Calculate the probability of reversal - page 10

New comment
 
secret:

It is:

https://en.wikipedia.org/wiki/P-P_plot

Take two different parabolas, for example. There is a linear relationship between them. Although both curves are non-linear.

In order to make the relation become linear, the coordinates must be transformed non-linearly. I will not talk about some data, but for the problem of estimating the probability of large deviations I enclose a book-sized article: Wentzel A.D. Rough limit theorems on large deviations for Markov random processes. Hardly read it myself, but it may help you. The 1976 edition.

 
Vladimir:

You have to transform the coordinates non-linearly to make the relationship linear.

Well, that's what the original question was about, a non-linear transformation of coordinates.
Perhaps I should have said "regression" rather than "approximation" to make it clearer.
It's just that by my peasant standards, in this case it's the same thing, since one of the quantities is a continuous line, not a set of points.
 
Serqey Nikitin:

As you all know ANY problem can be solved in SEVERAL DIFFERENT ways...

For example:

1. You can try to PREVENT a FUTURE trend reversal...

2. You may document a trend reversal in a CURRENT situation in the market...


As you understand, variant №1 is VERY difficult to solve with a high degree of reliability...

Option #2 is much easier, as you don't have to be a psychic like Vanga, and the positive results will be much higher than in the first option...


All in all: The RIGHT way of setting the problem gives more than half of its solution!

It's not about trends at all
 
Vladimir:

For the relation to become linear, the coordinates must be transformed non-linearly. I will not talk about some data, but for the problem of estimating the probability of large deviations I enclose a book-sized article: Wentzel A.D. Rough limit theorems on large deviations for Markov random processes. Almost haven't read it myself, but it may help you. The work is from 1974.

I think Elena Sergeevna was.

 
Guys, fat tails are just a confirmation of TV's unsuitability for the trade. They're very thick. Thicker than the top.
 
Maxim Romanov:
It's not about trends at all
A pivot is a pivot... Or are you calculating tick charts?...
 
Алексей Тарабанов:

I think Elena Sergeevna was.

Spouse

 
Алексей Тарабанов:
Guys, thick tails are just a confirmation of TV's unsuitability for trading. They're very thick. Thicker than the top.

and TV, too.

but I read about the Cauchy distribution yesterday.

There's a funny thing about probability densities.

and sincethe probability of entry is 0.5, therefore the probability density function should be equal too

In this case we are no longer interested in the shape of the distribution, and by and large the theorist has already done his part ;)

here's a drawing from a parallel thread

if you calculate the area of the triangles for buy and for sell, they are not equal

i.e. this kind of theanalysis is doomed to fail.

 
Vladimir:

Spouse

А. D. - son, spouse - D.A.

 
Vladimir:

For the relation to become linear, the coordinates must be transformed non-linearly. I will not talk about some data, but for the problem of estimating the probability of large deviations I enclose a book-sized article: Wentzel A.D. Rough limit theorems on large deviations for Markov random processes. Hardly read it myself, but it may help you. 1976 edition.

In this case the theory underlying the Kolmogorov-Smirnov test would be more accurate. The empirical distribution function is treated as a Poisson process and its deviation from the theoretical distribution function in the limit (when the sample size tends to infinity) will converge to a Brownian bridge. You can read it in Borovkov's textbook on matstat.

1...345678910111213
New comment
Reason: