Random Flow Theory and FOREX - page 64

 
begemot61 >> :
Well to put it simply, you have Gaussian distribution (in case of other distributions it can work too, the main thing is that it is known and stationary). [...] Wait for the price to bounce from the average "far enough" and open a trade in the direction of the average.

A stationary distribution with thick tails can play a cruel trick on such a strategy, as the notion of "far enough" in this case is extremely vague or non-existent (the second point is, say, infinite).

 
begemot61 писал(а) >>
To put it simply, you have Gaussian distribution (in case of other distributions it may work too, as long as it is known and stationary). You have a bell, which shows that the price is often close to the average. You wait for the price to bounce "far enough" from the average and open a trade in the direction of the average. The price will always return to the area "close to the average". What is "far enough" and "close to the average" can be determined from the distribution.

That's exactly the kind of simple scheme used by those who are dismissive of mathematics. It is a zero expectation game. Therefore in the limit of large numbers the result of such a strategy will be zero. And bearing in mind the presence of the spread in reality it will be very negative. If you do not trust this statement you can check on your real account.

 
AlexEro >> :

"Known" by whom? Or maybe it is "Known - one of the top five dozen ALL known theoretical distributions, all of which are easily reduced/derived to a normal distribution"?

Keep in mind that 97% of the formulas of probability theory and statistics refer to a normal distribution of a random variable. If its distribution differs in any way from normal or the "standard dozen", the formulas simply do not work. So there are immediately a bunch of problems that ROBAST probability theory deals with, i.e. little dependence on non-normality of the distribution. But before applying the robust one (there isn't much of it yet) you need to have a distribution function at hand and I would ask you to clarify - how do you or anyone else know the distribution of our price series and whether there is such a thing as a "probability distribution" for a price series at all? How and at what interval are you going to calculate it?

Please don't be clever and read what the comment refers to.

I have nowhere claimed that the process is stationary and with a known distribution.

I just said that if it is so, it's easy to see how you can make money on it.

 
People, has anyone managed to get the stratora library up and running? What am I doing wrong, can you tell me? Should I just put Probability.dll in the libraries folder and Probability.mqh in the include folder? Or something else?
 

Yes, Probability.dll in the libraries folder. You should also write something like:

#import "TrueRandom.dll"
   int TrueRandom();
#import

This is how I handled the other library.

 
Where should it be written?
 
begemot61 >> :

Please don't get clever and read what the comment was referring to.

I never claimed that the process is stationary and with a known distribution.

All I said was that if it is, it is easy to imagine how to make money from it.

Who said that? Who said, or maybe even proved, that with a known empirically-calculated distribution - one can predict the behaviour of a random variable over time? (That is, assuming for a second that the value is random)? Who was it?

hint:

1).If the distribution is a Russian letter L - you won't get anything out of it, the value will bounce between two or three clouds.


2). the LTCM company went bankrupt (3 billion of their own, setting up other banks for 100 billion) precisely because they (the two Nobel laureates) believed that

a). price fluctuations in the masses are random ;

b). the distribution of price fluctuations is always normal and even if slightly non-normal, there are "no thick tails" in the distribution of random variables.

 
Mathemat >> :

A stationary distribution with thick tails can play a cruel trick on such a strategy, as the notion of "far enough" in this case is extremely vague or non-existent (the second point is, say, infinite).

Well that's obvious. And it doesn't have to have 'fat tails'. It can be two-humped, etc. etc. You can give a sea of examples where something primitive will not work, at least not.

It's just that before asserting something, you need to find out the properties. And we (at least me) do not know them.

So the answer was purely hypothetical, as well as the statement "It is impossible to earn on this process - it is a mathematical result", which is simply nonsense and the result of incorrect formulation of the problem.


By the way, how about this definition of a trend:

A trend is the difference between a real price series and a stationary distribution.

 
begemot61 >> :

By the way, how about this definition of a trend:

A trend is the difference between a real price series and a near stationary distribution.

You have to define which stationary process we are talking about. But to be honest, I don't like this definition yet.

 
begemot61 писал(а)