Machine learning in trading: theory, models, practice and algo-trading - page 2036
![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
The question is fair, because this approach belongs to the methods of agent-based modeling and it is necessary to describe the rules of individual behavior of agents. The rules of the "minority game" describe only the "reward" received by the agent from the environment.
Scientific articles on this topic usually do not ask the question "how to create a profitable system?") It rather sounds like "how exactly do these idiot agents create crises in the market?") Therefore, what they call "trading strategies" looks rather cheesy.
If to the question of what is the TS seriously and try to combine the trader's and scientific approach, the formalization of this concept escapes me. The obvious definition through the concept of algorithm suggests itself. But if you look closely at the whole life cycle of the TS, there appear ideas like "algorithm of the TS algorithm tuning" and so on ad infinitum).
I am slowly reading the articles on the subject and I come to the conclusion that we will get to the models with complex properties of the simplest model elements - traders)
Atoms are taken apart piece by piece)
Let's consider the situation with the search for information on the basis of reliability and usefulness in the future,
- how long is the life cycle of the TS ? (the trader's folklore about the testing in 10 years and 10 nights, "sucked" out of the trader's finger, who turns everybody's knuckles - we should not take it into account)
- what are the optimization/reconfiguration tasks ?
TS is living on a stationary series, stationarity of series is ability of one mathematical model to describe this series with acceptable error, when the error is more than acceptable, it is considered that another model can describe this series and lag occurs. And I agree with Alexei Nikolaev, the strategy is the change / formation of new behavior from the analysis of new data.
I would still like to put emphasis on the discrepancy between the concept of algorithm and TS. Rather, the TS is not some fixed code, but the process of changing it) "Put the Expert Advisor on the chart and forget about it" - this, in my opinion, is an unattainable ideal)
the method of exceptions or negation does not allow to formalize the search
Then we again come to the beginning of the discussion, what is the abstract TS? Or what should a researcher look for?
SZZ: with algorithms it is easier, the concept has long been formalized, or rather for the concept of program it is formalized: the program is data and the algorithm that processes this data, i.e. data + algorithm = program. I would like to hear the same minimalistic definition for the TC
TS is alive on a stationary series, stationarity of a series is ability of one mathematical model to describe this series with acceptable error, if the error is more acceptable, it is considered that the series can be described by other model, and there is a lag. And I agree with Alexei Nikolaev, the strategy is a change / formation of new behavior from the analysis of new data.
But I'm trying to find a general definition for the TS - you can look for a lot of things, at least beautiful balance charts in the tester are searched for by everybody and even many people find them, but as shown by observations of the signals service and active forum participants - this is not what we should look for when searching for (evaluating) the TS
Then we're back to the beginning of the conversation, what is an abstract TS? Or what should a researcher look for?
TS is when out of a lot of information and variants of its processing you compress everything to a single binary YES/NO decision.
TS is a filter, there are different filters, adaptive, intelligent, primitive...
The method of exceptions or negation does not allow us to formalize the search
Then we are back to the beginning of the discussion: what is an abstract TS? or what should a researcher look for?
SZZ: with algorithms it is easier, the concept has long been formalized, or rather for the concept of program it is formalized: the program is data and an algorithm that processes this data, i.e. data + algorithm = program. It would be desirable to hear the same minimalistic definition for TC.
Let me put it this way - TS is an algorithm, some part of which is always in the trader's head. It can be just an intuitive understanding of when to switch the Expert Advisor to virtual trading or re-optimize the parameters (without waiting for the drawdown). In principle, this can be a quite rational decision based on the news, but not necessarily.
In general, a "piece of the TS algorithm" is always left in the trader's head, which indicates the impossibility to formalize it completely. If we try to formalize it, it will result in endless recursion - the algorithm that changes the algorithm, that changes the algorithm... etc.
I agree with Valery that the reason is non-stationarity. It cannot be removed by detrending, switching to increments and other similar methods from econometrics.
I don't know, let's do the first one.
In the sample there all the posts are involved in the training, is the last one targeted?
let's get to models with complex properties of the simplest model elements - traders)
Still, I hope that "the transistor will not be able to measure the hero's wide heart")
I would not like to appear on History's garbage heap.)
TS is when you compress all information and variants of its processing to a single binary decision YES/NO
TS is a filter, there are different filters, adaptive, intelligent, primitive ...
You're writing about testing strategies, you need to go back to the previous step - what did you want to find?
Let me put it this way - TS is an algorithm, some part of which is always in the trader's head. It can be just an intuitive understanding of when it is necessary to switch the Expert Advisor to virtual trading or re-optimize the parameters (without waiting for the drawdown). In principle, this can be a quite rational decision based on the news, but not necessarily.
In general, a "piece of the TS algorithm" is always left in the trader's head, which indicates the impossibility to formalize it completely. If we try to formalize it, it will result in endless recursion - the algorithm that changes the algorithm, that changes the algorithm... etc.
I agree with Valery that the reason is non-stationarity. It cannot be removed by detrending, switching to increments and other similar methods from econometrics.
So we can't formalize the concept of TC?
It turns out that TC is inspiration? Or playing a musical instrument?
Or let's return to our ... - It turns out that TS is first of all the analysis of market information and decision-making.
In the sample there all columns are involved in the training, the last column is the target?
The last column is the target, feed the rest to the input
Writing a neural network from scratch just to see how it learns is dubious fun). When you can test it on ready-made ones and not to suffer from nonsense
Also you have to write parallelization, a quality optimizer, GPU support and make it scalable. All this for the sake of understanding that NS don't work at Forex
And then you have to state that NS takes 24 hours to learn (like in recent articles) and that no research can be conducted on such an architecture (due to the peculiarities of mql or God knows what else).
Why such a global pessimism? ))) I "looked" how they were trained before all modern packages in NeuroShell Day Pro. I already got robust results that I didn't know how it worked inside and it was almost impossible to tighten to MT4.
I agree that it would be desirable to bolt on the GPU.
The question is what kind of NS they are and what paradigm they have been built/learned in, mine are evolving.
Yes, the first robust version can be trained even for 24 hours (though in practice it takes 8 hours on an ancient home laptop). But to return to the necessity of further evolution of the first variant at the expense of its robustness will be necessary in a month. I.e. even at ten working tools in real life in advance there will be a new variant.
Now about the architecture, the NEAT algorithm is taken as the basis, supplemented with its own features. At the output the architecture will evolve, including the architecture.
So it goes like this.
At the same time I recommend you to read books/lectures on microbiology etc.
In arguments unfortunately one is a fool (arguing without knowledge), the other is a bastard (arguing with knowledge), I prefer an exchange of opinions with arguments/reasoning.
The main thing is to have an impact, to hell with it, let's go)).