Machine learning in trading: theory, models, practice and algo-trading - page 827

New comment
 
The forum moderators surprise me. It's as if they don't exist.
 
Grigoriy Chaunin:
I am surprised at the moderators of this forum. It seems that they do not exist.

No... There are. But they are responsive when you complain. So back from the ban now I know who's a jabida-jabida.

And about my grail. It is not deflated, and continues to work at all costs. I'm just making it up as I go along. Because money loves silence...

 

Now it's so funny, if someone says something to me. I'll say back.

"You hear me, I've been banned for five days." Sounds serious and threatening :-)

 
Aliosha:

Looks like you're kidding, I wrote above, for those who are in the tank: chips - returns with exponential window from the past, target return(s) from the future, MO - XGB, RF, whatever MLP, SVM, the result is not much different. The result is 52-54% accuracy, Sharp ratio in the tester is 2-3


PS: This is of course a test example, to refute the use of ZZ as a target, in combat systems, everything is much more complicated

Alesha - well done. Writes everything to the point, without fear. That's right, there's nothing to be afraid of. No one will do anything anyway, it's been checked. People just need the Grail in the form of a trading signal, and that's it. And those few who really examine the market, because they have invested time and soul into their work, can't abandon their ideas by virtue of inertia. The 1% that they take away from the market is THE 1%, their merit, and they will never turn from their path. If they add something from the outside and get another +1%, it will not affect your personal achievements and well-being. There is no competition.

Thank you for your attention.

 
Alyosha:

Frankly speaking strictly formally I have not proved why so and besides not I about it have guessed, the information has been received some years earlier, in private correspondence with influential enough (on the market) algotrader, but then when some gentlemen openly here have written about it, I too have considered that not a secret and it is possible to support.

In my opinion there is nothing to prove, everything is quite obvious. For example, when we do regression we approximate some linear vector function {y1,..., yn} = f(x1,...xk), probably not linear, but by mixing the x's to the target, we artificially add a linear component, {y1,...,yn} = f(x1,...xk) + k1x1+...+ knxn which is easy to regress, instead of the sought f(x1,...xk). You yourself wrote that:


all right, linear dependence will be obtained, on which the model is overfitted

But using a zigzag is low-level, why do you need it at all :)

 
Grigoriy Chaunin:
How do I post it? I'm posting it. https://github. com/RandomKori/PythonForTraders There is a profit margin. I have a very good Expert Advisor on my trading robot. I am too lazy to write a document. I made it for myself. There is no history export script. I will show it in a minute.


Thank you!

 
Grigoriy Chaunin:
Shall I post it? I will, sort it out. https://github.com/RandomKori/PythonForTraders There are some prerequisites for profit. I don't know why I decided to trade on this article. I am too lazy to write a document. I made it for myself. There is no history export script. I will show it in a minute.

write everything in the form of articles - and the presentation will be complete and you will be compensated for your time. It is not convenient to study in fragments :)

I don't want to learn useful things in particular:

  • a quality MT5+Python bunch - you'll get +100500 to your karma for this article.
  • And if you want to have a look at some packages and ideas, you'd better have a look at some ideas.
If you don't like it, it's a banal disrespect to the forum members - ask some advice, try it out... nobody will waste time for nothing
 
In the Python MT bundle, I still wrote the docks. I will not write a big article. I don't know how. I don't have time for it either. In addition, your advice would be in your ears. I was asked to lay it out. I did it only for myself. There all the same genetic programming. And in general I am in a trance. The opener lost its second trial. You can find the details on the Internet. Keeping money there is too risky. BCS hasn't bothered to open an account for me for over a month. Also to hell with it. There is no broker on the Russian market with MT5 anymore. Finam sucks. They do not hesitate even with binary options. Where do I look for a broker with MT5?
 
Grigoriy Chaunin:
I wrote some docks in Python MT bundle. I won't write a long article. I'm not really good at it. I also don't have time for it. In addition, your advice would be in your ears. I was asked to lay it out. I did it only for myself. There all the same genetic programming. And in general I am in a trance. The opener lost its second trial. You can find the details on the Internet. Keeping money there is too risky. BCS hasn't bothered to open an account for me for over a month. Also to hell with it. There is no broker on the Russian market with MT5 anymore. Finam sucks. They do not hesitate even with binary options. Where do I look for a broker with MT5?

For the opener is useful information. Thought about working with them in the end...

 
This is Kolmogorov, not just anyone. I recommend it for reading.
1...820821822823824825826827828829830831832833834...3512
New comment
Reason: