Machine learning in trading: theory, models, practice and algo-trading - page 3477
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
Interesting article, 80% pop but interesting nonetheless....
---------------------
In continuation of this #34691 #34692 #34700 discussion.
The author of the article above has the same result
So think, see traders...
My hypothesis explains all the incomprehensible things that happen with TS/MO on new data, and reasonably explains.
I can answer that if you explain the terminology.
Well, you seem to be using the terminology yourself:
In-sample is data that has been used to develop and optimise a trading strategy. Usually it is historical price data, which have been tested with the help of various technical indicators and algorithms.
Out-of-sample is new data that has not been used before. They help to test how well a trading strategy works in new market conditions.
Independent testing area - the one that the algorithm was not aware of during training/optimisation/tuning/fantasy.
The percentage of selected results during the training period is the number of hypothesis-models/tunings that were obtained without using the independent testing section.
The question is what percentage of these hypotheses were confirmed and what percentage were incorrect.
Well, you seem to be using your own terminology:
In-sample is data that has been used to develop and optimise a trading strategy. Usually, it is historical price data that has been tested using various technical indicators and algorithms.
Out-of-sample is new data that has not been used before. They help to check how well a trading strategy works in new market conditions.
Independent testing area - one that the algorithm was not aware of when training/optimising/tuning/fantasy testing.
Percentage of selected results during the training period is the number of hypothesis-models/tunings that were obtained without using the independent testing section.
The question is what percentage of these hypotheses were confirmed and what percentage were incorrect.
My hypothesis explains all the incomprehensible things that are going on with TC/MO on the new data, and reasonably so.
The noise is just catching...
The noise is just catching...
why isn't it 50/50 ?
Noise has an average = 0.
why not 50/50?
Noise has an average = 0.
So noise can be 70%-80% initially - that's what I'm getting at here.
So the noise might be 70%-80% initially - I'm getting to that figure here.
It doesn't matter how much noise there is, noise has a mean of 0.
So there shouldn't be an inverse correlation, but there is.
By your logic, it should be like this.
It's not like that at all.
Well, you seem to be using your own terminology:
In-sample is data that has been used to develop and optimise a trading strategy. Usually, it is historical price data that has been tested using various technical indicators and algorithms.
Out-of-sample is new data that has not been used before. They help to check how well a trading strategy works in new market conditions.
Independent testing area - one that the algorithm was not aware of when training/optimising/tuning/fantasy testing.
Percentage of selected results during the training period is the number of hypothesis-models/tunings that were obtained without using the independent testing section.
The question is what percentage of these hypotheses were confirmed and what percentage were incorrect.
The algorithm to find the maximum FF was interrupted after computing 3000 FFs. Then sorting the 3000 results by FF value and running the best 20 of them on OOS. Among them, OOS passes sometimes 50%, sometimes 5% or 0%. This percentage does not tell us anything about the robustness of the TC. Because the search algorithm is monomodal.
It seems to me that we should wait a bit until the optimisers are tired of their FFs. Then they will start to absorb the information :)
After all, if nothing has happened in 20 years, what are the chances that something interesting will happen?
So far they have settled on the need to choose the right FF, which nobody knows how to choose, and if they do, they will never tell. @quote one of.Let the hype, which is at its peak with all sorts of "AI" stuff, subside with all sorts of optimisdatins. :) Otherwise they will run around in all the topics and FFK.
At least DSP was defeated in its time, and that's already a wine
Otherwise it's just crying 😂