Fully automated futures trading

[tokei]

I used to live in Japan for a couple of years, and speak a bit of Japanese. Which is why I recognized your handle name.


Your CL example makes sense: there was a relative large price spike recently, so the volatility of the instrument went up. And, due to that, the maximum position size went down. In your case this made it cross the hard limit that your software had set. You will experience this many more times: your software detects that a certain border is crossed and takes action. Whereas you would have left things unchanged based on your own "intuition", or "gut feeling".
I do know about that thresholding system of 1 contract. But I feel that that level is too low and the software can not properly adapt the position size to a change in forecast. Rob once wrote a blog entry where he showed the dependency of SR on the maximum position size. It is this post which made me decide that I prefer to only trade instruments where I can have 3 or more contracts at maximum forecast.

Edit: I found that blog entry. https://qoppac.blogspot.com/2016/03/diversification-and-small-account-size.html
Halfway, in the section titled "Putting it together" there is a graph. My approach is based on what I see there.

Thanks, HobbyTrading. I hope you enjoyed living in Japan. Japan is a good country to live for Japanese for sure, but not sure for non-Japanese. Firstly language is a problem, and also many complicated procedures in government related areas. It is flustrating that this country is under depression for 30 years. However I hear living in countries of depression, while earning money from growing countries, is a nice choise. I am not sure of it but that is where I am.

My learning from you is that what matters is to run the system you believe in. I will try to keep up!

 

[globalarbtrader]

.... I'm too lazy to do all this, so I decided to leave it out.

Someone after my own heart... good a reason as any to do anything.

GAT

 

[globalarbtrader]

right now I am blindly following Rob's method....

Please don't blindly follow anybody, even me!

However I noticed some difficulty in dealing with this thresholding methodology. Some instruments cross the boundry between "with-thresholding" and "without-thresholding", as their vol changes. For example yesterday my CL does not require thresholding, as the max position size was greather than 2, but suddenly today my CL requires thresholding, as its max position has now get a little smaller than 2 (1.998 or something). And what is worse is that my system was telling me yesterday to hold 1 contract for CL, but today it tells me to sell the 1 contract, even though the signal has not weaked. The system is right, because the CL today requires stronger signal to hold 1 contract, because it falls under the category of "with thresholding". Same happens for boundary between "with-thresholding" and "not tradable anymore even with thresholding".

Not sure if I made myself clear above... but the point I guess is that I am facing difficulties (major or minor) I need to find solution somehow, and I find myself enjoying the process!

To be clear: I don't exactly do this; I don't actually dynamically change the definition of whether something does or does not require thresholding, they have remained fixed for several years.

As a result my thresholding would probably require an occasional check to see whether instruments are classified correctly; but I plan to replace thresholding with a more elegant process at some point anyway.

GAT

 

[globalarbtrader]

Marsten been around for so long, and thru very different market types, that it is hard to discount his bottomline. I personally known him since around 2013 or 2014, and never seen anyone so honest and down to earth. Quite often, people are better at Doing rather than explaining how to do it to others. He is kinda like that. He has a great feel for what is outside of expected parameters and when a change to a system is actually needed. It certainly evolved over time from what you are describing, and he doesn't change his systems as often as he use to.

You can count me as a descendent of his method and I write about my trading in a satellite journal to Rob's - Fully Automated Stocks Trading. That might give more details for those who are interested.

Btw, really appreciate all work Rob has done for the community. Incredible stuff. Math you guys are talking about here is above my pay grade so I just keep doing what works for me

Val

Val's journal is here https://www.elitetrader.com/et/threads/fully-automated-stocks-trading.346201/

GAT

 

[tokei]

Please don't blindly follow anybody, even me!

Hi GAT! Yes I am on my way to gain eyesight. It needs to be done before it's too late...

To be clear: I don't exactly do this; I don't actually dynamically change the definition of whether something does or does not require thresholding, they have remained fixed for several years.

How come I cannot come up with such a simple solution! It is somewhat similar to your answer for my question on how to pick trading rule variations to apply for each instruments, of which volatility changes over time. I will do so starting now. Checking once a while is much easier comparing to checking every day and adjust the the risk weights manually (this is what I have been doing lately...pretty annoying). Thanks!

 

[tokei]

Edit: I found that blog entry. https://qoppac.blogspot.com/2016/03/diversification-and-small-account-size.html
Halfway, in the section titled "Putting it together" there is a graph. My approach is based on what I see there.

Thanks HobbyTrading. I did read this blog post. My takeaway was that if you have 2 choices, A) instrument with max contracts of 3 or bigger, B) a instrument with max contracts of 2, and a new instrument with max contracts of 1, go with B at any time. The 3 curves looks pretty proportional regardless of x.

 

[ValeryN]

I'm sure all the statistical arbitrage (I presume? Mix of interday and intraday) stuff you are doing is not trivial either - mathematically, programming wise.

No stat arb. ~80% mean reversion with a tiny bit of TF for diversity and better combined stats.

There are several systems each targeting their own setups in stocks, both sides - long and short. Math is so simple that I don't even perceive it as math. It is relatively complex on the execution side though. Both me and Marsten independently created our own backtesting and execution pieces. I eventually swapped mine with his for backtesting but keep mine for execution. It is just under 10k lines of code in Kotlin. Most of my complexity comes from massaging orders to get good executions, error handling associated with that and some self-healing routines as it runs on a server and I don't typically touch it.

Val

 

[tokei]

Hi all,

A thought somewhat related to market diversification. What would be the downside of keeping more instruments in your portfolio beyond your risk target? You will be able to pick trending market effectively from broarder market, particularly when your risk is not fully used up by your position. You just need to set the rule not to place order that exceeds your risk limit... does this sound like a bad idea? Or is this something you all do already??

 

[globalarbtrader]

Hi all,

A thought somewhat related to market diversification. What would be the downside of keeping more instruments in your portfolio beyond your risk target? You will be able to pick trending market effectively from broarder market, particularly when your risk is not fully used up by your position. You just need to set the rule not to place order that exceeds your risk limit... does this sound like a bad idea? Or is this something you all do already??

It's something I am considering. It's more complicated than it sounds, but may still be worth doing.

GAT

 

[tokei]

It's something I am considering. It's more complicated than it sounds, but may still be worth doing.

GAT

Hi GAT! Oh I see. I put this in the very end of my long to do list. Priority for me is to learn coding so that I will be able to run backtest with the very basic strategy. Thanks!