Suggest a topic for research - page 4
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
If you have a good MAKD EA, use a standard one (from MT4 delivery) and make it profitable for the last 5 years at least.....
hint - there's only one virtual trade in it and only one of them goes to real :)
i think it's good for the pocket, but the head of the department does not accept it. this is a technical analysis. we need mathematics. while at university you have to develop your brain mathematics, not indicators, which are also good in their own way.)
then think of the MACD as a digital filter, describe the maths through the z-transform, see the frequency response, etc., there's plenty of water to pour.
Thanks, I'll think about it. in fact, everything you wrote is familiar, but has been studied without much understanding.
And in general, it's nice that people respond to requests for help with the topic.
And in general, it's nice that people are responding to requests for help with the topic.
Look, there are two approaches broad linear and linear, here we have all methods linear in essence, we were taught AR, SS, ARPSS, etc. classics. Here if we estimate ACF and CHAFC of the first quote difference, we will see similarity with ACF and CHAFC of white noise, consequently the model y(t)-y(t-1) gives stationary error that is a good sign, stationary time series models will not give us adequate results as there is no lag in ACF and CHAFC (one of the criteria to start selection with), I built these models anyway and made sure of it myself. I concluded that in the linear approach a random walk model is appropriate and hence the best predictive value is the value at the previous point in time. But it is nonsense, isn't it? Who would trade like that?
Constructed a model and indeed for the first differences, y(t)=b*y(t-1)+e. The coefficient. Always means and close to 1. And I found in the book, one how you can try to predict the random walk. After all we need in fact only sign of increment on the next step, because I worked with first differences, coursework not yet passed, so I want to try, like when close to zero the situation is uncertain, and the further from zero the more likely that sign of increment will really be so.
The main idea is: maybe the truth is far away and it is not correct to describe the behaviour of the increment (first differences) as a random walk, but it can be done safely within the linear approach.
There are more conclusions, but they are trifles, I rather dispelled my own doubts.
You have a half tool in your hands: you know how to make a forecast, but you don't have elaboration of how to use it. Without the other half (the use of the forecast) it is meaningless to do the forecast and its refinement. NS -= it's just classifications and it's loved here because it's very close to TA, which is looking for patterns. Taking wavelets is promising, but pointless at this stage, as the problems of using the forecast are not solved and it's impossible to evaluate the forecast itself.
I would distinguish two directions in using the forecast:
1. to account for forecast error
2. to take into account the direction of the forecast due to the gradient and the derivative model of the forecast. The latter is not just a diploma.
Listen less to the visitors to this site - this is the site of programmers, cyclists and TAs. The word "econometrics" is a swear word here. Look at my thread "Econometrics. One step forecast" and you will understand the miserable level of the site in the field of econometrics.
Good luck.
faa1947:
a lousy level of econometrics on this site.
- What is the purpose of the study?
- To program something.
Commendable.
We have been taught AR, SS, ARPSS etc classics. Here if we evaluate ACF and CHAF of the first quote difference, we observe similarity with ACF and CHAF of white noise, consequently the model y(t)-y(t-1) gives stationary error, which is a good sign, models of stationary time series will not give us adequate results, because there is no lag in ACF and CHAF (one of criteria, with what model to start selection), I built these models anyway and made sure of it myself. I concluded that in the linear approach a random walk model is appropriate and hence the best predictive value is the value at the previous point in time. But it is nonsense, isn't it? Who will trade like that?
Constructed a model and indeed for the first differences, y(t)=b*y(t-1)+e. The coefficient. Always means and close to 1. And I found in a book, one how you can try to predict a random walk...
Too many letters. From what I understand, you should go to the faa1947 thread - a similar train of thought.
You just understand one thing, the purpose of any research is to try out ideas. If there are no ideas, then doing research is pointless.
That's commendable.
Too many letters. From what I understand, you need to branch faa1947 - a similar way of "thinking".
You just realise one thing, the purpose of any research is to test ideas. If there are no ideas, there is no point in doing research.
Really doesn't make sense? and I thought it did)))) "MQL is just a tool to try out ideas) not the subject of the post as there are no ideas.) - MQL is just a tool for approbation of ideas) not the subject of the post, because there are no ideas, that's why I'm asking. is it not clear: "Please pass by if you don't have any ideas. MQL is just a tool to try out ideas, that's why I'm asking: "'MQL is just a tool to try out ideas" - that's the point, to avoid lyrics in conversations aka bloating for life, for approbation, for mathematical ponces=)
You have half the tool in your hands: you know how to make a forecast, but you have no elaboration on how to use that forecast. Without the other half (the use of the forecast) there is no point in doing the forecast and refining it. NS -= it's just classifications and it's loved here because it's very close to TA, which is looking for patterns. Taking wavelets is promising, but pointless at this stage, as the problems of using the forecast are not solved and it's impossible to evaluate the forecast itself.
I would distinguish two directions in using the forecast:
1. to account for forecast error
2. to take into account the direction of the forecast due to the gradient and the derivative of the forecast model. The latter is not just a diploma.
Listen less to the visitors to this site - this is the site of programmers, cyclists and TAs. The word "econometrics" is a swear word here. Look at my thread "Econometrics. One step forecast" and you will understand the miserable level of the site in the field of econometrics.
Good luck.
Some of it is understood, some of it is not. Thanks for the wishes!
Really pointless? I thought there was a point)))) "You've taken something out of context)). - MQL is just a tool for approbation of ideas) not the subject of the post, because there are no ideas, that's why I'm asking. isn't it clear: "please pass by if you don't have any ideas. "This makes sense, so as to avoid lyricism in conversations aka flubbing for life, for approbation, for mathematical ponces=)
No offence. Just turned out a nice collage, that's all.