How can I tell the difference between a FOREX chart and a PRNG? - page 19

 
AlexEro:

3. Let me say in more detail. You gave an autocorrelation formula for a normally distributed time series of random variables. The standard deviation is a good criterion for the mean only for a Gaussian distribution. In the general case of price series, the standard deviation is not only not the best criterion of optimality of the so-called expectation, but leads to the wrong one. That is why in trading the masks (MA) either work or do not work at all.


Before posting them, I carefully checked all calculations. I know three ways of calculating the ACF, all three are shown below in the screenshot and in the Matcadet file (attached). The results of the calculations are the same for all three methods. If you know a more correct calculation of ACF please share the formula. I posted only the third way of calculation, the one in headfirst form. And when I was porting the code I caught a bug in MQL and suggested a more perfect variant of linear regression calculationhttps://www.mql5.com/ru/forum/107017/page6

Files:
akf.zip  45 kb
 

When and if you KNOW EXACTLY that your random variable distribution is normal, these are autocorrelation methods. Only then do these formulas give a more or less reliable estimate of "autocorrelation", the statistical repeatability of a series. For a rough estimate (of the degree of repeatability of the series, or lack of repeatability in the residuals of the model when subtracting a series from it, that is, to check the validity of the model - as they do in ARIMA or something else) they can certainly be used (except all kinds of Fourier). But for highly variable systems these methods give a big error. But how big is this error and is the error acceptable for trading with 1:100 leverage and 1-2% volatility per day?

If the distribution of a random variable is unknown (price series), then one MUST apply other, more complex non-parametric (ranked, ranked) methods of calculating correlations (and auto-correlations).

https://ru.wikipedia.org/wiki/Корреляция

They are often used in the social sciences for "correlations", because it has long been known there that technical "mean-square" theorist methods just stupidly don't work there. There's even a special non-parametric statistics package called SPSS for these people.

https://ru.wikipedia.org/wiki/SPSS

Exactly the same should be done for auto-correlations.

http://www.hr-portal.ru/statistica/gl13/gl13.php

In statistics, the term non-parametric statistics has at least two different meanings:

  1. The first meaning of non-parametric covers techniques that do not rely on data belonging to any particular distribution. These include, among others:
    • Distribution free methods, which do not rely on assumptions that the data are drawn from a given probability distribution. As such it is the opposite of parametric statistics. It includes non-parametric statistical models, inference and statistical tests.
    • non-parametric statistics (in the sense of a statistic over data, which is defined to be a function on a sample that has no dependence on a parameter), whose interpretation does not depend on the population fitting any parametrized distributions. Statistics based on the ranks of observations are one example of such statistics and these play a central role in many non-parametric approaches.
  2. The second meaning of non-parametric covers techniques that do not assume that the structure of a model is fixed. Typically, the model grows in size to accommodate the complexity of the data. In these techniques, individual variables are typically assumed to belong to parametric distributions, and assumptions about the types of connections among variables are also made. These techniques include, among others:
    • non-parametric regression, which refers to modelling where the structure of the relationship between variables is treated non-parametrically, but where nevertheless there may be parametric assumptions about the distribution of model residuals.
    • non-parametric hierarchical Bayesian models, such as models based on the Dirichlet process, which allow the number of latent variables to grow as necessary to fit the data, but where individual variables still follow parametric distributions and even the process controlling the rate of growth of latent variables follows a parametric distribution.

https://en.wikipedia.org/wiki/Nonparametric

 
AlexEro:

They are often used in the social sciences for "correlations", because it has long been known there that technical "mean square" theorist methods simply don't work there. There is even a special package of non-parametric statistics for these people


Why do they need all this for trading?
 
Avals:
What's the point of all this in relation to trading?
They haven't seen each other for a long time and miss each other. And how can you not prove who is cooler in terminology?
 
AlexEro:

...

If the distribution of a random variable is unknown (price series), then other, more complex nonparametric (ranked, ranked) methods of computing correlations (and auto-correlations) MUST be applied.

https://ru.wikipedia.org/wiki/Корреляция

...

Professor! (at the last desk a student's hand timidly reaches out) How can correlation help you make money in the market? The correlation between the dollar index and the euro is -0.98. What should we do? Sell the euro? Buy the dollar index?
 

The distribution of the incremental series. One series is PRNG, the other is forex.

P.S. No"division, multiplication and other multiple GSCh.". Still the same dumb gpsh from excel.

 
Is the forex on the left? although it looks more like the forex on the right.
 
C-4:
Professor! (a student's hand timidly reaches out to the last desk) How can correlation help make money in the market? The correlation between the dollar index and the euro is -0.98. What should we do? Sell the euro? Buy the dollar index?

I don't have the slightest idea. I don't know how a calculated "correlation" with the illegal currency "euro" by an unknown person can "help make money in the market" in an unknown, unspecified trading system.

The science of statistics tests hypotheses.

 
AlexEro:

I don't have the slightest idea. I don't know how the "correlation" with the illegal currency "euro", calculated by nobody knows, can "help make money in the market" in an unknown, unspecified trading system.

The science of statistics tests hypotheses.

Professor, at least teach me how to deal only with "legal" currencies. How do you tell the difference between an illegal "Euro" type currency and an illegal one?
 
C-4:


How can correlation help make money in the market?



There is an article by Statistical Carry Trading on how to make money on positive swaps using correlations.

In theory, nothing complicated or abstruse. And even the screenshot to the article draws the answer to the question, "where does the money lie?

Another thing is that correlations can change sign to the exact opposite and then instead of earning you get a loss.

Simply put, solving one problem involves another problem: "how do I predict the sign of the correlation?