experienced prop trader details his custom Python algo trading setup

Jack_Larkin

Jack_Larkin

From: https://fxgears.com/index.php?threads/python-development-environment-jacks-technology-stack.1090/

Snippets:

My primary language is Python. Despite some uninformed beliefs that Python is too slow for algo trading, and that algorithmic trading is best left to C/C++ or some hardware programmed FPGAs, Python is perfectly suitable and more than fast enough for any retail trader who wants to get into algorithmic trading.
More...

Despite what some believe, you don't need very beefy hardware to trade or algo trade. A few of my best and most profitable trades have been done on an old netbook with a 10", low resolution, screen. People who invest into expensive hardware, computer setups, etc.. before they even have a prof of concept hammered out (be it a trade plan, or algo strategy,) are putting the 'cart before the horse.'
More...

Algos are good for when you don’t have enough hands, eyes, or aren’t fast enough to do something manually. If you already can do something efficiently on your own, then turning it into an algo will only make performance worse because I can’t encapsulate your years of experience into lines of code. Also, if you ask me to code up a moving average crossover strategy, we aren’t friends anymore.
More...

Everything is tied together with Python, ZeroMQ, and open source libraries. Even uses free data feeds like IEX.

Thoughts? Anyone else take the same approach?
 
fan27 said:
Interesting work! What advantages do you feel you have with your current configuration vs just using NinjaTrader or MetaTrader?
More...

Platform agnostic; nothing is dependent on a given broker or platform's proprietary software. When porting to a new broker, I just need to write an interface to said broker and all strategies previously coded up work as expected. Sometimes a desired order type might be different between brokers/platforms, but that's just a settings thing, not code dependent.
 
sle said:
How do you integrate back-testing with the live trading (e.g. for perfomance attribution)?
More...

Three approaches:

1) 3rd party tools vet the idea first (I've used BuildAlpha for example.)

2) Anything based on the order book, or relates to things that would be difficult to test -- mid-point / dark orders, or sweeping a level where your trade would materially change the book -- I might take straight to live walk-forward testing and only scale up after spending some time here. In the last two+ years I'm finding that most single name equity strategies have to fall into this category because their average trade size is so little and the book reacts to even a few shares going off so even small size can start skewing strategy results.

3) Since the agents connecting my platform to NT8 and MT5 are chart indicators and strategy files (in contrast to plug-ins,) then running either platform's back tester (or market replay modes) will effectively let my platform do trade logic testing and back testing.. for simple time series analysis this is fine, but it's a little slow compared to what some engines can do just traversing massive tables of data.

95%+ of what I do falls into the 2nd category. Theory > Implementation > Measure > Adjust > measure > adjust > repeat...
 
ZeroMQ is great. I use it to send out the buy signals from a commercial trading software package and then my own trade manager takes care of the rest. Reason for this is that I may want to take the same futures trade in multiple accounts with possibly different sizes, one usually being an IRA account.
 
qlai said:
You run co-located? Do you build book or get from provider?
More...
Not co-located as that hasn't proven to be required, yet.

Book feeds depend on the source. I have one source that is a composite book of a fractured market by one provider, and another method that takes books from different sources and creates my own composite book. So it depends on the intended use.
 
You must log in or register to reply here.
Back
Top