Machine learning in trading: theory, models, practice and algo-trading - page 1898

 
Maxim Dmitrievsky:

everything is already laid out in the idea - clustered seasonal patterns that supposedly repeat (and in fact sometimes do)

But... wrong coat. Or the tree is heavily overeducated and you need to train and parse a neural network

But it's all bullshit, if tree doesn't show anything then there is no regularity. There is no sense in deep learning.
The sample is either too small. The number of patterns or types of series is infinite by definition. Seasonality does occur, but probably not so obviously. Since there is no detectable
 

Who wants to test this insanity?

This is a strategy to return to the average at 1-4 hours. Each hour is traded according to its own logic.

The code in the inluder is generated in python in a few seconds, put it in the folder <include>

just compile the bot (st hours), apply the set

you also need an mt4Orders lib from fxsaber

I can't promise good tests on new data

In this f-i at the end of the inluder you can play with deviations. For example, multiply them by 1.5, 2, etc. The more, the less trades, but they should supposedly be more accurate.

void trade_func(double cl, double mcl, double std, double p) { 
if(cl == mcl) {
      if(countOrders(0) == 0)
         if (iClose(NULL, 0, 0) - p < -std)
            OrderSend(Symbol(),OP_BUY,LotsOptimized(), Ask,0,Bid-Stop_loss*_Point,Ask+Take_profit*_Point,NULL,OrderMagic,INT_MIN);

          if(countOrders(1) == 0)
         if (iClose(NULL, 0, 0) - p > std)
            OrderSend(Symbol(),OP_SELL,LotsOptimized(), Bid,0,Ask+Stop_loss*_Point,Bid-Take_profit*_Point,NULL,OrderMagic,INT_MIN);
   } 
 } 
 

It is not difficult to get such graphs. It can be obtained without any indicators and without MO.

I would also like to add that many users are testing on such demo accounts, where spreads from 0-5 pips (0-0.5 pips) and no commission. In this case you can make millions.

There are no such accounts on the real market. But for some reason it happens on the MetaQuotes Demo server.

 
Petros Shatakhtsyan:

It is not difficult to get such graphs. It can be obtained without any indicators and without MO.

I would also like to add that many users are testing on such demo accounts, where spreads from 0-5 pips (0-0.5 pips) and no commission. In this case you can make millions.

There are no such accounts on the real market. But for some reason it does on the MetaQuotes Demo server.

I'm not "getting charts", but testing the approach. I'm not interested in the MO theme - you're interested in the MO theme. If you're not interested in the MO theme, go to Edith Nahir.

 
Maxim Dmitrievsky:

Who wants to test this insanity?

It's a strategy to return to the mean at 1-4 hours. Each hour is traded according to its own logic.

The code in the inluder is generated in python in a few seconds, put it in the folder <include>

just compile the bot (st hours), apply the set

you also need an mt4Orders lib from fxsaber

I can't promise good tests on new data

In this f-i at the end of the inluder you can play with deviations. For example, multiply them by 1.5, 2, etc. The more, the less trades, but they supposedly should be more accurate.

You have to watch. And a good way to interpret the resulting logic to the real world. Saber bible stands already)
 
Maxim Dmitrievsky:

I'm not "getting graphs," I'm testing the approach. You've had enough of the flubbery. If you're not interested in the MO, you can go to Edith Nahir.

If the program has inputs where take profit and stop loss values are specified, then this result can be obtained without inputs, by optimizing a very simple strategy.

And what results do you get if you test it with the same parameters on another pair?

 
Petros Shatakhtsyan:

If the program has inputs where take profit and stop loss values are specified, then this result can be obtained without the MO, by optimizing a very simple strategy.

And what results are obtained if you test with the same parameters on another pair?

I have everything to try it myself. There is even no need to install Python. Just a saber bible.
 
Petros Shatakhtsyan:

If the program has inputs where take profit and stop loss values are specified, then this result can be obtained without the MO, by optimizing a very simple strategy.

And what results are obtained when testing with the same parameters on another pair?

The model is matched to a specific instrument.

 
Valeriy Yastremskiy:
Seems to have everything you need to try it yourself. You don't even need to screw in a python. Just a saber bible.

I would give the generator itself on python, but no one will use it

All the more that I redo something and will have to explain the version changes

 
Maxim Dmitrievsky:

The model is fit for a specific instrument

What does it mean? If everything is done with MO, you don't need any Take Profit or Stop Loss parameters. They should be generated automatically.