Discussing the article: "Fast trading strategy tester in Python using Numba" - page 2

 

Cool article, thanks Maxim !

The speed is really impressive.


But it did take some work before running the scripts.

For those who can't run the script:

1) install catboost

pip install catboost

2) download the EURGBP_H1.csv data and put it in the Files folder.

3) download all python scripts and put them in one folder

4) tweak the first line of the script Tester_ML.py to be as follows

from tester_lib import test_model

5) write the path to the file in the Tester_ML.py script, I have it like this

p = pd.read_csv('C:/Program Files/MetaTrader 5/MQL5/Files/'+hyper_params['symbol']+'.csv', sep='\s+')


Tester_ML

 
Evgeniy Chernish #:

Cool article, thanks Maxim !

The speed is really impressive.


But it did take some work before running the scripts.

For those who can't run the script:

1) install catboost

pip install catboost

2) download EURGBP_H1.csv data and put it in Files folder.

3) download all python scripts and put them in one folder

4) tweak the first line of the Tester_ML.py script to read as follows

from tester_lib import test_model

5) write the path to the file in the Tester_ML.py script, I have it like this

p = pd.read_csv('C:/Program Files/MetaTrader 5/MQL5/Files/'+hyper_params['symbol']+'.csv', sep='\s+')



You're welcome. Yes, everyone works in different environments. For example, conda may already have all packages installed. That's why I didn't describe it separately, the interpreter will give warnings about what's missing.

I know from my own experience that a huge amount of text in an article scares you off. It's easier to figure it out on the spot than to read it all :)

 
Evgeniy Chernish #:

Cool article, thanks Maxim !

The speed is really impressive.


But it did take some work before running the scripts.

For those who can't run the script:

1) install catboost

pip install catboost

2) download EURGBP_H1.csv data and put it in Files folder.

3) download all python scripts and put them in one folder

4) tweak the first line of the Tester_ML.py script to read as follows

from tester_lib import test_model

5) write the path to the file in the Tester_ML.py script, I have it like this

p = pd.read_csv('C:/Program Files/MetaTrader 5/MQL5/Files/'+hyper_params['symbol']+'.csv', sep='\s+')

Great! You have done the work that the author of the article should have done.

 
😁
 
Write about trading strategies

Be above it all
 
Ivan Butko #:
Write about trading strategies

Be Above It All

It's expensive, unfortunately. If only the basic stuff.

 
Maxim Dmitrievsky #:

It's expensive, unfortunately. If it's just the basics.

SmartMoney has started to publish basic stuff, but it's kinda... dry. You can't get away on FVG alone.

In general, the field is open.
 
Ivan Butko #:
On SmartMoney they started to publish basic stuff, but somehow..... dry. You can't go on FVG alone.

In general, the field is open.
I don't know what it is and I don't want to :)
 
Maxim Dmitrievsky #:
In my heart I don't know what it is and I don't want to :)
Understood, no more insistence.
 

Turning trousers into shorts. Bonus material.

Strongly retraining the basic model as in the article. Two before and after versions:


Optimise now not sl/tp, but inputs by meta_labels (trade/not trade):



Additional library with modified optimiser in the attachment.

If you need a second article for this, with extended description, post here.

Everything works very fast! The TC generation process takes seconds. All that remains is to convert and export the models to the terminal.

Можно идти другим путем. Сначала экспортировать модель в терминал, затем делать оптимизацию. Интересно будет сравнить, есть ли выигрыш в скорости.

Files:
opt_lib.py  5 kb
This website uses cookies. Learn more about our Cookies Policy.