Machine learning in trading: theory, models, practice and algo-trading - page 1556
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 does the deposit consist of? From the buy/sell/wait commands.
These commands will be taught to the final NS. And then to predict them. What should the intermediate nets be trained on?
Here you can simply maximize profits.
In other words, the points of opening and closing of positions are identified that, as a result, give profits. The potentially higher profit is obtained when entering the determined entry point, the higher is the signal level on the grid exit.
Here you can simply maximize profits.
That is, the places of opening and closing positions, which as a result yield a profit, are identified. The potentially higher profit is obtained when entering at the identified entry point, the higher is the signal level at the grid exit.
Intermediate ones need something specific to teach. If there isn't something specific, they are useless.
What does the deposit consist of? From the buy/sell/wait commands.
These commands will be taught to the final NS. And then predict them.
What should intermediate networks be trained on? ZigZags? For a network to learn something, it needs to show an answer. What algorithm of the zigzag and with what parameters would you like to use as a training signal?
We don't need any zigzag algorithm. The result here is somewhat similar to that used in AlfaZero. The network detects the entry points by itself.
All of the zigzags and indicators are crutches. The market is not a stationary system. Fractals appear in different forms and at different timeframes. They all have a simultaneous effect on the market.
What is a zigzag or a muving? It's just letting the data flow through a certain algorithm that is a certain filter. The algorithm doesn't take into account the size of the resulting fractals. It just grinds the information into something average. And if you feed this hospital average to the input of the network, that is, in fact, feed distorted information to the input, then the output will also be not quite adequate. The beauty of neural networks is that they themselves identify and classify the fractals present in the received signal.
You don't need any zigzag algorithm. Here you get something remotely similar to what is used in AlfaZero. The network itself detects the entry points.
All zigzags and any indicators are crutches. The market is not a stationary system. Fractals appear in different forms and at different timeframes. All of them have a simultaneous effect on the market.
What is a zigzag or a muving? It's just a passing of a data stream through a certain algorithm that is a certain filter. The algorithm doesn't take into account the size of the resulting fractals. It just grinds the information into something average. And if you feed this hospital average to the input of the network, that is, in fact, feed distorted information to the input, then the output will also be not quite adequate. The beauty of neural networks is that they themselves detect and classify fractals present in the received signal.
That is, you don't need intermediate networks. One network will do everything. In the end we come to what we have. That no one has a decent signal to confirm the success of the MO.
I agree.
That is, you don't need intermediate networks. One network will do everything. In the end we come to what we have. That no one has a decent signal to confirm the success of the MO.
This is done by one network.
The intermediate ones should be trained. If there is nothing specific, they are useless.
Each network identifies places to open positions for potential profits. And signals from these networks are fed, for example, to the full-connection network. Groups of inputs from networks from different timeframes receive potential signals.
And the full mesh network identifies where it is better to open/close positions on the basis of getting the maximum rate of deposit increase.
If we buy EUR 1-10-2000 and hold position till 1-07-2008 following the monthly timeframe, we will gain one profit. But if during this time trades will be opened both to buy and to sell, say, on a daily timeframe, the profit will be potentially larger. The Full Boundary Grid is designed to detect those places where the profit will be growing the maximum.
And if we open on H4 or H1... or on the minutes. The final grid will identify where the profit is better and will also take into account signals coming from higher timeframes. Here, redtrader.ru, for example, you can see how signals from different timeframes are considered. But there everything is done manually.
Each network identifies places to open positions for potential profits. And signals from these networks are fed, for example, to the full-connection network. Groups of inputs from networks from different timeframes receive potential signals.
And the full mesh network identifies where it is better to open/close positions on the basis of getting the maximum rate of deposit increase.
If we buy EUR 1-10-2000 and hold position till 1-07-2008 following the monthly timeframe, we will gain one profit. But if during this time trades will be opened both to buy and to sell, say, on a daily timeframe, the profit will be potentially larger. The Full Boundary Grid is designed to detect those places where the profit will be growing the maximum.
And if we open on H4 or H1... or on the minutes. The final grid will identify where the profit is better and will also take into account signals coming from higher timeframes. Here, redtrader.ru, for example, you can see how signals from different timeframes are considered. But there everything is done manually.
Unfortunately, in all your reasoning there is one gross mistake. One word "itself". Unfortunately, this is a misconception that the network itself can not do anything and until you understand this nothing will change.
You saw the link to AlfaZero above. And this grid itself learned to play Go. And then the DeepMind team, using similar ideas, created grids that began to win in various computer and not only computer games. Think about it.
You saw the link to AlfaZero above. And this grid itself learned to play Go. And then the DeepMind team, using similar ideas, created grids that began to win in various computer and not only computer games. Think about it.
Well, if only on their own field, the computer and would lose. Shoo choo :-) In fact, Alpha has learned to play the game by playing with itself, and here's the idea from the shoulder of the bargain. In our case, too, there is competition between buyers or sellers. Now let's introduce conditions: The signals should alternate buy-sell-buy-sell. But set the competition between purchases and sales to maximize profits. As a result, the arrows will try to increase their profits by tending to the extremes. In other words, we tell her. Yes, you set your trades any way you want, but stick to these recommendations.
In the process of optimization, the arrow will move back to the extremum improving its position may not reach it for a couple of bars but it will not lose its relevance, while ZZ puts the arrow at the extremum itself and tries to attract the optimization algorithm there where there is no generalization. That is, when the arrow reaches an extremum, it is key; when it is at the extremum itself, it is not. This is the reason why it is better not to use ZZ as a target