Discussion of article "Creating Neural Network EAs Using MQL5 Wizard and Hlaiman EA Generator" - page 9
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Eh, it's all so depressing :( And really, there's no point in even debating....
Where are the results of trading?
Let scientists debate, traders trade, and programmers program....
As mentioned in the article, the information on trained neural networks is stored separately from the source MQL code, in the corresponding data files, which are loaded by Expert Advisors when they are run in the strategy tester or on a chart.
Now Hlaiman EA Generator has the ability to convert any of such neural network data files into the source code of two separate MQL4 and MQL5 indicators, which after compilation can be used independently, for example, for manual trading in MT4, MT5 terminals or when creating other Expert Advisors.
Moreover, this feature can be useful as a means of indirect debugging and optimisation of indicators by means of debugging and optimisation of Expert Advisors with neural networks, which then become prototypes of the generated indicators.
This may be especially relevant for MT4 users, where there is no option to directly run indicators in the tester.
The appearance and settings of the generated indicators are the same as those of the free test indicators published earlier in the Market. https://www.mql5.com/en/market/product/2551 https://www.mql5.com/en/market/product/2553.
The difference of the new indicators is that they are calculated according to the pattern of formed bars and are not redrawn.
Now with the help of Hlaiman EA Generator you can try to increase the trading efficiency of other, ready-made Expert Advisors, if they are presented in the source code and based on price movements, for example, on technical analysis.
For this purpose, a neural network filter is added directly to the source code of such EA, which can be initially included for training when running the EA in the tester, and then can be included in the work.
Variables are added to the EA settings to control the filter modes and the required degree of filtering.
Free sample of the Expert Advisor on the example of the standard Moving Average can be downloaded in the market, there you can also watch a video of training and testing processes.
https://www.mql5.com/en/market/product/8460
Now, with Hlaiman EA Generator can try to improve the efficiency of the trade of other ready EAs if they are presented in the source code and are based on the movement of prices, such as technical analysis.
To do this, directly in the source code of the adviser added neural network filter that can be initially included in the training run advisor in the tester, and can then be put into operation.
In EA settings added variables to control, filter mode, and the necessary degree of filtration.
Free Sample advisor on example Moving Average, can be downloaded from the Market, where you can also see the video, processes, training and testing.
https://www.mql5.com/en/market/product/8460
It's like some kind of double standards on the part of the administration. I remember I was banned a few minutes after the first link to the Market. ;-)
https://www.mql5.com/en/market/product/8460
In this example, the training of the neural network filter was performed based on the results of trading of the original Moving Average for 2014, the last update of the Expert Advisor - March 2015.
In order to check the effectiveness of the filter, I ran the Expert Advisor in the tester for the entire period after the publication, i.e. from April to the current date of August.
The first run was made with the filter disabled (corresponds to the original Moving Average), and the second with it enabled (see the marked variable UseNeuro = true), here are the results:
Thus, we can see that the neural network filter, trained last year, has not lost its efficiency over the past time, and it can increase the efficiency of trading almost twice.
Now, with Hlaiman EA Generator can try to improve the efficiency of the trade of other ready EAs if they are presented in the source code and are based on the movement of prices, such as technical analysis. To do this, directly in the source code of the adviser added neural network filter that can be initially included in the training run advisor in the tester, and can then be put into operation. In EA settings added variables to control, filter mode, and the necessary degree of filtration.
Free Sample advisor on example Moving Average, can be downloaded from the Market, where you can also see the video, processes, training and testing.
https://www.mql5.com/en/market/product/8460
In this example, the training of the neural network filter executed on the trading results of the original Moving Average for 2014, the latest EA update - March 2015.
In order to test the effectiveness of the filter I ran adviser in the tester for the entire period since the publication, ie from April to August the current date.
The first run made with a disabled filter (corresponding to the original Moving Average), and the second with the enabled filter (see. The marked variable UseNeuro = true), here are the results:
Thus, we can see that training in the past year, the neural network filter, over time, has remained effective and can increase the productivity of trade almost doubled.
Thus, we can see that the neural network filter trained last year has not lost its effectiveness over the past time, and it can increase the productivity of trading almost twice.
The pictures given by you say just the opposite: you should not use your Expert Advisor under any circumstances, because at the very beginning there is an unexplainable jump in profit, which is then squandered for a long time. And if this jump in profit is removed (who said that real trading will start with such a jump?), then in the first picture we see a fall, and in the second picture - eventually profit with intermediate drawdowns.
My article was published on the site, which shows that the problem is not in the model (neural networks or something more efficient), but in the initial data. The application of Rattle is shown, those who wish can buy a book from me, which is an extended version of the article. So with the help of Rattle you can understand one very simple and extremely important thing: the problem is not in the algorithm, but in the source data, which may or may not generate overtrained models. Here Rattle helps to experiment with input datasets in order to select those that do not lead to overtraining (overfitting).
And the choice of model is a tenth matter.
PS.
According to my research, using any kind of MA gives overtrained models, i.e. models that show excellent results on historical data and absolutely unprofitable on real data.