Machine learning in trading: theory, models, practice and algo-trading - page 3542

New comment
 
mytarmailS #:

I don't get it, deviation of what from what?

The price from your model. Values. To formulate sets.
 

I review it on different pairs and TFs and TG rules.

Sometimes it draws such a picture that I jump up from my chair.)



but there are still more of them, so it's not all that pretty.



 
Maxim Dmitrievsky #:
Prices from your model. Values. To formulate setups.

Model prices from real prices?

Yes this is in the plans +

to add many models

 
mytarmailS #:

the model prices from the real prices?

Yes, this is in the plans +

add many models

Yes, the difference. That's also how arima and garch fics are cooked up, along the same lines. Regression models as features. But you'd have to go through a lot.
I think yours is called a distributed lag model.
 
Maxim Dmitrievsky #:
Yeah, the difference. That's also how arima and garch fiches are prepared, along the same lines. Regression models as features. But it's a lot to go through.
I think yours is called a distributed lag model.

The hardest part is finding the right target function to make it work.

 
mytarmailS #:

the hardest part is getting the target function right to make it work.

You can't pick it from the ceiling, you have to look at the data 😀
 
Maxim Dmitrievsky #:
You can't pick it from the ceiling, you have to look at the data 😀

Well, the zigzag is definitely not in the kasa, and the turns are not clear either.

 
mytarmailS #:

Don't make up things that aren't there.

AMO only has access to one last value of the fly and the last value of RSI... ALL!!!!

There are no matrices, shmatrixes.

THERE IS. It's a system with memory. In the process of learning, you fill it up.

It is probably not quite clear with matrices, people here mostly think in algorithms.

The basic steps of the alg. "reconstructing a price series from the SMA(period) series".

- initial K=1, consider last price Last = K*last_value of SMA.

- using the inverse incremental SMA function we obtain an approximate price series P

- vary K, minimising e.g. the price spread (maximum-minimum) in the price series P

MA is not a filter, it does not throw out anything, it is a scaling. These are the same prices in a curved form, the equivalent of a price series.

This is actually the reason why TC is categorical about using prices and their MAs in training. It would be silly to use prices and prices again.

 
Maxim Kuznetsov #:

IS. A system with a memory. In the process of learning, you fill it up.

God bless whoever understood what you wrote.
 
mytarmailS #:
God bless whoever understood what you wrote.
It's a bunch of random words
1...353535363537353835393540354135423543354435453546354735483549...3616
New comment