Machine learning in trading: theory, models, practice and algo-trading - page 326
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
This branch is huge.
Can someone give me a hint...
I have charts of movements of several currency pairs. How can I use machine learning to select parameters (lot, direction) for opening/closing orders so that the result is positive as often as possible?
What should I do, how should I train the program?
If you have a good idea, you may use the following search engine, but if you don't have any problems, just go to the "search engine" and select "neural networks", there is a lot of useful information.
In our case we use the optimizer to adjust neuron weights, that's all... what difference does it make whether you train it in logic or through the optimizer... And in terms of speed I think that in the cloud learning is much faster via the optimizer
Is 1000% in two months bad or what? :) improved the logic a little bit
But it's true that the bulk of the profit was made in April. Since mid-May there is even a stable tendency
2000% in 3 months, but the drawdown is 64%, which is logical for such a yield :) I'm done spamming... but Reshetov's Tax Report is definitely the thing, the main thing is to think up cool predictors
And what about this amazing run, 3000% in 3 months with a 55% slump, it's almost perfect.
2000% for 3 months, but the drawdown is 64%, which is logical for such a yield :) I'm done spamming, but Reshetov's PH is really something, and the main thing is to think up good predictors.
And here's an amazing run, 3000% in 3 months with a drawdown of 55%, almost perfect
This is not spam.
I've been waiting a long time for the results.
Otherwise there is no point in even reading.
2000% for three months, but the drawdown is 64%, which is logical for such a yield :) I'm done spamming ... but Reshetov's PH is definitely something, the main thing is to think up great predictors
And here's an amazing run, 3000% in 3 months with a drawdown of 55%, almost perfect
Ok. What happens on the demo?
I haven't put it in yet, it will be the same, because the test is based on opening prices, the results are very reliable.
I have not fully implemented all my ideas, these are intermediate variants... I may have some more funny ideas...
This branch is huge.
Can someone give me a hint...
I have charts of movements of several currency pairs. How can I use machine learning to select parameters (lot, direction) of order opening/closing, so that the result is as often as possible in the plus?
So what do you need to do, how to teach the program?
If out of the box - you can take the history of bars from MT5, export it into csv, and train the neuronics to trade every bar in the positive direction. The lot will be constant during training. It would be better to determine it later on by the Expert Advisor during trading depending on the current balance.
In the Strategy Tester you will get a perfect trade on this timeframe. But it will fail on new bars. This is not a trading strategy, but a rigid fitting to the history.
And it is suitable only for large timeframes, such as M30, maybe M15. On smaller spreads will eat up all the profits.
Do you want? :) If yes, then I will add here the example code of how to do it, but you have to set R (https://cran.r-project.org/bin/windows/base/) in order to start training of the neuron.
If it is a quick fix - you can take the history of bars from MT5, export it to csv, and teach the neuron to trade every bar in profit. The lot in the training will be constant, it is better to determine it later by the Expert Advisor during trading, depending on the current balance.
In the tester you will get a perfect trade on this timeframe. But on new bars you'll lose. This is not a trading strategy, but a rigid fitting to the story.
And it is suitable only for large timeframes, such as M30, maybe M15. On smaller ones the spread will eat up all the profit.
Do you want? :) If yes, then I will add here the example code how to do it, but you will have to set R (https://cran.r-project.org/bin/windows/base/) in order to start training the neuron.
Still intriguing. I will have slightly different input data.
Gotta
If it is a quick fix - you can take the history of bars from MT5, export it to csv, and teach the neuron to trade every bar in profit. The lot in the training will be constant, it is better to determine it later by the Expert Advisor during trading, depending on the current balance.
In the tester you will get a perfect trade on this timeframe. But on new bars you will lose. This is not a trading strategy, but a rigid fitting to the story.
And it is suitable only for large timeframes, such as M30, maybe M15. On smaller ones the spread will eat up all the profit.
Do you want? :) If yes, then I will add here the example code on how to do it, but you have to set R (https://cran.r-project.org/bin/windows/base/) in order to start training of the neuron.
Can I show you a little manual on R and how to train simple neural networks with examples?
Can I suggest a little manual on R and how to train simple neural networks, with examples?
There are examples in R itself. True, not on our topic, but they explain a lot.
But in general, I like SciLab (MathLab analog) better. In terms of neural networks it is poorer - only 3 packages (R has 11), but the help is better, and the orientation of R is different, while SciLab is focused on scientific and technical calculations and real-time modeling.
I work with both, but with SciLab more. Feel free to join in.
There are examples in R itself. True, not on our topic, but they explain a lot.
But in general, I like SciLab (MathLab analog) better. In terms of neural networks it is poorer - only 3 packages (R has 11), but the help is better, and the orientation of R is different, while SciLab is focused on scientific and technical calculations and real-time modeling.
I work with both, but with SciLab more. Feel free to join in.
Will it be possible to connect SciLab to mt-chemicals?