Machine learning in trading: theory, models, practice and algo-trading - page 431
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
What should I set in the EA, in the magic number, so that the order I have placed, it will consider it as my own. now it is 777. platform mt4.
Open an order using a script... Or try to put 0 in the Expert Advisor.
What should I set in the EA, in the magic number, so that the order I have placed, it will consider it as his. now it is 777. platform mt4.
any number, and you will use it to refer to the positions. IMHO
Machine learning is a very interesting topic, I am a complete zero in it. But, I have a question, has anyone tried to train the system on open orders on the history - that is, the training was aimed at analyzing the situation at the time of opening - here it opened in profit - yes, here was an indicator or candlestick pattern, and here - a loss of a different model - we will not do so.
Machine learning is a very interesting topic, I am a complete zero in it. But, I have a question, has anyone tried to train the system by open orders on history - i.e. the training was aimed at analyzing the situation at the moment of opening - here it opened in profit - yes, there was an indicator or candlestick pattern, and here - a different pattern made a loss - we will not do that.
On the whole you are right, but this is a slightly different approach. Anyway, this is how I do it:
An event occurred (a signal from the underlying strategy, which by the way is a counter-trend, but never mind) At the time of this event, a set of indicators had that value. And at each event, we have certain values of the indicators, which we feed to the input of the NS. We train the NS on the history, where the result of each event is known to us. Accordingly we train our NS on historical data, when the result of the event is known to us.
NS works in real time. An event happens and we took the values of the indicators and fed them to the input of the TC, in response we got the result of working of the TC in the form of 1 - signal will bring profit. -1 signal will bring a loss. This result is calculated by the NS on the basis of training on past data. In short, it looks like this. If you want more details, here. read the article
In general, you are right, but there is a slightly different approach. Anyway, this is how I do it:
There was an event (a signal from the underlying strategy, which, by the way, is a counter-trend, but not the point) At the time of this event, a set of indicators had a value. And at each event, we have certain values of the indicators, which we feed to the input of the NS. We train the NS on the history, where the result of each event is known to us. Accordingly we train our NS on historical data, when the result of the event is known to us.
NS works in real time. An event happens and we took the values of the indicators and fed them to the input of the TC, in response we got the result of working of the TC in the form of 1 - signal will bring profit. -1 signal will bring a loss. This result is calculated on the basis of ES training on past data. In short, it looks like this. If you want more details, here. read the article
To continue the idea, many hand traders say that they feel the market, but cannot formalize it - so I think that their unconscious makes a decision based on its vast experience, and there are two options: 1 - Trade purely on emotions (getting angry - selling, hungry - buying), and 2 - Trade by picture (unconscious analysis of the situation). Here, if you take reports from signals, especially from people trading hands, or your own reports, it would be possible to identify this unconscious pattern, wouldn't it?
And I think that analyzing indicator readings on 1-2 bars is not enough - you need to make a mold of the screen - i.e. consider the direction and nature of the movement of indicators and price in general.
Have you tried to train on ZZ patterns?
I will continue the idea, many hand traders say that they feel the market, but cannot formalize it - so I think that their unconscious makes a decision based on their rich experience and there are two variants: 1 - Trading purely on emotions (angry - selling, hungry - buying), and 2 - Trading by picture (unconscious analysis of the situation). Here, if you take reports from signals, especially from people trading hands, or your own reports, it would be possible to identify this unconscious pattern, wouldn't it?
And I think that analyzing indicator readings on 1-2 bars is not enough - you need to make a mold of the screen - i.e. consider the direction and nature of the movement of indicators and price in general.
And you didn't try to train on ZZ patterns?
TC keep your voice down about ZZ. You should not say it aloud in this thread, or you'll get tomatoes. :-)
First, I analyze data for the period of the window, which I have a floating. This is one.
Secondly. I think it would be possible to analyze my trade history, or rather it would be possible to analyze it, but not to apply it. I will tell you once again. In order to begin to analyze the market you need a fait accompli. In my case these are signals from the counter-trend strategy. In my case, too. Or you can analyze each bar. The trader has opened a position emotionally, so this event has no regularity of occurrence. How do we know that it is time to analyze the market? When the trader opened the position? Well, then it will be too late as the position has already been opened. So. First, we formalize the moment of decision-making. Usually this is a basic strategy.
Well, the fact that the price went in both directions during the same pattern can be considered either a random process or the tricks of a puppet who tries to break strategies of both pattern finders directly on charts and users of NS in order to worsen forecasts, or rather to reduce them to near zero, as in the figure.
I have partially solved this problem by increasing the number of bars in a pattern (template) and moving to a larger timeframe. Then the patterns start to give less scattered predictions. But it's still far from the grail, something else is needed.
For example on eurusd H6 - I have a pattern of about 1000 bars and the search for similar patterns goes to the depth of history also about 1000 bars (for example in the extreme case a similar pattern is taken from bars with indexes 2000-1000).
And I also decrease "similarity" for those patterns, which are too far away in the history.
Unfortunately all this mess will not give even 100% profit per year (the risk is 20%).
All right, Vasily. Tell me how you open a trade???? Based on what? Let's say it's a trading day, you have selected your TF and are ready to rush into action. So the question is, under what circumstances do you open a trade???
This I can only tell my students, knock in a personal account, let's agree, now one student has just taken a study leave, I can teach you an in-depth course of exchange trading based on psychology and understanding the motives of the major players (insiders, puppets). So far I can tell you one thing:
It's not at all like you imagine, almost the opposite!
Hurry up! I'm not the Lord God, I can't train hundreds of students at a time.
This I can tell only my students, knock in a personal note, let's agree, now one student just took a sabbatical, I can teach you an in-depth course in stock trading based on psychology and understanding the motives of the major players (insiders, puppets). So far I can tell you one thing:
It's not at all like you imagine, almost the opposite!
Hurry up! Soon enrollment will be closed, unfortunately, because of the excess of those who want to learn, I'm not God, I can't train hundreds of students at a time.
Unfortunately psychology is primarily a work on yourself, I prefer to work on the AI and to hell with that psychology. Although I would not refuse to train my stamina.....
Well here write that the price comes to equilibrium after large deals and news for a few minutes, that is, something can be caught probably only on M1
This probably applies to those who work with limits, takeaways, stops. My MO is not that strong yet.
I'm trying to predict trends so EA always sits in position with a fixed lot and only occasionally flips when trend reverses. For me M1 is noise and a lot of problems.