• Registered: 2024-02-23
  • Posts: 25

Topic: IS and OOS

Dears,

I would like to set the Reactor to generate strategies according to the following process:

1. Generate strategies (doesn’t matter the entry/exit rules) one optimize them to pass my acceptance criteria in a back test In Sample

2. Afterwards, testing all strategies validated under a back test Out of Sample without optimizing them

3. Then all strategies validated OOS should go to the Montecarlo for the stress test.

4. Out of the MC landing in the collection.

What is the best way to setup the Reactor to get this process in place?

Thanks
Vincenzo

  • footon
  • footon
  • ◄≡≡≡►
  • Offline
  • Registered: 2009-08-17
  • Posts: 2,925

Re: IS and OOS

It depends whether MC is done on IS or the whole dataset (IS+OOS). If the latter, then you have to first generate collection(s) in the Reactor and then validate them in the Validator.

  • Registered: 2024-02-23
  • Posts: 25

Re: IS and OOS

I am sorry, but I did not get it.
Could you try to elaborate a bit more?
Thx

  • footon
  • footon
  • ◄≡≡≡►
  • Offline
  • Registered: 2009-08-17
  • Posts: 2,925

Re: IS and OOS

If I understand your points correctly, then you first must generate collection(s) in the Reactor by using data you regard as IS. After that you validate the collection(s) in the Validator with data you designate as OOS, there's only MC to be done. After that you are left with collection(s), which correspond to your development points written in the first post.
Try it out, you'll learn it much quicker then.

  • Registered: 2024-02-23
  • Posts: 25

Re: IS and OOS

Thanks a lot.

I will give a try with this set up:
Generator settings
- search best: net balance
- out of sample: in sample (I suppose that this allow to generato strategies on the whole set of data)

Full data optimizations ON
- out of sample 30% (I suppose this will not optimize 30% of the data sample)
- numeric value range +/- 10 steps
- optimize SL and TP
- optimize preset indicator
- use common acceptance criteria

Walk forward optimization OFF

Normalization ON
- all flags as usual
- optimization: in sample, 20 steps, net balance
- normalize preset indicator on
- use common acceptance criteria

Robustness
walk forward validation ON, 2 segments, out of sample 30%, +/- 20steps, net balance, optimize SL & TP OFF, optimize preset indicator OFF, validated segments 2

Montecarlo ON as usual.

Is this reflecting my intention ?

  • Registered: 2024-02-23
  • Posts: 25

Re: IS and OOS

Dears,

May I have a quick answer here ?
If I want to generate and optimize strategy on only 70% of the historical data, should I set the generator and the optimizer in this way?

Thank you in advance
Vincenzo

Post's attachments

IMG_1785.jpeg 357.13 kb, 5 downloads since 2024-03-21 

IMG_1786.jpeg 327.44 kb, 3 downloads since 2024-03-21 

You don't have the permssions to download the attachments of this post.
  • footon
  • footon
  • ◄≡≡≡►
  • Offline
  • Registered: 2009-08-17
  • Posts: 2,925

Re: IS and OOS

You set OOS to 30%.

  • Registered: 2024-02-23
  • Posts: 25

Re: IS and OOS

Okay, thanks

  • Registered: 2024-02-23
  • Posts: 25

Re: IS and OOS

Testen, all good !
Thanks a lot.
Vincenzo


footon wrote:

You set OOS to 30%.