From theory to practice - page 50
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
So calculate the autocorrelation of the increments. But this is unlikely to help the case, because dependencies of this kind are weak and impermanent.
Looking for patterns in prices is much more productive.
Suppose I tell you that the average volatility at a given time of day changes only slightly from one day to the next. Would that work as a "market memory"? And there are only two points in the prices, high and low, taken once a day. Tics are not even close, with their increments).
Wait a minute. Where has this been proven? ))) To prove no memory, you have to rule out the presence of ALL possible (i.e. any) patterns. I don't recall where this has been done)
And again - a distribution of any kind contains no information about memory. Why have you suddenly decided the opposite?
And yes, I can easily build you an earning system on ticks with any readout method, even exponential, even completely random.
Yes, yes, of course...
Only in the evening - I need to get away from such a knockout... After all, I've already run the program today on a demo account, and it turns out all wrong there!
The reason - I still can't understand - how exactly the tick data should be taken, so as not to destroy non-marking, to be able to use historical archives...
Well done, Alexander! We are delighted!) It's finally come to an understanding that analysis and modelling should start with data preparation. So that the baby's water isn't wrong).
And memory is in the process - I won't vouch for all the tools, of course - maybe there are some that wander around in the SB all the time)
Well done, Alexander! We are delighted!)) It has finally come to an understanding that analysis and modelling needs to start with data preparation. So that the baby isn't wrong with the water).
And the memory is in the process - of course I will not vouch for all tools - may be there are some that wander in the SB all the time)
Exactly, Dimitri! I understand that you need to take the data not just any way, but the RIGHT way. But how? I don't know what to do with it yet.
If I read every tick - each brokerage company has its own data flow, it means that the trading robot will work differently with different brokerage companies. We need a universal mechanism. What is it? I don't understand...
When reading data in that sequence, which I applied, I got geometric distribution of probability of increments, actually, that was pointed out to me by respected Vladimir. This is the proof of the lack of memory of the process under given conditions.
I'm sorry, but this is the reasoning of a first grader. It's as if you haven't studied the subject matter. Memory is the dependence of something in the future on something in the past. A distribution (any distribution) contains no dependence and proves or disproves nothing. It's like "fly agarics are red, so you can't eat red strawberries."
A basic, elementary detection of memory is autocorrelation of adjacent increments. Learn the basics before you mess with the forum)
I can easily generate a series with any distribution (even exponential, even geometric, even uniform), which will be full of temporal dependencies, and on which you can earn, knowing them (within a sample, of course).
I'm sorry, but this is the reasoning of a first grader. It's as if you haven't studied the subject matter. Memory is the dependence of something in the future on something in the past. A distribution (any distribution) contains no dependence and proves or disproves nothing. It's like "fly agarics are red, so you can't eat red strawberries."
A basic, elementary detection of memory is autocorrelation of adjacent increments. Learn the basics before you mess with the forum)
I can easily generate you a series with any distribution (even exponential, even uniform), which will be full of temporal dependencies, and on which you can earn, knowing them (within the sample, of course).
Have you achieved anything here? Have you even written a single advisor? ))
You are wrong and you know it. Where there is a geometric distribution, there is no memory and there cannot be one. And no autocorrelation will help you. Study the math before you start arguing with me. That's how it is!
Please give me a link to any textbook with that phrase)
And I'm not arguing with you, I'm telling you how to shorten your ordeal)
By degrees of freedom I mean the classical definition:
https://ru.wikipedia.org/wiki/Степени_свободы_(physics)
It makes a very beautiful interconnected picture - if the current price is related to the previous price by a vector and the next price is related to the current price by the same vector, then we have the notorious 2 degrees of freedom which completely describe the system. What is 2 degrees of freedom in statistics is approximately the same and in the market there simply MUST be a t2-distribution. And I can't find it... How so? I don't get it...
IT DOESN'T MUST.
SO
this t2-distribution thing is a bit of an obsession...
DO NOT MUST.
SO
just some kind of obsession with this t2-distribution...
You are wrong and you know it yourself. Where there is a geometric distribution there is no memory and there cannot be. And no autocorrelation will help you. Study the math before you argue with me.