Recognise changes in the "behaviour" of a financial time series (Trading on the news)

New comment
 

Comrades, I would like to ask for help with the following question:

Question 1.

I want to write an indicator that would be able to take into account changes in the "behaviour" of a financial time series.

In my head I imagine the following:

Step 1. I know the time of news publication, time moment t;

Step 2. The market is waiting for the news release and begins to change its "behavior" up to the moment t and the market agents are waiting for the news. It is clear that in some agents the information arrives promptly.

Step 3. An indicator captures the change that has occurred, the specific model is ordered to enter into action.

Indicator - what can act as this indicator? Not too complicated an apparatus.

Question 2.

Is there any literature, references, on forecasting models of financial time series, taking into account the news release?


 

Do you have your own ideas?
If not, I don't think someone else's will help

 

Question 1: Estimation of variance, standard deviation, but the point is that the law of the distribution is unknown, as well as from which considerations to take the n that is involved in the unbiased estimate. Phase analysis. Wavelets (they are complex).

Question 2: Econometric models, minar choice model 1 - there is news, 0 - there is no news, but it seems wrong to me.

 
This reminds me of Elder: "The analysis soon reaches a level beyond which the payoff goes down".
Is it worth making things so complicated?
 
Which point does it refer to?
 
I would attribute this to the problem statement as a whole. Simple solutions are unlikely to be available, and there is no adequate return from complex ones.
IMHO
 

How will wavelets help you? They'll smooth out the price... (you won't even be able to tell the difference :-)

And the news - 50/50 - will move the price once - but in the wrong direction (Analysts will say later that the market took into account this news) - or will move in the right direction... but you don't have time to get on the train - the brokerage house won't give a price or requotes...

 
orb:

Comrades, I would like to ask for help with the following question:

Question 1.

I want to write an indicator that would be able to take into account changes in the "behaviour" of a financial time series.

In my head I imagine the following:

Step 1. I know the time of news publication, time moment t;

Step 2. The market is waiting for the news release and begins to change its "behavior" up to the moment t and the market agents are waiting for the news. It is clear that in some agents the information arrives promptly.

Step 3. An indicator captures the change that has occurred, the specific model is ordered to enter into action.

Indicator - what can act as this indicator? Not too complicated an apparatus.

Question 2.

Is there a literature, references discourses, on forecasting models of financial time series, taking into account news releases?


Full.

This is the basic, unresolved problem of kotir forecasting. It's called "breakpoint", in my translation "fracture".

There are a number of kotir tests that identify these breakpoints. I posted them in my thread. Econometrics: one step ahead forecast.

You don't want to look for a specific place.

 
Aleksander:

How will wavelets help you? They'll smooth out the price... (you won't even be able to tell the difference :-)

And the news - 50/50 - will move the price once - but in the wrong direction (Analysts will say later that the market took into account this news) - or will move in the right direction... but you don't have time to jump on the train - the OC won't give a price or requotes or something ...

i see your point. the point is that all problems can be solved with dc and requotes. maybe you can tell me what the indicator may be?

 
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
 
faa1947:

Full.

This is the basic, unresolved problem of kotir forecasting. It's called a "breakpoint", in my translation "fracture".

There are a number of kotier tests that identify these breakpoints. I posted them in my thread. Econometrics: one step ahead forecast.

I don't want to look for a specific location.


can you give me some advice? how, what to look for. what to look for, my econometric models are insignificant, SB all the time except for the model with fractional integration.

12345678...10
New comment
Reason: