Wish List

Synergy data mining application developed by AdaptiveTradingSystems.com
jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Wish List

Post by jamess » Mon Jan 23, 2017 9:53 pm

Thanks Carlos, I understand what you are saying now. Behavior would have changed when I implemented zooming for the model performance charts. I'll see what I can do.

Kind Regards,

James

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Wish List

Post by CarlosR » Thu Jan 26, 2017 3:34 am

James, here's another one for the wish list: it would be nice to be able to put in models by hand, and have Synergy analyze them. What I'm thinking of is that sometimes I get a Synergy model that looks good when I'm doing a run not using the PSO, and I'd like to know how sensitive it is to varying the parameters. It would be nice to just be able to change the values by hand and see what happens.

Just a thought, not a biggie.

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Wish List

Post by jamess » Thu Jan 26, 2017 10:50 am

That does sound like it would be an interesting feature to have. Maybe something like a sliding range for each parameter so we could drag and slide and view the impact.

Kind Regards,

James

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Wish List

Post by CarlosR » Mon Jan 30, 2017 3:17 pm

Yes, that would be great!

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Wish List

Post by CarlosR » Sat Feb 11, 2017 5:46 pm

Hi James,

Would it be possible to incorporate into the User Guide a list of formulas for each of the function blocks? (You might want to leave out any proprietary ones, of course)

While many are obvious, some of us who are less mathematically inclined aren't quite sure what the "log modulus transformation" is, just to pick one of many.

Or if you don't want to put it in the User Guide, perhaps it could be available as a separate file?

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Wish List

Post by jamess » Sat Feb 11, 2017 10:47 pm

Hi Carlos,

I'll do that. Good suggestion thank you.

Kind Regards,

James

Schutten
Posts: 7
Joined: Thu Jan 12, 2017 11:02 am

Re: Wish List

Post by Schutten » Fri Jul 07, 2017 8:02 am

Hi James,

Would it be possible to implement a merged model just like in profit? All models for example which are marked for export are combined by average signal or majority voting. For me this would help if I want to use the non adaptive signals directly from Synergy. But I need to have this 'master' signal exported as well with signal export. In this case I don't have to use Dakota anymore.

Regards,
Dennis

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Wish List

Post by jamess » Fri Jul 07, 2017 10:23 pm

Hi Dennis,

I'm currently in the design phase of a set of related applications for building systems using Synergy models. It will walk the models forward and combine model signals in an adaptive manner. There will be two types of walk-forward adaptation:

1. The model parameter values will be adapted. No need to limit the model parameter ranges. The optimizer will work with the same parameter ranges that you see on the Function Blocks tab. For example, the period of a simple moving average could be anything from 2 to 100 or even wider if you configure the range to be wider on the Function Blocks tab. This is how the walk-forward simulator in Synergy works.

2. A system will have a pool of models to work with. You will be able to add to the pool, or remove models from the pool while the system is 'live'. The system will use a subset of the models at any time. You won't be able to remove models that the system is currently using, but you can mark them for deletion with the possibility to force a re-optimization of the ensemble. Models that are added will be available for use the next time the ensemble optimization runs.

There would be the option to use a fixed set of models in a system as well. In that case there would be no automatic walk-forward adaptation of the ensemble of models used to produce the trading signal.

Kind Regards,

James

CarlosR
Posts: 51
Joined: Sun Jan 08, 2017 10:52 pm

Re: Wish List

Post by CarlosR » Sat Jul 08, 2017 12:31 am

James, that sounds very interesting. I currently do that externally to Synergy, using Excel. One nice thing about doing it that way is that I can aggregate sub-sets of models, have them vote, and then combine the votes of the subsets.

For example, I might have 10 Synergy models that generate a signal to trade ES based on the VIX, 21 that generate long-only trading signals based on the ES data itself, 4 that generate a short-only signal based on the same ES data, and 7 that use another intermarket data stream as input. I can combine each of those subgroups separately, yielding 4 signals, and then combine those 4 signals with equal (or unequal) weights to generate the master trading signal.

Do you think you can build in that type of flexibility, or will that make it too complicated?

I do like your idea of having the process be adaptive. That's the one thing I miss about not doing the combining in Dakota, and I'd love to have that capability available.

jamess
Site Admin
Posts: 55
Joined: Wed Jan 04, 2017 9:52 pm

Re: Wish List

Post by jamess » Sat Jul 08, 2017 12:42 am

Hi Carlos,

That can definitely be done. I'm glad you mentioned it now.

Kind Regards,

James

Post Reply