From theory to practice - page 263
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 one of the main questions. I'm not ready to comment on it now, a lot will depend on the data you give, Alexander, I think 300,000 won't be enough, I want more! Is it realistic? I've been thinking about it for a long time, I don't know, dr. I don't know, Dr. Trader, can you help me? As Alexander will provide the data, you need to mix up this sample in such a way as to clearly pull out the exponent, and all that remains is to identify what kind of distribution it is. I doubt it's logarithmic. But let's see...
OK. I'll post the bases for the 12 pairs with time stamps and intensities tomorrow.
Good. Tomorrow I will post the bases for the 12 pairs with time stamps and intensities.
Thank you very much! Thanks from everyone, not just me!)))
I would just like to understand the purpose of your research.
Here look, the course of my reasoning:
1. it is known that all mathematics of equations of diffusion is constructed for processes without consequence, i.e. for Markovian processes. These are ordinary differential equations, NOT integro-differential.
2. I WANT a stream of events (quotes) without aftereffects, i.e. with exponential time intervals between events.
3. I look at real time intervals between tick quotes. They are NOT exponential, but logarithmic.
4. I forcibly read tick quotes through the exponent defined by a discrete geometric distribution generator.
5. I obtain time series both with real quotes, i.e. when at time T1 a real quote with a time stamp has come, and with pseudo-quotes, i.e. when at a given time there was no tick and the previous tick is taken as the accepted value.
6. I have an amazing time series with exponential time intervals between tick data, with real and pseudo-quotes, i.e. I know the trading intensity in the moving observation window.
7. calmly apply drift and diffusion coefficients to solve the problem.
What do you want to achieve? A secret?
Why is there no 3rd mod - European???
What does bimodality have to do with sessions?
You measure trading intensity against the time scale and you have bimodality against the price scale.
Bimodality suggests that there is no transient between the changes of directions.
Or maybe there is, but it's hidden in the reversion. It means that increments actually tend to the mean of their modes, but neighboring increments may be of different modes (anti-seriality).
Try to build a histogram of seriality distributions (all gradients in one direction in a row, combined into one gradient).
By the way, you can also build a histogram of Z-bills, but increments there are not in primes, but in units, was a tick up then 1 up tick down 1 down. Then, counting how many ticks up, you will get the series length. A series can consist of a single tick or of any number of ticks. The Z-count also takes into account series of coming back, when each next tick is a U-turn from the previous tick. It will give us some insight into the processes. Maybe then we will have some ideas how to build TS.
I would just like to understand the purpose of your research.
Here look, the course of my reasoning:
1. it is known that all mathematics of equations of diffusion is constructed for processes without consequence, i.e. for Markovian processes. These are ordinary differential equations, NOT integro-differential.
2. I WANT a stream of events (quotes) without aftereffects, i.e. with exponential time intervals between events.
3. I look at real time intervals between tick quotes. They are NOT exponential, but logarithmic.
4. I forcibly read tick quotes through the exponent defined by a discrete geometric distribution generator.
5. I obtain time series both with real quotes, i.e. when at time T1 a real quote with a time stamp has come, and with pseudo-quotes, i.e. when at a given time there was no tick and the previous tick is taken as the accepted value.
6. I have an amazing time series with exponential time intervals between tick data, with real and pseudo-quotes, i.e. I know the trading intensity in the moving observation window.
7. calmly apply drift and diffusion coefficients to solve the problem.
What do you want to achieve? A secret?
Are you worried?)) Do you know how scared I am!))) I'm walking on thin ice right now.
You're going to crush me with points like this))) It's like loading a man walking on fresh ice with a bunch of bags on his back.)))
OK, let's go paragraph by paragraph, in principle, these are not questions, these are statements, OK:
1. It is known that all mathematics of diffusion equations is constructed for processes without consequence, i.e. for Markovian processes. These are ordinary differential equations, NOT integro-differential.
-----No big deal.
2. I WANT an event stream (quotes) without consequence, i.e. with exponential time intervals between events.
---- You have it.
3. I look at real time intervals between tick quotes. They are NOT exponential, they are logarithmic.
----Exponential and something else.
4. Forced reading of tick quotes through an exponent set by a discrete geometric distribution generator.
----If they are exponential in basis, then it is redundant.
5. I get a time series both with real quotes, i.e. when at time T1 a real quote with a time stamp has come, and with pseudo-quotes, i.e. when at a given time there was no tick and the previous tick is taken as the accepted value.
---- I agree.
6. I have an awesome time series, with exponential time intervals between tick data, with real and pseudo-quotes, i.e. I know the trading intensity in the sliding observation window.
-----I agree." with exponential time intervals between tick data"-you already have that.
Alexander, you already have a very good algorithm, you just need to do some more research, as you say, so that it's not 80% of successful trades, but 100%).
Maybe I'm wrong, I'm repeating myself, I'm walking on thin ice.
Look, we are just hung up on these exponents! Dr.Trader, this analysis has done your own data, and it has a direct bearing on tick lags. Thanks a lot, Dr. Trader!
https://www.mql5.com/ru/forum/221552/page261#comment_6878470
https://www.mql5.com/ru/forum/221552/page262#comment_6880004
What a forest of exponents,
https://www.mql5.com/ru/forum/228822/page4#comment_6773278
This is for those who have been doing series quantization on thresholds. Exponents!
Somewhere on some forum I came across a pairwise distribution. So there's a two-way exponent. Laplace distribution. Just like here, Laplace, that dr. Trader did. If anyone can find the link, please paste it in for an example.
And you say no, there's only the logarithmic.
PS. Alexander, please don't be offended by me, I will still be useful! ))))
Good. I'll post the bases for the 12 pairs with time stamps and intensities tomorrow.
By the way, ASK has to be memorised too, right?
Eh
Saving up ticks again. That chip turned out to be without ascs.
Here's the one with asks //corrected the code
Did I understand correctly that there is already an exponent there, but you just have to "see" it and work at that frequency, and discard the other ticks that don't fit into that exponent?
Have you ever read about non-gentropy? Can you give your opinion - is it a promising direction? Logically, yes. We will really see the measure of non-randomness of the process, i.e. whether we are in the trend or not.
Alas, I've never dealt with her. There's only hope for you!
Did I understand correctly that there is already an exponent there, but you just have to "see" it and work exactly at this frequency, and discard the other ticks that don't fit into this exponent?
I agree with the first part, but I am not ready to give an answer to the second part.
6. I have an awesome time series, with exponential time intervals between tick data, with real and pseudo-quotes, i.e. I know the trading intensity in the sliding window of observation.
Alexander, I must have misunderstood one thing from previous readings - is the sliding window a constant size? That is, you select a window of 20 units and move it in your calculations deep into the data? Or does the size change?
Somewhere you wrote that you saw a constant (constant diffusion value or so you put it) in the market when you applied the sliding window. I assumed at the time that your window size was variable!
Also, a few pages ago someone asked you a question how do you check that the conversion gives a "no memory" flow? You didn't answer? At least I didn't find an answer.