Machine learning in trading: theory, models, practice and algo-trading - page 2389

 
elibrarius:
2-3 days of studying Python and something simple - like launching a catbust. All the more, there are examples in Maxim's articles.

There still need to work with analysis of arrays and with files - well, I'm not a born programmer.

If it is not interesting, I will do what I can do in MQL5 - I have a lot of ideas and have little time to check them.

 
Aleksey Vyazmikin:

You just didn't understand the point I was trying to make - the best model in terms of classification statistics doesn't mean the best in terms of profitability. It is so only in the case of fixed SL and TP.

I'm looking for a method to influence the income and expense curves - the green and red curve.

This is what the probability distribution of the model response to the sample looks like when trained:

This is how it looks when the independent sample is fed:

As you can see, the curves almost merged, while the patterns didn't deteriorate as much-the aqua curve is zeros and the magnet curve is ones-they're quite acceptably spaced, and the patterns are sort of globally preserved, but the cost of those patterns wasn't sort of weighted in terms of income/expense.

of course I don't get the point, because the 1st time it was about trying out traits and finding the best ones, and now it's about a way to influence income curves

I will get bogged down with such "help with code" for a long time and pointlessly, because it is not even clear what problem is being solved

So, the best solution would be to do nothing :)

 
Maxim Dmitrievsky:

of course I don't understand the point, because the 1st time it was about enumerating the signs and finding the best ones, and now it's about the way to influence the income curves

I will get bogged down with such "help with code" for a long time and senselessly, because it is not even clear what problem is being solved

That's right - the selection of predictors that positively affect the increase in profitability without reducing the statistical performance of the model.

I would explain in more detail if I had interest, but I just don't. Well, maybe another time.

 
Aleksey Vyazmikin:

That's right - the choice of predictors that positively affect the increase in profitability without reducing the statistical performance of the model.

I would explain in more detail if I were interested, but I just don't have it. Well, maybe another time.

I'd rather write an article.)

I have it all in my articles. And learning in the loop and selection of models by external criteria

 
Maxim Dmitrievsky:

I'd rather write an article.)

Well, of course - at least they'll pay for it :)

 
Maxim Dmitrievsky:

I have it all in my articles. And learning in a loop and selection of models by external criteria

I read it all - there's not even a representation of the model that I posted in the form of a distribution graph - without it it's impossible to evaluate the model - I've come to this now.

 
Aleksey Vyazmikin:

I read everything - there's not even a representation of the model that I posted in the form of a distribution graph - without it, it's impossible to assess the model - I've come to that now.

It's all a monkey's work, these graphs... Charts for the sake of charts. It doesn't solve the main problem.

It is easy to visualize the model by different means
 
Maxim Dmitrievsky:

It's all a monkey's work, these charts... charts for the sake of charts. It doesn't solve the main problem.

Well, that's if you do not feel the essence of what is displayed.

And what is the main problem, please, state it.

 
Aleksey Vyazmikin:

Well, that's if you don't feel the essence of what's being displayed.

And what is the main problem, state it, please.

I have accuras directly affects the profitability

I'm not interested in the rest. I'm not interested in other similar problems.

 
Maxim Dmitrievsky:

My accuras has a direct impact on profitability

The main thing - the search for patterns, the rest does not interest me at all. Drawing and counting - not interested at all.

I completely forgot, that you have classes, which are responsible for the trade direction, and I have classes, which allow/forbid to trade - that's why you can't feel the usefulness of a chart :)))

If I wanted to handle this problem I would have to search for regularities by enumerating the various correlations of predictors, simply the increments are unstable and it's necessary to widen the range, at least add a daily ATR(3) to it.