Probability assessment is purely mathematical - page 11
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
This is the reason why I don't like forex. I don't know about others, but I don't understand currency pricing at all. And then there are two... What fundamental factors add up to the price ? In equities everything is precise and clear, you buy a piece of equity which has a real value. With this calculation, I understand firstly, that I am buying part of the company and secondly, I can understand what price I am paying, high or low. In some (very rare) cases I can say with a guarantee that I paid less for the company than it's worth and even if it goes bankrupt tomorrow I will get a profit. And if it works and brings profit, it will bring me profit. It is profitable for everybody.
With forex it is not clear. We chase prices, but we only understand them when we understand what we are paying for. Some people think that there are no prices on forex, yes there are... Just presented as a ratio of these prices. What am I paying for (changing my currency) and what do I own after payment and who benefits from it anyway?
I think currency trading is only needed in times of repression. When you save your dough in the currency of another, stronger country.
Inflation is thought to determine the price of a currency, but that mostly determines the rate of depreciation of the currency. A currency with a maturity expectation < 0. Of course there is positive inflation, but not with us and not in many other countries. Hence forex trading itself is not sensible.
Have you ever considered that currency is the same as shares, only the word firm has been replaced by the word state...
https://www.mql5.com/ru/code/8295 yes it is possible and anyone who thinks can download this indicator - install it and see that there is a pattern in forex
I installed it and did not understand anything. My indicator does not look like the one you see in the picture attached. But I don't understand your picture either. For some reason the ACF is exceptionally monotonous. How can it be? To my mind, ACF shows the correlation (connection) between "0" and "1" bar, between "0" and, etc. Why should this relationship decrease monotonically and smoothly?
I do not want to look for a probable error in your indicator and for me it is prismatic for the following reasons.
You should take a ready-made package that makes statistical calculations and use it. For ACF, for example, STATISTICA. This particular package has been around for over 20 years and hundreds of thousands or millions of users before us have agreed on the formulas and fished out all the mistakes the developers have made. It is methodologically correct to use the results of other people's work.
Hiding the inner workings of the package allows you to concentrate on preparing the raw data and interpreting the results. At least the package automatically calculates confidence intervals, that your indicator doesn't and it's not clear if you can trust the result you get.
I have that package somewhere and I will calculate the ACF and post it. As far as I remember (I may be wrong) the ACF from the packet has a completely different look, giving rise to various speculations.
Can we not have a reference. ARPSS has a different opinion: the autocorrelation can be used to judge the model of the series.
I'm too lazy to look it up. I'll just repeat the principle of proof.
1. Choose a timeframe, for example M15. Plot on a quite long history interval (say 10 000-20000 bars) the distribution of frequency of price increments (number of times depending on points). We obtain (approximately, but due to a large amount of data it is a good approximation) the probability density function. (I'm pretty sure it is exponential, but for this problem the type of distribution is unimportant).
2. We make the very realistic assumption that if we take a segment of history shifted 1 bar to the left (or right) of what we took in item 1, the probability distribution will change very little.
3. we measure in the same way and on the same period of history probability density function for the increment in price for 2 bars.
4. Further is a contrary proof. Suppose that the neighboring increments are independent. Since the price increment for two bars is the algebraic sum of increments of the first and the second bar and the density of the distribution of increments on the neighboring bars is the same (see step 2), according to the known rule, the probability density of the sum must be a simple convolution of the density of each of the summed values. Performing the convolution and comparing it with the obtained distribution at step 3, we make sure that they are not even close to each other (you can see everything with a naked eye there, you do not even need to check it). Having arrived at a contradiction, we conclude that our assumption about the independence of neighboring increments is incorrect.
That's it, it's pretty rigorous and without "scientific fiction". This method is suitable to check for independence of increments of any series. In addition I would like to note that if increments of a series are distributed by the exponential law (it seems to be so for the price) which is preserved at higher timeframes (which is probably true), the mentioned proof can easily be theoretically obtained due to the calculation of the appropriate convolution integral. However, it has long been known in probability theory that the exponential distribution is not stable.
I took it not out of my head before I posted it to the code. I tested it long and hard with readings of well-known and tested ACF calculation algorithms. It matched me completely, checked up to 16 decimal places (maybe more, I don't remember exactly now, but there was no difference with built-in function in MathCad).
And about
Почему эта связь должна монотонно и гладко убывать?
I took it not out of my head before I posted it to the code. I tested it long and hard with readings of well-known and tested ACF calculation algorithms. It matched me completely, checked up to 16 decimal places (maybe more, I don't remember exactly now, but there was no difference with built-in function in MathCad).
And about
The ACF has a whit delta function, because there is no relation between the data, they are random. But for forex, if ACF is built correctly, there is a relationship, the data are correlated and the nature (type) of ACF may help determine the type of the process. It is not always the same as in the example I posted. The selected section shows that at the moment the movement corresponds to the oscillatory chain of the 2nd order.Judging by the indicator, you do not differentiate the price series before calculating the ACF. Therefore, there is no sense in comparing it with the ACF. But it makes sense to apply the indicator to the CGS integral.
p.s. I think it is impossible to draw conclusions about dependencies from this indicator (or you need serious substantiations)
Judging by the indicator, you do not differentiate the price series before calculating the ACF. Therefore, there is no sense in comparing it with the ACF. But it makes sense to apply the indicator to the CMP integral.
p.s. It is impossible to draw conclusions about the presence of dependencies from this indicator's readings (or serious substantiation is needed).
Judging by the indicator readings, you are not differentiating the price series before calculating the ACF. Therefore, there is no sense in comparing it with the ACF of BGS. But it makes sense to apply the indicator to the GBS integral.
p.s. I don't think you can draw conclusions about dependencies from this indicator (or you need strong justification).
First justify why you need to apply differentiation. simple example. A car moves at a certain speed and by constructing the ACF of speed we will see that it has speed (correlated), in simple words "the trend is more likely to continue...". By applying differentiation you will no longer be investigating velocity, but acceleration - which in turn can be random.
P.S. To conclude that speed is random because acceleration is random is wrong in principle. We can be moving at a constant speed (trending upwards) and the acceleration will be BGS...
I agree about the presence/absence of dependencies. But I would argue about differentiation: each differentiation operation nullifies one order of dependence, if represented polynomially. So even if we get that there is no dependence in the differentiated series, it does not mean that there was none in the original series.
5 points. I'd give it a 10, but that's what spoils the score :-) "about the presence/absence of dependencies I agree..."