Machine learning in trading: theory, models, practice and algo-trading - page 3476

New comment
 
Maxim Dmitrievsky #:

You can evaluate the result of your optimisation in a simple way.

  • Optimise the TS for different periods (take the best ones according to your subjective criteria).
  • average the parameters, the TS will become unbiased
  • run on the whole dataset + new data

Far from a fact...

There are two TCs( aprimitive but illustrative example):

on one plot the best periods are 10, 21,

on the second one, the best periods are 30, 150.

on the third, 5, 145.

etc..


Averaging the periods into one TS, we get a plum.

 
mytarmailS #:

Far from it.

TC is aprimitive but illustrative example:

in one section the best periods are 10, 21,

on the second section the best periods are 30, 150.

on the third, 5, 145.

etc.


Averaging the periods, we get a drain.

So it's a losing trade that memorises history.

It won't work on a new history.
 
It is clear that the more parameters the TS has and the spread of their values, the more optimisations on different intervals should be made. Statistics, Karl, is a stubborn thing!
 

One can do a mental Python experiment on a pre-predicted BP (sinusoid etc.) how the TC parameters will converge to the optimum in this way.

 
I've been reading about p-hacking all day today...I'm sick of it.
 
mytarmailS #:
I've been reading about p-hacking all day today. I'm sick of it.
What's new?
 

I can't figure out the old one properly.

I realised Prado plagiarised the idea from the scientists when he wrote about the PBO.

 
mytarmailS #:

I can't figure out the old one properly.

I realised Prado plagiarised the idea from the scientists when he wrote about the PBO.

And now he's a goat teaching students.
If you can find him on Twitter, he's on there all the time.
 
Tell him the guys in the MO thread are wondering what's up with the strategies 😀
 

thegrail is buried in generating TCs with a probability of earning >50%.

The task has already been asked ridiculous in essence...


It is necessary to develop a method for selecting TCs, at least according to the same PBO from Prado.

1...346934703471347234733474347534763477347834793480348134823483...3674
New comment