Machine learning in trading: theory, models, practice and algo-trading - page 3601

New comment
 
Maxim Dmitrievsky #:
I remember a case. Some time ago, I was invited to the office of a DC to give a lecture on currency markets to the department where the programmers were sitting. I told something that I had prepared, because the programmers did not understand the sphere at all. And the main question after the lecture was about data providers. That is, where to get good quotes. They were given such a task and they could not solve it.
That is, if you think that all DCs have a good understanding of what they do - no, they don't understand anything either.

and you didn't send them to Ice-Data ?

didn't know that ... :-)

 
Maxim Kuznetsov #:

and you didn't send them to Ice-Data?

didn't know ... :-)

I don't know what it is, it's not like I'm a broker. There are a lot of different quote providers.
I know only the fastest ones, for arbitrage.
 
Maxim Dmitrievsky #:
Which bank? Banks have a different line of business. Retail forex has nothing to do with banks.

Absolutely not!

Puppet.

 
Dmytryi Nazarchuk #:

Absolutely!

A doll.

Very interesting conclusion )
 

Here I sit and wonder if there is any point at all to these digital perversions.....

Strategy from 2007 to 2010

The same strategy on a longer horizon 2010-2024 - here I got it by the statistical method in the sense, but in fact I took one quantum segment after marking by this indicator - 200 take and 50 stop. It is traded, but the stop is 300 and the take is 800. Buy and sell is determined by the direction of the last segment ZZ.


In 2024 it works like this for now - I did not adjust it initially to this period - I just checked it.

Anyway, here you go. Yes, you can apply more MO methods, although there are not very many deals there - a little more than 1000 for 2010-2024, but that's not the main thing. The main thing is how to choose strategies to work at a certain point in time. Whether their readings are random or not... If a strategy did not work for 3 years, then 5 years at breakeven, and then showed growth - what e there is so different....

Here I think, maybe it is necessary to do some A-B testing, but again, you can just fit will be - and it is not clear whether the reason for the different behaviour of the strategy in the identified condition, or it is random....

And so, the method itself allows you to make grails - there are a lot of variants of such quantum segments....

Strategy logic

   if(Strateg==Test_Q_Otrezok_V_00)
   {
      if(isNewBar_TF(PERIOD_H1)==true)
      {
         double Get_iD_P_21_H4_1627=iDeltaf(Symbol(),PERIOD_H4,7,6,0);
         if(Get_iD_P_21_H4_1627>0.20489999651908875 && Get_iD_P_21_H4_1627<=0.27416667342185974)
         {
            int V_ZZ=(int)ZZ_Vibor(Select_ZZ,7,0);
            if(V_ZZ==1)  BuyNow=true;
            if(V_ZZ==-1)SellNow=true;
         }
         if(CountMarketOrder_OS>0 || CountMarketOrder_OB>0)
         {
            BuyNow=false;
            SellNow=false;
         }
      }
   }
 
Aleksey Vyazmikin #:

And so, the method itself, allows you to rivet grails - variants of such quantum segments wagon ...

The logic of strategy

There are a lot of minuses, we need to filter further

 
Maxim Dmitrievsky #:

There are a lot of minuses, we need to filter further

It's easy to do that, but how to know whether the filter will include more of the necessary signals on the new data or not - that's the question...

 
Aleksey Vyazmikin #:

It's easy to do, but how to know if more of the right signals will get into the filter on the new data or not - that's the question...

cv

 
Maxim Dmitrievsky #:

cv

Naive.

 
Aleksey Vyazmikin #:

Naive.

Not naive without her?)
1...359435953596359735983599360036013602360336043605360636073608...3713
New comment