Probability assessment is purely mathematical - page 15
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
...My suspicions about your indicator ...
is not my indicator:)))
Confusion in many ways...
;)
for bars it depends on the number of ticks that are hit. I don't quite understand how it all averages out, etc., but there may be some errors because m15 bars have a steady intraday change in volatility (and consequently in increments). We should perform a more detailed analysis. Maybe it is not that simple.
Here is, for example, a similar study: we measure the average incremental length modulo m15 and h1 for example. For SB, according to Einstein's f-law the average body length h1 will be 2 times larger, in reality there are significant deviations for different periods as well. But again, here we need to analyse the increments which do not have a systematic difference in volatility - for example to average for each hour separately, or to take the timeframe of the day and above.
This may surprise you, but on equivolume bars, where the number of ticks in each bar is set in advance, the picture is even more definite.
Don't doubt about the method, it is proven and can detect dependencies even when correlation analysis does not work, i.e. when correlations are not visible behind the noise. The only disadvantage, however, is the need for a large sample - the larger the sample, the more reliable the conclusions.
it's not my indicator:)))
I was actually referring to Prival. I got it wrong.
I was actually referring to Prival. I got it wrong.
I don't understand how to general base. Reece, I argue that you can't get another from one period. What do you mean?
Frequency = 1/period. If you shortened the series by skipping every other observation, what happens to the frequencies?
And this statement requires your explanation -
Temperatures do not flow from Brownian motion, ticks do not flow from timeframes.
;)
From another thread, but since the figure with the different frequencies that surprised you is also present here, I'll repeat it.
I don't understand how to general base. Reece, I argue that you can't get another from one period. What do you mean?
Frequency = 1/period. If you shortened the series by skipping every other observation, what happens to the frequencies?
And this statement requires your explanation -
Temperatures do not flow from Brownian motion, ticks do not flow from timeframes.
;)
Let's go to the same thread. My answer is there.
A profound misconception. My suspicions about your indicator were intuitive and wrong. The indicator is most likely correct, but the use of it is methodologically incorrect.
What does your indicator (ACF BP) show? That there are dependencies in BP. Sorry, but this is a triviality. No one is denying the presence of trends and so it can be seen without any mathematics. Moreover, it is not correct to investigate the regular components of BP by methods of mathematical statistics. Your post has once again convinced me of the need to stick to software packages - this will avoid methodological errors. In our case we need to exclude the regular components - trend and cyclic component, if we want to see in BP what is not visible to the naked mathematical eye.
What do we want to see? We want to see the parameters of a model, by which we could not only analyse historical data, but also predict the future. This is what the ACF of differences, difference-in-differences, etc. are built for. For example, when identifying the ARPSS model we initially get two possible answers: the model can be identified and the model cannot be identified. Please agree that this result is already worthy of taking differences, and your arguments about loss of information have no basis, as we are excluding an established fact (t rand) from consideration, and trying to get information that is not initially visible at all.
1. You still haven't figured out what is being calculated there. If you can't see that the trend is subtracted there before the ACF is built ... I'm sorry.
2. ACF is just a formula, just like the well-known Mashka, and what to give it as input is up to the researcher. And to argue that if someone gives Mashkas input minutki, it is methodologically incorrect, more correct let us say daily ... makes no sense.
3. As for getting the model, I have it, good or bad, that's not the point. You can read the methodology at https://www.mql5.com/ru/forum/105740/page25#54080.
Candid 11.12.2007 23:21
...
That is, the preformation which reduces BP prices to white noise will be a market model. Now that's what I understand :)
The check is simple, just with ACF - if ACF is a delta function (or close to it), all components are found...
2. ... And to argue that if someone feeds minutiae to Mashka's input, it is methodologically wrong, more correct let's say diaries .... makes no sense.
Actually, I was talking about something else and I attached kotir spectra as a proof of my thought. Once again: each timeframe has its own statistics. The stats of different timeframes can complement each other, but you can't get one from the other.
3. As for getting the model, I have it, good or bad, that's a different matter. The methodology of model derivation can be found at https://www.mql5.com/ru/forum/105740/page25#54080.
Read, for example, ARPSS. Maybe your model is more successful for profit than ARPSS, but what you have described is not a model: I have not seen confidence intervals, p-levels, or any estimates of confidence in the results.
1. You still haven't figured out what is being calculated. If you don't see that it subtracts the trend before plotting the ACF ... I'm sorry.
Yes, you're right, I didn't understand your indicator, but I figured it out in essence. The way the trend was subtracted didn't lead to any interesting results. I promised to bring the results of the analysis in STATISTICS - here you go.
ACF. By ACF we can talk about a trend. Please note that this calculation is accompanied by very useful additional information.
ACF Close is the trend that is approximated by regression. The formula is below. I think this ACF calculation is entirely the same as yours. IMHO, subtracting the trend from the regression, didn't yield anything. The regression probably doesn't approximate the real trend well.
Here is a chart of Close without trend by regression. We can see that it does not look much like white noise. ACF is higher.
Let's take a look at the Close chart without ACF. It looks more like noise. And we are trying to get rid of the regular component.
Now the ACF is for that.
It seems to be more interesting.
I haven't worked with the statistics package in a long time. this package often has very unclear graphs.
1. In the link I gave, there is no model there, only a description of the methodology to build it. The model is in the same branch, but much further away. I know what ARPSS are, and I gave up on them a long time ago, and regression models too (I think you have the same conclusion).
2. Your graphs are strange. Can you cite the ACF of white noise to make sense of what you have drawn here and(or) the constants. By definition the ACF at point 0, equals 1 (for any data) and gradually decreases to 0 (when the bias is completely out of sample). I have not seen this in any of your graphs.
3. These are the phrases that baffle "...Let's look at the Close chart without ACF. It looks more like noise. But we are trying to get rid of the regular component..." You have a wrong understanding of ACF.
Any function has ACF, it is unclear how you build a graph without ACF, most likely it's some kind of unfortunate wording, you wanted to build something, you built it, but unfortunately you failed to explain it.
Also, by trend I mean the straight line equation y=a*x+b. I subtract it from Close before plotting ACF. And then I calculate the ACF - and its form clearly shows that in addition to the trend (straight equation) there is always an oscillatory process in the market.
I do not know how to explain.
1. We take Close.
2. Subtract the trend (y=a*x+b) from the Close.
3. Construct the ACF
4. we get a picture like this https://www.mql5.com/ru/code/8295 in red is what we get. This is an oscillatory chain of the 2nd order at a certain point in time. There is also ACF plot obtained from the model with correspondingly calculated coefficients (blue plot), it's almost a perfect match... But it's not always the same, today this ACF may look different, and consequently, another model is valid in the market...
Z.U. I'll answer right away to everyone who messaged me in person. I'm like, "How do you trade with ACF?" You can't trade with ACF (or rather, I have no idea how to do it). ACF allows you to visually see and calculate the parameters of the processes occurring at the moment in the market, but you need to understand what ACF is and what it is all about. Try to build ACF without a computer, a piece of paper and a pen and you will understand a lot...
Have you sorted out the 40 and 20 periods?
Or do you keep insisting that you should expect "miracles" with the original and shortened one by removing every other observation?
This reminds me of Krylov.
NEARDERTALEC and the "statistics" package...
;)