Random Flow Theory and FOREX - page 21
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
Prival, not much left! - to cram the market model into this filter:-)
We definitely can't do with one model here, it would be great if we could manage 3-4 models, to cover at least 30-40% of the time.
The only thing I think we need to do is to help programmers creating MQL5. Think through and offer them an algorithm for storing these bars + having two such archives of quotes, try to adequately reproduce the flow of ticks when the tester works.
P.S. I am a common person and I can always make a mistake. But only he who does nothing is not mistaken. I would be glad to hear your objections, or support. If this information is needed by many traders, I think MetaQuotes Software Corp. will certainly help.
I think it is a necessary and useful thing. But while the necessity can be estimated by the number of those interested (and there have always been plenty of those on this forum and the question of ticks has been hotly debated), the usefulness can hardly be estimated yet. If someone implemented it for their own purposes, he did not report the results. But even if those few who did it had no results at all, it does not mean that there cannot be any. Everyone's head, as we know, works differently.
However, I doubt that they will be able to "help programmers who are developing MQL5. The MQ's EA is quite calm, or even cool, considering the initiatives born in the MQ forum.
But with the storage algorithm, I think there's no problem. One can easily store ticks in a single file of a simplified time-price structure, and dynamically construct bars of a given equal volume in each tick chart autonomously. The number of ticks in the bar of such a chart should be just one of its properties, changeable in the same way as the colours or appearance of candlesticks change now.
But I don't understand why people completely uninterested consider it their duty to put a fat cross on tics. In public.
I don't know why you don't understand. There's nothing to understand. I'm just trying to bring to the attention of beginner traders some details of life.
These are the volumes of one brokerage house. See how they change over time. If you're interested you can compare them with the ones in your computer. And try to explain how you will build a grail that works for this particular brokerage company and at the same time, for other quotes and volumes of other brokerage companies.
We can not do with one model here, we should make 3-4 models, if they cover at least 30-40% of time it will be perfect.
I appreciate the joke!
There is only one adequate model to build... The paradox is that once you've built a model, you don't need anything else. This is the basis which can be put into any tool, including a digital filter. In this light, the primary focus should be on building the model and not on discussing the use of its possible reification.
We will not manage with one model only, we should make 3-4 models, if they cover at least 30-40% of time that would be perfect.
I appreciate the joke!
There is only one adequate model to build... The paradox is that once you've built a model, you don't need anything else. This is the basis which can be put into any tool, including a digital filter. In this light, the priority should be given to building a model rather than discussing the use of possible reification.
That's what I'm doing. Maybe you didn't pay attention to the phrase in the post where I posted the algorithm "....But then there are questions with the divergence criterion of the filter...", it is just a question of checking the adequacy of the model which is laid down in the filter.
Prival, am I correct in assuming that the function defined in the code as NORM(a,s)=a+s*SQRT(-2ln(rnd(1)))*cos(2Pi*rnd(1)) is equivalent to the generator of a normally distributed random variable rnorm(1,a,s)o?
Prival, I understand correctly that the function defined in the code as NORM(a,s)=a+s*SQRT(-2ln(rnd(1)))*cos(2Pi*rnd(1)) is equivalent to the generator of normally distributed random variable rnorm(1,a,s)o?
Yes it's just from an old version, matcad in version 6 did not give a normal distribution when using the built-in generator. The noise was coloured. Using this formula we get a really normal law from a uniform rnd(1). If you need I can look up the mathematics of these transformations how to get any desired distribution law of a random variable from rnd(1).
Simply very long ago I stepped on this rake and very long searched, where error was in program. And found that the noise is dyed, after that I use this formula, as I checked it you can see by chi-square it is a normally distributed random variable.
In this problem it is not so important, so you can change it to rnorm().
Good.
Moving on. I compared performance of Kalman filter with your settings, with optimal (in terms of FS) Butterworth filter, and could not find any noticeable difference not in smoothing properties, not in size of FS, see fig.
Please comment. The script is attached.