let's rub the noses of the best signal traders!)
 
Dmitriy Skub:
Why do I need signals from the Market? They are begging for training. Let's rub the noses of the best signalers!))
It's all good, except signal traders have no trade history. Or am I wrong?
But in general, I don't see how TC can get anything useful out of 200 trades. Three or five thousand is all right. Well, 1-2 thousand, maybe.
I wonder if it works. I will have to reconsider my views on the IO.)
 
Yuriy Asaulenko:
Everything is good, only signalers have no transaction history. Or am I mistaken?
But in general, I don't see how TC can get anything useful out of 200 trades. Three or five thousand is all right. Well, 1-2 thousand, maybe.
I wonder if it works. I will have to reconsider my views on the IO.)

Why absent? Just present).

You can pick it up with a lot. The main thing is that the expected payoff should be larger.