Download MetaTrader 5
To add comments, please log in or register
Share links on the Forum. Now they have Rich Content!
Behzad Sabet
3257
Behzad Sabet 2015.08.22 19:56 

Dear friends!

how do you think about modern methods to forecast the market. ARE YOU FAMILIAR WITH THESE METHODS LIKE NEURAL NETWORKS OR ARIMA OR GARCH. these are used by international organization like IEA to predict oil gold dollar and ... is it applicable to use them in our predictions?

Marco vd Heijden
Moderator
4079
Marco vd Heijden 2015.08.22 20:34  
not possible.
Behzad Sabet
3257
Behzad Sabet 2015.08.25 21:32  

WHY BUT I THINK THERE ARE MANY TRADERS WORK BY THIS SYSTEMS AND MANY EAs  WORK BY THESE METHODS DO YOU HAVE SEARCHED IN THIS FIELD?

Marco vd Heijden
Moderator
4079
Marco vd Heijden 2015.08.25 22:23  

nobody or no system can PREDICT the future.

calculation of PROBABILITY is possible but this is NOT the same it is EXPECTATION no PREDICTION.

Anybody who tries to sell a system claiming it can predict the future is a liar and is after your money.

Young Ho Seo
14068
Young Ho Seo 2015.08.25 22:53  
behzad sabet:

Dear friends!

how do you think about modern methods to forecast the market. ARE YOU FAMILIAR WITH THESE METHODS LIKE NEURAL NETWORKS OR ARIMA OR GARCH. these are used by international organization like IEA to predict oil gold dollar and ... is it applicable to use them in our predictions?

Hi.

prediction (i.e. using Neural networks, ARIMA, etc ) is done for two purpose.

1. Expected future direction - note I am talking about expected future direction. This is often used as good referencing level of the security prices by many firms.

2. Possible risk analysis - some times knowing prediction interval is more important than knowing direction


Such a mathematical method is good because they always state their own limitation unlike other trading system. For example, Regression, ARIMA and Neural networks will implicitly and explicitly state their expected forecasting accuracy plus their worst and best outcome based on their model.

That is to say that you will know how accurately the model predict on past data (performance on the data) plus you will also know what is potential risk of using such a model in real world application too (i.e. prediction interval and coefficient interval).

Note that prediction interval and coefficient interval gets smaller and smaller with more number of historical data.

When there is little number of data, you have to watch out that model become less accurate because large prediction interval and coefficient interval (more influences on coefficient interval though).

 

Then how these compare to typical trading method. For example comparing to Japanese candlestick pattern it is used by many traders but they don't give you what is the potential risk of using such a model in real world application.

For example, there is no ways of telling the prediction interval of the Doji candle appeared in EURUSD or any other currency pairs.


Note that I am not here to discuss which trading method is superior.  However generally professionals using Regression, ARIMA, etc knows that what is the limitation of their models before they use assuming they have learnt and mastered these models.

So they don't make overly stupid decision either.


Just one interesting discussion. I hope this answers you some of your questions.


Kind regards.




Here is some reference about prediction interval for your information.


http://www.real-statistics.com/regression/confidence-and-prediction-intervals/

Confidence and prediction intervals for forecasted values
  • www.real-statistics.com
This means that there is a 95% probability that the true linear regression line of the population will lie within the confidence interval of the regression line calculated from the sample data. In the graph on the left of Figure 1, a linear regression line is calculated to fit the sample data points. The confidence interval consists of the...
Behzad Sabet
3257
Behzad Sabet 2015.08.26 14:28  

Dear Friend

Thanks for your good answer. Are you familiar with these method? are you work in this field?

Young Ho Seo
14068
Young Ho Seo 2015.08.26 14:52  
behzad sabet:

Dear Friend

Thanks for your good answer. Are you familiar with these method? are you work in this field?


I have fair amount of experience of working with analytical professionals using these methdos inside industry and academia.

Kind regards.

Behzad Sabet
3257
Behzad Sabet 2015.08.26 15:05  
do these method really applicable in science and industry? if so that it can be also good to predict market? are you agree?
Marco vd Heijden
Moderator
4079
Marco vd Heijden 2015.08.26 15:32  
behzad sabet:
do these method really applicable in science and industry? if so that it can be also good to predict market? are you agree?

Yes they do, they are very good at predicting the Future.

Some can only look a few day's into the Future, but others,

more complex ones can Predict and look several Years ahead into the Future.

They just print out a list of prices and dates so you can look up tomorrow's and next weeks trend.

This is how many people became a Millionaire. 

ADRIAN MATUSIAK
1119
ADRIAN MATUSIAK 2015.08.26 16:10  
Neural networks? Yes, it's very good thing. But it requires a lot of memory and very fast proccessor to make all those calculations. I still suggest to use "old methods".
Behzad Sabet
3257
Behzad Sabet 2015.08.26 20:32  
But if you can predict next day prediction you have one day time to calculate and predict it? isn't enough? do you know anyone using these methods in practical?
/
To add comments, please log in or register