Machine learning in trading: theory, models, practice and algo-trading - page 2970
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
It wouldn't work like a normal one. You need some clever auto-tuning for unknown data, screening out noise signals, etc. to do about the same thing as a human.
Yeah, I've done that and thought along the same lines.
Yes, with such a filter the result on the test was much more reliable.
Specifically with FF on profit, I did something like a Monte Carlo simulation on the current balance and if the forecasts from simulations on the test were all in growth, it was like a filter that the found current balance will grow in the future. Which was confirmed...
I can't figure out what, but something protests in me against your idea with the target as a balance.
A person wanted to train on balance, not me, for him I explained how to do it, then you wanted a code, I made an "easy" example to make it clear how it works at all....
Why do you fixate on this balance? There are other variations of FF that are much better. Switch off the tunnel thinking and go with the song....
PS.
Not to the night mentioned zigzag is the same balance, but marked into longs and shorts.
that is if you do not think about it very much, but if you think about it you will understand the limitations of ZZ compared to FF in terms of profit, only now 3 things came to my mind.
a person wanted to train on balance, not me, for him I explained how to do it, then you wanted the code, I made an "easy" example to make it clear how it works....
Why do you fixate on this balance? There are other variations of FF that are much better. Switch off the tunnel thinking and go with the song...
that's if you don't think about it too much, but if you think about it, you'll realise the limitations of ZZ compared to FF in terms of profit, I've just now come up with 3 things.
Fuck the ZZ.
The disadvantage of ZZ is that it replaces the unsteady quotient with regular links, no sudden movements. Balance plays the same role. They are similar in this.
FF has to filter out the non-stationary, rare movements. In fact, the only method of predicting non-stationary processes is MO with its ability to form patterns. Since undesirable patterns are extremely rare, we will get the same extremely unbalanced classes. If no special effort is made for unbalanced classes, you will easily get overfitting, i.e. a pattern rigidly tied to the training sample.
Why is your FF not an overfitting tool?
Why isn't your FF a tool of over fitting?
Maybe there's something I don't understand.
Doesn't FF do the selection of feature values?
Maybe there's something I don't understand.
Isn't it the FF that does the feature value selection?
In the balance example, no.
Thanks, I'll have to think about it
Couldn't you take a family of standard benchmarks (you can borrow from Georges) and teach DL/ML to trade in the same way as the best ones, or at least not worse ?
and do exactly the same as the notorious midjourney ?
PS. if I were Georges, I would raise the price tag for the history of zoo trades now...by 2-3-5 orders :-)
The sellers have one thing on their mind.
Yeah, profit.
What else do people have in mind, what's the goal? A way of counting the final number of stars in the galaxy??
I mean, no.
---
It's well recognised that NN training requires "real samples". That is, long-lasting steytes, preferably made by humans (or controlled by them) with strictly specified options. And in significant quantities. And there are none
Georges started his zoo for a different reason, but his steytes are in the top now. There are no live samples, so at least they have them.
oops...I replied and others didn't...either she deleted herself or the moderators did.