Grail indicators - page 7

 
yosuf:

Past (P) + present (N) + future (B) = the single process in question. And although there are functional distinctions in types of functions P(c), H(c) and B(c), where c is time, the normalization condition : P(c) + H(c) + B(c) = 1 is always fulfilled at any moment of time. Time boundaries of these stages depend on the considered time unit. If we consider millennia, the "present" time = 1000 years, strange as it may seem. If we consider years, "present"=1 year, etc.

In our case, "present" is events occurring during the current bar. It turns out that if we don't consider price behaviour before the current bar, we don't use historical data either.



Interesting thought ;) In other words, the future is represented through the present and the past by a very simple relationship: B(c) = 1 - H(c) - P(c)

And the fairness or unfairness of such a normalisation condition P(c) + H(c) + B(c) = 1 can easily be checked on the basis of the following considerations.

If

present == H(c),

then the past == P(in)=H(in-1),

and the future == B(c)=H(c+1).

From

P(in) + H(in) + B(in) = 1

we get

H(in-1) + H(in) + H (in+1) = 1

i.e.

H(in+1) = 1 - H(in) - H(in-1).

or alternatively

H(in) = 1 - H(in-1) - H(in-2).

We have the simplest recursion. Consider bars -- hour, day, year, millennium.

You should preliminarily fit the process into the range [-1;1] which is not difficult, and having done this preliminarily, you can check your statement about the H-H-B relation for any process.

But this check is unlikely to give a positive result ;)

 
avtomat:


Interesting thought ;) In other words, the future is represented through the present and the past by a very simple dependence: B(c) = 1 - H(c) - P(c)

And fairness or unfairness of such a normalization condition P(c) + H(c) + B(c) = 1 can be easily checked based on the following considerations.

If

the present == H(c),

then the past == P(in)=H(in-1),

and the future == B(c)=H(c+1).

From

P(in) + H(in) + B(in) = 1

we get

H(in-1) + H(in) + H (in+1) = 1

i.e.

H(in+1) = 1 - H(in) - H(in-1).

or alternatively

H(in) = 1 - H(in-1) - H(in-2).

We have the simplest recursion. Consider bars -- hour, day, year, millennium.

You should preliminarily fit the process into the range [-1;1], which is not difficult, and having done this preliminarily, you can check your statement about the H-H-B relation for any process.

But this check is unlikely to give a positive result ;)

Absolutely correct. Don't doubt, if you have thought up such check in the specified original way, because, found by me, the normalization condition from the mathematical point of view is absolutely faultless, although it was necessary to introduce in B(c) unknown before "two-parameter integral exponential" function E - "primogenitor" of all exponents, so that B(C)=1-E, and E itself, miraculously, decomposes on sum H(C)+P(C) at its integration in parts (look at paper). And the normalization condition is sounded as a single orchestra (!), so that "a mosquito can't sharpen its nose" (c).
 

As far as I understand the physics course. Time is the transition of a system to a new state with increasing entropy. Hence time can be described by the spatial position of elementary particles (how elementary is a question).Here again, space is discrete or continuous, so time will also be discrete or continuous. Everything would be fine if god didn't play dice. A random variable introduces its corrections. It turns out that the future is described by the laws of interaction, corrected by the probability fan, and the further from the moment now, the more unpredictable the result. For example, with a probability close to 1, I would call you a wonderful person. In 10-15 minutes the probability is clearly lower, in a year, who knows. Going back to our mutton, we must know the position of the system now, the position of elementary particles read traders, somehow model their behavior (let's assume that there is a model of behavior accurate enough to account for randomness). There is also the question of accuracy of representation of initial data. It may turn out like with the butterfly causing a hurricane and the trader with a deposit of quid causing the collapse of the financial system.

The mushrooms are very strong this year.

 
avtomat:


Why are filters suddenly a utopia?

However, the concept of "filter" is too broad and therefore rather vague. After all, the term "filter" is quite applicable to (18) as well.

My point is that an incipient useful signal can be roughly "cut off" by a filter. You need a clever filter, but once you've invented one, you don't need anything else. So, it's a vicious circle.
 
ivandurak:

As far as I understand the physics course. Time is the transition of a system to a new state with increasing entropy. Hence time can be described by the spatial position of elementary particles (how elementary is a question).Here again, space is discrete or continuous, so time will also be discrete or continuous. Everything would be fine if god didn't play dice. A random variable introduces its corrections. It turns out that the future is described by the laws of interaction, corrected by the probability fan, and the farther from the moment now, the more unpredictable the result. For example, with a probability close to 1, I would call you a wonderful person. In 10-15 minutes the probability is clearly lower, in a year, who knows. Going back to our mutton, we must know the position of the system now, the position of elementary particles read traders, somehow model their behavior (let's assume that there is a model of behavior accurate enough to account for randomness). There is also the question of accuracy of representation of initial data. It may happen like with the butterfly causing a hurricane and the trader with one dollar deposit causing the collapse of the financial system.

The mushrooms are very strong this year.

And to successfully describe the result of random actions of an ormada of traders, there is an excellent example of the solution of such a problem known as the gas laws, only in this case the "craziness" of gas molecules is leveled by relations connecting the volume, temperature and pressure. The analogue of temperature can be price, the volume of the market can be taken, as a first approximation, constant. But what is the "pressure" parameter? This is the reason for our troubles - it is impossible to describe the pricing process by the parameter "price" alone! One parameter is missing - the analogue of pressure. Think about it, gentlemen. We need a parameter that can be estimated unambiguously at any given time. Maybe the total number of buy and sell contracts announced, at all prices, will do?
 
yosuf:
Quite right. Don't doubt, if you have conceived such check in specified original way, because, found by me, condition of normalization from the mathematical point of view is absolutely faultless, though it was necessary to introduce in B(c) unknown earlier "two-parameter integral exponential" function E - "primogenitor" of all exponents, so that B(C)=1-E, and E itself, miraculously, decomposes on sum H(C)+P(C) at its integration in parts (look at paper). And the normalization condition is sounded as a single orchestra (!), so that "a mosquito can't sharpen its nose" (c).



.

Drive the process into the range [-1;1]

Here the comments show a clear inconsistency.

.

And here's what it looks like on the story:


Y[j]=1 - X[j+1] - X[j+2];

.

This looks more like a conjugate process relative to the original process.

But you must agree that this is far from what is stated.

 
avtomat:


.

Drive the process into the range [-1;1]

Here the comments show a clear inconsistency.

.

And here's what it looks like on the story:

Y[j]=1 - X[j+1] - X[j+2];

.

This looks more like a conjugate process relative to the original process.

But, agree, this is far from what is stated.

A conversion error has crept in unnoticed. The statements are wrong:

then past == P(c)=H(c-1),

and future == B(c)=H(c+1).

P(c) and B(c) are integral functions, while H(c) is a differential function and cannot be equated in this way.

B(c) = 1- E

E = Integral(from 0 to t) (t/τ)^(n-1)/G(n)*exp(-t/τ)dt - introduced, by me, function, so that E=H(in)+P(in) .

H(c)= (t/τ)^n/G(n+1)*exp(-t/τ)

P(B) =Integral (from 0 to t)(t/τ)^(n)/G(n+1)*exp(-t/τ)dt

G(n+1) =Integral(0 to infinite) x^n*exp(-x)dx -Hamma Euler function

G(n+1) = 1*2*3*....*n = n! -for integer values of n;

The sign of the integral is not shown, I think you will see.

 
yosuf:

An unnoticed error in the transformations has crept in. The statements are wrong:

then past == P(c)=H(c-1),

and the future == B(c)=H(c+1).

P(c) and B(c) are integral functions, while H(c) is a differential function and they cannot be equated in this way.



OK, let's correct. Give the formulas for P(c) and B(c).
 
avtomat:

OK, let's correct. Give the formulas for P(c) and B(c).


B(c)=f(P(c),H(c))

f-? :) These formulas are of no use. You have to study the processes - their internal times and phases. In the market it is complicated by the fact, that there are a lot of processes and their price is resulting, processes are not periodic (period in astronomical time is not a constant) and they change.) It remains to consider only a part of the processes and expect that they will not disappear quickly.

 
yosuf:

1. I agree, but we should still try to understand natural processes.

2. I do not agree, the process is in progress. It's the second month of stalemate: initial deposit of 2K is rotating around its axis with the amplitude of 1.7 - 2.4K. The market cannot overpower the algorithm, just as the algorithm has no noticeable advantage over the market, despite the huge number of transactions (2 orders are continuously set every 15 minutes with 0.1 lot). At the moment, equity = 2.109K.



1. You can safely use something without understanding it.

2. The market is not overpowered by any algorithm, leave this fatal business and invest in pams, they use a simple idea - the inertia of the process.

And no huge number of transactions - a maximum of one per day.