Machine learning in trading: theory, models, practice and algo-trading - page 2285

[Evgeny Dyuka]

Rorschach:

Have you ever used Brython? It's python for the browser.

the only way to get free graphs a la tradingview is the tvjs opsor project, which is node.js

[Renat Fatkhullin]

Jonathan Pereira:

It would be interesting to have access to the data in the order book.

If we are talking about ticks, there are functions copy_ticks_from and copy_ticks_range

Here is more details: MetaTrader for Python.

[Maxim Dmitrievsky]

Renat Fatkhullin:
Can you share the information:

1) Do you use the python library of MT5?

2) Do you use it outside or inside MT5?

3) What functions does the library lack? Access to indicators?

We are preparing an upgrade of MQL5 adding fast matrix operations. This will allow to perform massive calculations normally.

We will also develop connectors to analytical packages and implement the standard WinML integration.

1. Yes, always

2. On the outside, as I'm used to VScode and jupyter (very convenient for research)

3. So far everything is enough, I don't use indicators. I don't use indicators yet, but I can use simple MT5 model porting tool like ONNX, but I haven't studied it yet.

Trade API also tried, everything works fine

[Renat Fatkhullin]

Evgeny Dyuka:

The ML theme is getting away from you.
I use MQL5 to collect data and then to prepare current data for polling the neural network. Everything else is in Python.
MQL in this chain only due to old inertia, because I started with it, otherwise all these issues are solved in Python. Of course MQL are speed and clarity, but at the same time:
- crutches to obtain data from crypto exchanges
- inability to directly interact with api of crypto exchanges to trade
- inability to put your EA in the Market without opening the code (it is impossible to pass automatic validation if there is a web checkout)
- total ignorance and inability to use MQL terminal (everybody use a browser).

It's not going anywhere.

There are no crutches with cryptocurrencies - just find an MT5 broker with direct integration with cryptocurrencies.

If all of the EA's logic depends on webrequest and the EA is not able to make independent trades, it will not pass. Because in this case, the trader's protection is violated and there is an absolute space for falsifications.

You are wrong out of ignorance and ineptitude. Everything has long ago reached the level of using a cell phone.

Your statements show that you imagine the whole world as "there is nothing but python". If you had taken up programming 20-30 years earlier you would have seen where the real work is going on (workhorses on tens and hundreds of millions of consumers) and where the research on crutches.

Research in R/Python for release into the commercial market (tens and hundreds of millions of users) necessarily requires almost complete conversion to other more productive languages. As a result, "garage vs factory" corresponds to "Python vs MetaTrader 5".

We are working to get rid of the transition process and develop MQL5 so that it allows to do the most of its work.

[Renat Fatkhullin]

Evgeny Dyuka:

You missed the moment when, about three years ago, trading went massively into the masses. And it's not just crypto.
All the plans you described are cool technically, but they're geek stuff, they won't save you. If you want to jump into the last train, you need to urgently make a web version of the tradingview level, with all the functionality of the mql5 terminal.
Take this project as a base and develop it, or the train will pass, leaving you with 1.5 geeks.

We have been dealing with trading platforms for over 20 years.

Our first mobile terminals were developed back in 2001 on Palm, we also developed web terminals long ago, including those embedded in other applications.

We will soon release a new web terminal for MetaTrader 5. The kernel is ready:

Here you can see it in the Quotes section and you can also easily insert it right here in the editor:

[Renat Fatkhullin]

elibrarius:

A suggestion for the MQ.

In MO most of them use OHLC bar data for training. I.e. real ticks do not make sense, because they are many times longer. For example, I use the tester by opening prices.

I would like to have the second candlestick with OHLC prices at Ask instead of the minimum spread for Bid bars.

On a flat spot double the amount of historical data?

No, use either M1 spread information or tick data for that.

[Renat Fatkhullin]

Aleksey Mavrin:

1) Not yet, but I will have to. While for ML I use solutions written in MQL by habit to work "all in one place".

2,3) Not figured out how to use inside, I think that mqh-wrapper interfaces for popular ML Python libraries would be in demand.

Will there be matrix operations with GPU computation capability?

We will develop direct integration with WinML, as we have done for OpenCL and DirectX.

In addition, we have a big project to include C++ modules/packages similar to other languages. That is, it will be possible to convert a lot of open source libraries into packages.

Matrix operations at least on CPU/Multithreads/AVX, but possibly with GPU.

[Maxim Dmitrievsky]

Renat Fatkhullin:

We will develop direct integration with WinML, as we did for OpenCL and DirectX.

In addition, we have a big project to include C++ modules/packages similar to other languages. That is, it will be possible to convert a lot of open source libraries into packages.

Matrix operations at least on CPU/Multithreads/AVX, but possibly with GPU as well.

Perhaps it makes sense to make it with a reserve for automatic differentiation (as in PyTorch, for example). I.e. tensors with possibility to calculate gradient on them, and on top it is possible to use any neural network

But it's a lot of work, probably like writing your own neural network engine

[Evgeny Dyuka]

Renat Fatkhullin:
It is obvious from your statements that you imagine the whole world as "there is nothing but python.

I see the whole world as a businessman. I have a product and promote it. I am looking for better ways to promote it. MQL5 does not work for product promotion, this is my personal experience, nothing more.

- I tried to post an article on my topic in MQL, but they told me to fuck off...If i tried to place it in MQL5 I was sent to Messages with the phrase"not enough volume, sorry, we won't even read it",
- MQL5 Market is not available for my product,
- out of 10 clients only 1 (at best) uses or has used MQL terminal,
- "just find MT5 broker with direct integration with crypto-borders" this is a crutch, that in real life does not work, because crypto-borders is the only direct broker, and they all have great api.

In my opinion, you (MQLs) are immersed in your world and measure who has longer and thicker code.

[Forester]

Renat Fatkhullin:

Double the amount of historical data on a flat spot?

No, for that use either the information about the spread in M1 or tick data.

The volume of M1 data (OHLC * 2) is several times less than the volume of real tick data.

Minimum M1 spread is not suitable for accurate trade estimation.

Yes, - now only from real tick data, so we will pump traffic.