Unix base

[Kevin21]

Hey,
a bit of a ramdom question here. I don't suppose anyone has designed and programmed an automated system on a unix based OS? If do you know of any unix based tools that would be helpful?

I'm pretty confident with one of my systems, so as side project i'm going to fully automate it on my IB paper account, but i want to base it around a unix system for stabilty and security. At the moment tho it looks like i'm literally gonna have to build up a base from scratch in C++. Last night i was even thinking to myself that in a hypothetical situation on another project i would like to build a system around the FIX protocol lol

Last year when i was still at uni i was building a multi dimensional econometrics model, and one of the lads that i worked on the project with had previously worked on making an automated VWAP (volume weighted average price) system with his old man. Since having multiple conversations with him about it last year, the idea has always been floating around in my head to build a fully automated system from scratch. It shouldn't be hard to do to be honest, just the fact that i want to to ditch windows for this system means that i have to ditch a few tools in my tool box, and my unix skills are a bit rusty lol.

Also has anyone ever looked into algorithmic trading? I know its beyond the scope of my trading activity, but due to my geeky side i'm interesred in algorithms, and as another side project i wouldn't mind building a black box system in a simulated enviroment...

Anyone have any thoughts?

Kev.

 

[options]

Have a word with grey one when he gets back of his hols in September. He auto trades and is seriously good at what he does. Very helpfull guy as well.

 

[dcraig1]

I'm some way down the track. I'd go for Java - you will get there sooner. Anybody that says it's too slow is mostly talking nonsense. No doubt there are some notable exceptions to this.

For a free FIX implementation you can go for quickfix.

There is a free ESP/CEP package called Esper, I think, that might be worth looking at. If I was starting over, I would have a decent look at this stuff.

Get your event processing model right at the start, or you will have eternal grief. Develop a backtesting/modelling "engine" that is event driven - bars, ticks, L2 events etc. The same code can then drive a live system.

And it will take much longer than you think.

 

[Kevin21]

I'm some way down the track. I'd go for Java - you will get there sooner. Anybody that says it's too slow is mostly talking nonsense. No doubt there are some notable exceptions to this.

For a free FIX implementation you can go for quickfix.

And it will take much longer than you think.

Don't worry i ain't expecting to get it done before i go to the pub lol... I estimate it will take me a min of 8months to make some real progress with it, with the time i have to spend on it.

Thanks options, i'll look out for him

 

[dcraig1]

Some possibly usefull libraries:

http://www.ta-lib.org (several languages)

For Java:

jfreechart - charts
jgap - genetic algorithms
joone - neural nets
jdom - xml parser

 

[Kevin21]

thanks, that is very helpful