Recognise changes in the "behaviour" of a financial time series (Trading on the news) - page 2
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
can you give some advice? how, what to look for. what to look for, i have econometric models insignificant, SB all the time except for the model with fractional integration.
Don't take someone else's advice - it's your skill, the other person has their own skill.
This skill can be obtained by using some econometric package and then reading the literature on the issues you don't understand in the package. This is a learning tip.
There is no universal model, at least not to my knowledge. At one point in time one works, at another point in time another works, and between them there is a breaking point and the whole problem is in them. Attached is a recent article that can be used as an input to the problem.
I am also dealing with this issue recently. I want to build an algorithm on the news. to start with, I have an indicator:https://www.mql5.com/ru/code/10741
Don't take someone else's advice - it's your skill, the other person has their own skill.
This skill can be obtained by using some econometric package and then reading the literature on the issues you don't understand in the package. This is a learning tip.
There is no universal model, at least not to my knowledge. At one point in time one works, at another point in time another works, and between them there is a breaking point and the whole problem is in them. Attached is a recent article which can be used as an input to the problem.
the book is good, but difficult because it is in english.
orb:
книга хорошая, но трудна из-за того, что она на англ. языке.
In Russian we are at least five years behind.
There is no universal model, at least not to my knowledge. One moment one is working, the other is working, and there is a breaking point between them and the whole problem is in them. Attached is a recent article which can be used as an input to the problem.
So identify the moment when the old model stopped working, what's the problem?)
I don't know how. A breakpoint on the history and by the time I managed to determine it a new breakpoint has arisen, which I will determine after some unknown time.
maybe write an EA that would unload Volume, Time, Close,Open,High,Low, candle body at break point calculate analyse. then based on all history evaluate the binary selection model.
maybe try a logit, a breakpoint model to build a result variable 1 - there is a breakpoint, 0 - there is no breakpoint?
I don't. A breaking point on history and by the time I managed to determine it a new breaking point has arisen, which I will determine after some unknown time.
alsu:
orb
Each model has a criterion of applicability to the current situation, roughly speaking, the minimum possible error at which we believe the model works. Why not narrow down the interval to make detection of outliers the fastest possible?
There is no complete solution, only approaches, as shown in the attachment above.
The problem is understood intuitively. To fit the model, the larger the sample, the better. But the bigger the sample, the less it takes into account the current situation. It would seem that AP(1) is ideal - only previous candle, but no, it works, but very rarely.
The model should predict the breaks, but how to do it?
maybe write an EA that would unload Volume, Time, Close,Open,High,Low, candle body at break point calculate analyse. then based on all history evaluate the binary selection model.
maybe try to logit, breakpoint model to build a result variable 1 - there is a breakpoint, 0 - no breakpoint?
write - it's easy to output the parameters to the log file via Print... or better yet to a separate file via file operations...
and then also make a table by Exodus when there was ++ and after which number of Minuses...
because it's possible that even though the total number of --- will be greater than +++
but by manipulating lots you may be able to display the overall result in +++