GSB Forums

New Equity Highs From ES system

heyligerb - 6-12-2018 at 09:04 PM

I built this system in June of 2017... it's by no means perfect, but it's making money, and that's more than I can say for other systems I've built.

Attachment: Login to view the details

admin - 7-12-2018 at 03:29 AM

Great to hear. Is this a few systems in one, or some sort of postiion sizing method?

cotila1 - 7-12-2018 at 04:17 AM

Quote: Originally posted by heyligerb  
I built this system in June of 2017... it's by no means perfect, but it's making money, and that's more than I can say for other systems I've built.


Same sort of comments with my systems. Sept 2018 - I've built few ES sytems with new GSB methodology discussed trough Peter's videos.
Tough may be a bit early, I have built systems with Nth and MultiTF approach (plus many other tips for validation-such as multimarket). Furthermore in the specific example of the screenshot I have used also an un-seen period since 1-1-2015. So period from 1-1-2015 till today has never been seen by GSB (see red line in screenshot). Slpg & Comm inluded.
SPR attached too.


Screenshot.jpg - 172kB

Attachment: Login to view the details

heyligerb - 7-12-2018 at 08:02 AM

The answer is yes with both! I have a 16 GSB different systems/signals in there, and I'm using different position sizing for each system/signal.

admin - 7-12-2018 at 03:26 PM

Quote: Originally posted by heyligerb  
The answer is yes with both! I have a 16 GSB different systems/signals in there, and I'm using different position sizing for each system/signal.

I take it you limited total contracts to 3?
You should get much better results now with newer GSB with 29,30,31 min bars and verify on 25_35

heyligerb - 13-12-2018 at 07:43 AM

correct... 3 contracts max, and depending on the signal(s) that fired, I would adjust the contract size manually (1-3) using something called adaboost which is another classifier that would give me a greater confidence in the signals firing together. It would just add another layer of ML to the overall picture.

But looking forward to getting into some more GSB. It's been a while, and time has been short...

admin - 13-12-2018 at 02:40 PM

Quote: Originally posted by heyligerb  
correct... 3 contracts max, and depending on the signal(s) that fired, I would adjust the contract size manually (1-3) using something called adaboost which is another classifier that would give me a greater confidence in the signals firing together. It would just add another layer of ML to the overall picture.

But looking forward to getting into some more GSB. It's been a while, and time has been short...


Is it worth explaining more about adaboost? Limiting the amount of contracts per day is likely to decrease a lot of metrics good and bad. Ie I expect pf, at to reduce, but also drawdown

boosted - 2-1-2019 at 11:30 PM

Quote: Originally posted by admin  
Quote: Originally posted by heyligerb  
correct... 3 contracts max, and depending on the signal(s) that fired, I would adjust the contract size manually (1-3) using something called adaboost which is another classifier that would give me a greater confidence in the signals firing together. It would just add another layer of ML to the overall picture.

But looking forward to getting into some more GSB. It's been a while, and time has been short...


Is it worth explaining more about adaboost? Limiting the amount of contracts per day is likely to decrease a lot of metrics good and bad. Ie I expect pf, at to reduce, but also drawdown


Peter, here is a white paper I came across yesterday that details trading a methodology with Adaboost using Logitboost algorithm. They describe a method of automated trading with boosting and expert weighting. Very interesting read. Unfortunately some of it went over my head but I got
the gist of what they were doing.

I really like the methodologies they used. It relies on a layered structure consisting of a machine learning algorithm, an online learning utility, and a risk management overlay.

I don't have the skillset to understand how to put together something similar to what they had done. I can imagine the possibilites with something as they described.

Great read. Any chance their methodologies make their way into GSB?

Attachment: Login to view the details

Attachment: Login to view the details

admin - 3-1-2019 at 02:18 AM

The maths is at a higher level than mine, but the lead programmer is really good at maths. I will pass it on.