Bitcoin Forum

Bitcoin => Project Development => Topic started by: OJsimpson on October 29, 2013, 12:24:32 PM



Title: Developing an opensource algorithmic trading system
Post by: OJsimpson on October 29, 2013, 12:24:32 PM
I haven't been online in a while, I've been quite busy with work for the past few months.

I'm looking for somebody located in either Hong Kong, Japan, Korea or here in Australia to partner up with.

My current system is coded in C++ and matlab, but I'm looking to port it over to Java and R for portability.
In a nutshell, what the systems does is it scans the market for mean reversion opportunities and co-integrated pairs, in the past 3 months, I've managed to make a 42% profit so it definitely works.

The system would let the user input their own rules for making trades and allow either semi-automated or completely automated trading.
Technical analysis tools (Linear regression && Bollinger bands seem to be the most reliable in the crypto market, but will include all the most common indicators)

A little about myself:

I'm 37 living in Melbourne Australia, I work at a small boutique investment firm with locations across Asia. I specialize in financial modeling, but develop trading systems for personal use. I travel quite a lot, so if you're interested maybe we can meet up to discuss things in more detail sometime soon.

You:

Preferably somebody who works in finance or at a hedge fund or with relevant work experience. Proficiency in Java and R.

I'm looking to raise money eventually to fund the project, so all of the money raised would go towards development costs.


Title: Re: Developing an opensource algorithmic trading system
Post by: daybyter on October 31, 2013, 05:40:58 PM
I do trading stuff in java, too. Maybe we can share some code, or so?

But I'm aiming at using drools for the strategies. I can already evaluate some rule sets in drools, but I need a compiler for a trading language, so I can just just describe my strategies as text files, load and execute them.

At the moment, I still have to hardcode my strategies, but I have some profitable bots already...



Title: Re: Developing an opensource algorithmic trading system
Post by: TheNoveauExpat on November 03, 2013, 06:39:18 AM
I've been thinking of doing trading stuff, too, though I'm fairly new to the subject area.

Out of curiosity, have you tried your Matlab code in Octave? Does it actually use some libraries only in Matlab?

I ask because I recently did the Computational Investing (Part 1) (https://www.coursera.org/course/compinvesting1) course on Coursera where they use Python/Numpy/QSTK and I know a number of people from there moved on to the Machine Learning course which uses Octave/Matlab (either works, code written for one often works without changes in the other) and some are implementing that code in Python. I would expect there may be more contributors to an open source project that was written in a relatively more accessible language like one of the ones those courses cover.


Title: Re: Developing an opensource algorithmic trading system
Post by: pulsecat on November 03, 2013, 07:13:47 AM
If you consider investing in existing business, please have a look at my project (https://cryptotrader.org). It is based on very flexible Coffescript language and provides support for 130+ trading indicators.


Title: Re: Developing an opensource algorithmic trading system
Post by: sidhujag on November 03, 2013, 07:16:51 AM
Mt4 is the best charting platform and i wouldnt trade via mt4 but algo scripts running with mql are pretty easy to do. Ive coded a few of my own strategies for forex. A trade bridge can be connected via dde thru excel if you dont want to use mt4 backend to place orders.

Infact i know someone who did exactly what you are talking about but mire complicated like creating synthetic hedges and pairs.. Ofcourse 42% profit is good but what was max dd time of trades on avg? He used r then created indicators and eas in mql...