Hey Everyone!
Author: Glitch
Creation Date: 9/1/2017 3:01 PM
profile picture

Glitch

#1
Who remembers me, Wealth-Lab user #0, from the good old days back in 2004. Yes, 2004 was my last post here, because at that time I became part of the Fidelity organization and due to compliance reasons was not able to communicate on the forums here any longer. Well, I'm no longer part of that organization as of today and it's a pleasure to be able to post here again! Who here remembers me?? I see that Cone and Eugene have continued doing an amazing job of support and development. But who's this "vk" guy that keeps popping up here and there??
profile picture

vk

#2
Yes, who is actually that VK guy? Anyway, I think he is the only one here who remembers you! ;)

WELCOME BACK!

vk
profile picture

Glitch

#3
Hopefully Cone and Eugene remember me too!
profile picture

Eugene

#4
Hi Dion!

Very nice to see you here again! :) For sure, your unprecedented expertise will be appreciated.
profile picture

Glitch

#5
Glad to be able to contribute here again! Thank you!
profile picture

Carova

#6
Remember?? Who of us from those days could forget. Welcome back!

What are your plans now?

Vince
profile picture

Glitch

#7
Hi Vince, someone remembers me! ;) I have some ideas brewing, backtesting has always been my passion so I think I want to develop something new in that area. Will keep everyone posted as ideas and results crystalize!
profile picture

Carova

#8
Glitch, if you have some concepts about the following it would be appreciated:

1) Stochastic sampling during backtesting (similar to bagging)
2) Concurrent OOS assessment during optimization (common in most supervised learning environments)
3) Easy access to the wealth of machine learning software (for those of us who are not C# programmers)

Thanks!
Vince
profile picture

Glitch

#9
I think my interest will lie in trying to expose concepts like this in logical and understandable pieces that can be easily used to build and then test a trading model. We have some of these things in WL already (Neural-Lab is a subset of machine learning) and WFO is a type of OOS assessment, but there could be much more done to integrate concepts like this into a backtesting platform. Still processing ideas ...
profile picture

abegy

#10
Hi Glitch,

I'm happy to see that you are looking to improve the platform. I agree with you that a better integration of existing concepts in WL is a good idea.

If you want also some new ideas, I can recommend you to implement the multi-objectives concept (https://www.wealth-lab.com/Forum/Posts/Request-add-multi-objectives-to-Particle-Swarm-Optimization-38794) for all optimizer algorithms.

But do not forget that the greatest missing concept is the multi-core implementation. Others software offers now this feature as a standard.

Can I ask you one question : what about all open issues ? Can we expect a maintenance release to correct them ?
profile picture

Glitch

#11
Hi abegy - I don't want to give anybody the wrong idea - I'm actually not working on the WL platform per se any longer. That's in the hands of Fidelity, of which I have just recently departed. I'm here to share ideas and thoughts about backtesting and to help out in any way I can as an unofficial WL person at this point. I have some idea churning right now for what to do next, and your comments regarding multi-objectives and multi-core are appreciated!