Machine learning in trading: theory, models, practice and algo-trading - page 220
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
(I think the man just complained a little)).
The point is that you don't need to take all sorts of "sweets" in the form of R, Matlab, python and so on. Anyway, then everything will be remade on C++, and this process is almost independent, R code is not a prototype for C++. People get hooked on all sorts of pops RAD and then bite their elbows, to get rid of bad habits after 2-3 years, if course one realizes that he got in a deadlock, often he doesn't.
In this case, the question is not whether there are any regularities in the market; they exist by the very nature of the market.
I guess there are inefficiencies that are either liquidity-limited or time-limited
What is the difference between inefficiency and regularity? Because it appears randomly in time and just as randomly disappears as soon as it exhausts its liquidity
well, in my opinion, inefficiency is more short-lived.
Regularity is something from the category of ten-year market cycles, there's a downturn-there's a growth. or seasonal patterns, and the rest is inefficiency
The point is that you don't need to get your head in the form of R, Matlab, python and so on. All the same then everything will be converted to C++, and this process is almost independent, the code on R is not a prototype for C++. People get addicted to various popular RAD and then bite their elbows, it takes them 2-3 years to get rid of bad habits, if a person realizes that he got in a dead end, of course, often he doesn't.
There is exactly one line about R in the article:
Sometimes, though seldom, I want to pull in Matlab or R ;-) There is one piece of advice - go straight to the doctor.
I must add that this phrase refers to deep neuronics, not to all R. The problem is that R is an interpreter, so the speed of mathematical calculations is a bit slower than in compiled programming languages like C++ or Java.
For high performance in R, it is necessary to write a C++ library that interfaces to R, and this is done in most cases. Or, logically, you can just create a neuronc at once in C++, without R. Working with data in c++ is more uncomfortable than in R, but apparently this approach is quite acceptable.
In general, R in production is good for data processing, you don't need a doctor.
Creating a neuronics on pure R is still a mauvais ton, I agree with you there.
The point is that you don't need to get your head in the form of R, Matlab, python and so on. All the same then everything will be converted to C++, and this process is almost independent, the code on R is not a prototype for C++. People get hooked on all sorts of pops RAD and then bite their elbows, to get rid of bad habits after 2-3 years, if course one realizes that he got in a deadlock, but often not realizing.
99% of people on this forum is the researchers, those who do not have a good working strategy and they are looking for it, those who have it, they rather silently observe this branch ...
So if I as a researcher wanted to test a neural network in the market, what's better? spend half a year to figure it out by myself and write the network in C++ and understand that it does not work, or check this idea on "pops" R using already prepared solution and understand that it does not work too, but it takes less than 30 minutes, that's the plus of all "pops" R, I'm not talking about production, production is when you know exactly what you're doing, when you have ToR and so on, and when you're researching - this is a method of groping, you're just looking, and if for every study to write ready-made solutions, then you'll lack a life, I saved myself one life with R.)
Understand, I'm not advertising R and say that it's the best, you can write at least in pascal if you like, but in this situation and in this context, when you explore the market, R is the best solution at the moment. imho. When there will be a production then you can already think...
99% of people on this forum are researchers, those who do not have a good working strategy and they are looking for it, those who have it, they rather silently observe this branch...
So if I as a researcher wanted to test a neural network in the market, what's better? spend half a year to figure it out by myself and write the network in C++ and understand that it does not work, or check this idea on "pops" R using already prepared solution and understand that it does not work too, but it takes less than 30 minutes, that's the plus of all "pops" R, I'm not talking about production, production is when you know exactly what you're doing, when you have ToR and so on, and when you're researching - this is a method of groping, you're just looking, and if for every study to write ready-made solutions, then you'll lack a life, I saved myself one life with R.)
Understand, I'm not advertising R and say that it's the best, you can write at least in pascal if you like, but in this situation and in this context, when you explore the market, R is the best solution at the moment. imho. And when there will be a production then you can already think...
It depends on what kind of research, if researches of the student in order to pass the term paper, then yes, it is much easier to take a function, run it and also get a standard report with nice graphics, there are no questions.
If the study of engineer quant, in the financial office, where they take money from the market, not only due to commissions, or near-market technologies, the situation is different. As a rule, the right offices build their own trading infrastructure for 5 years, where there are 95% of all necessary tools, which are conveniently wrapped and you can use them not much harder than in R, everything is transparent and available for editing and dancing with diamonds. At this level the quantum thinks in the system of abstractions that its trading infrastructure gives it and the further this system of abstractions becomes unique and not reducible to common parts. Most trading ideas are variations of the structure of high level abstractions of the trading infrastructure, which in principle cannot be checked in R or in Matlab, since it is not a question of feeding MLP with stochastics with different windows.
Sometimes I can use R(Matlab, math) as a kind of advanced calculator when I need to draw some exotic function or something like that for students to try, but it's far from looking for trading strategies. But if you write code longer than 100 lines in R with the logic of strategies, etc., then you gradually develop a habit of thinking in terms of R , which then is very harmful, as programming with "cubes", etc. this is the danger. Standard problems are solved quickly, complex ones often reach a deadlock, because they cannot be approximated by the available tools or require much more complicated tambourine dancing than in pluses. The coder starts to think like a designer thinking he is programming and then he just variates parameters of ready tools and when you need to program it, it turns to be very unfriendly and inconvenient.
It depends on what kind of research, if it is a student's research in order to pass a term paper, then yes, it is much easier to take a function off the shelf, run it, and get a standard report in addition with beautiful graphs, there are no questions.
If the study of engineer quant, in the financial office, where they take money from the market, not only due to commissions, or near-market technologies, the situation is different. As a rule, the right offices build their own trading infrastructure for 5 years, where there are 95% of all necessary tools, which are conveniently wrapped and you can use them not much harder than in R, everything is transparent and available for editing and dancing with diamonds. At this level the quantum thinks in the system of abstractions that its trading infrastructure gives it and the further this system of abstractions becomes unique and not reducible to common parts. Most trading ideas are variations of the structure of high level abstractions of the trading infrastructure, which in principle cannot be checked in R or in Matlab, since it is not a question of feeding MLP with stochastics with different windows.
Sometimes I can use R(Matlab, math) as a kind of advanced calculator when I need to draw some exotic function or something like that for students to try, but it's far from looking for trading strategies. But if you write code longer than 100 lines in R with the logic of strategies, etc., then you gradually develop a habit of thinking in terms of R , which then strongly harms, as programming with "cubes", etc. this is the danger. Standard problems are solved quickly, complex ones often reach a deadlock, because they cannot be approximated by the available tools or require much more complicated tambourine dancing than in pluses. The coder starts to think like a designer thinking he is programming and just varying the parameters of ready-made tools, but when you need to program it, it is too inconvenient and unusual
This is not the first time on this forum that a dispute about "Which language is better?
Personally I remind every time that MT4/5 are TRADING tools, not programming tools.
What are you going to program in C or R? Who needs this process anyway? We need money, and it results from the block of decision making on positions. And there are a large number of ready-made tools in R for this purpose, and the problem is not in development of these tools, but in their application. There is no need to program anything in R - it has much more than one person can master, is it really impossible to understand such a simple idea?
And lastly on the topic that in C here someone will create a more efficient code.
Such statements can only be made by people who have no idea WHAT one should program. If they can understand the object of programming, they will find out that they will have to compete with Fortran and C libraries, matrix arithmetic bibbles, and all that in the mode of loading all the cores of the computer.
C supporters! Sit down, at least a little understand R, not with R, but with R packages, such as those listed in the caret shell, which includes up to 200 packages (a few thousand functions), relevant to trading. And then, hopefully, you will lose the desire to publicly demonstrate your ignorance.