Bayesian regression - Has anyone made an EA using this algorithm? - page 22

 
MikeZv:
You didn't answer the question...
Have you achieved anything in this direction?
Certainly, the "years and years spent" have not gone unrewarded.
 
Олег avtomat:
Of course, the "wasted years and years" did not go unrewarded.
Is there a real working TS, for 5 years bringing at least 100% a year with a drawdown of no more than 20% ?
Only without weekly optimization ... :)
 
MikeZv:
Is there a really working TS, for 5 years, bringing at least 100% per year with drawdown not more than 20%?
Only without weekly optimization ... :)

For a TC to meet these requirements, it must have been established at least 5 years ago. But those are your criteria. My criteria are different.

As for"weekly optimisation" - you also have no understanding of the essence of the matter.

 
Олег avtomat:

And on the subject of"weekly optimisation" -- you have no understanding of the merits of the case here either.

Share ...
 
MikeZv:
Share ...
Share what? Understanding? ;))
 
Олег avtomat:
With what? Understanding? ;))
Yes.
 
MikeZv:
Yes.
Cut off a piece? ;))
 
Олег avtomat:
Cut off a piece? ;))
"Then will also come the understanding in which direction your irony should be directed."
 
MikeZv:
"Then will come an understanding of the direction in which your irony should be directed."
Open the books in that direction -- perhaps along with understanding the essence of identification will come understanding the essence of optimisation.
 
MikeZv:
This initial sample has to be very large ...
There should be at least 100 trades in a subset, and at least 100 subsets.

The other way round. This method was developed for short samples. 500 observations is a lot. Usually 50-100.

For example. There are 30 predictors - input data. From these, we need to select some subset of predictors that have the strongest impact on the target variable. In my examples selects 10-15 predictors that are relevant to the target. Gives statistics with confidence intervals.