Machine learning in trading: theory, models, practice and algo-trading - page 1969

New comment
 
Maxim Dmitrievsky:

I was leaving... now it's been tested for over an hour since the beginning of 2019 to this day, I'll send you a screenshot soon

I can try to retrain the neural network on almost every bar.

I tried to retrain on every bar in the sliding window as an indicator, it did not work, how this one works I don't know

 
Maxim Dmitrievsky:

Holy...fuck.

:D

======

i want to know how it works!

 
mytarmailS:

:D

======

I want to know how it works!

so do i) and it would be nice to be able to write an analogue

if you write it and it works the same way, consider it solved)
 
Maxim Dmitrievsky:

Me too ) And it is desirable to be able to write an analogue

if you write it and it works the same way, you've solved it )

well 100% )

 
Maxim Dmitrievsky:

Holy crap.


It's beautiful. Maybe he's peeking.

You have to put the demo on to check.

Is the balance on the chart in money/items or in guessed/unguessed?
If the latter, then it is possible that "guessed" may be less in money than "didn't guess", and then the graph will be different.
 
elibrarius:

It's beautiful. Maybe he's peeking.

You have to put a demo on to check.

Is the balance on the chart in money/pips or in guessed/unguessed?
If the latter, it's possible that "guessed" may be less when converted to money than "didn't guess", and then the chart will be different.

Not a peek, in pips. The only thing I forgot to put in the markup, but it is 15 min at closing prices, spread should not cheat much.

I will check it on the demo.

I will check it on the demo.

80 000 points 5-digit profit.

Spread = 1600 trades *20 pips = 32 000. It's still positive.

 
Maxim Dmitrievsky:

I did not look, in points. The only thing I forgot to put in the markup, but it is 15-minute closing prices, the spread should not cheat much.

Plus the spread will be put in the revolds at once

And then I'll include the spread

I will check it on demo, there is nothing to correct for real

It is better to read the spread not from the bars, but from the tick history, because the minimum spread for all 15 minutes while the bar lasted is recorded in the bar, at the time of closing it will not necessarily be equal to the minimum.

I hope there will be a demo with the signal? It will be interesting to watch.
 
elibrarius:
It is better to read the spread not from the bars, but from the tick history, because the bar records the minimum spread for all 15 minutes while the bar lasted, at the time of closing it will not necessarily be equal to the minimum.

I hope there will be a demo with the signal? It will be interesting to watch.

I will do it. I have no VPS now, I do not want to work 24 hours a day.

 
Maxim Dmitrievsky:

I will. I don't have a VPN right now, I don't want to spend 24 hours on the computer.

There is a feeling of uncertainty... that even the electricity will not pay off as a result.
 
Maxim Dmitrievsky:

Max, how do you install this library? It's kind of a user's library

1...196219631964196519661967196819691970197119721973197419751976...3493
New comment
Reason: