Automated Trading for a Living

ofthomas · May 4, 2013

Quote from gmst:

How does the idea of taking a year off and going back to university to pursue an M.S. in computer science sounds for a person like me? Post this MS, I have no intention of joining a tech role, rather I would like to use this tech knowledge to design my own stuff and implement my infrastructure for systematic trading. Idea is to have enough skills/capital to develop into a systematic trading firm in 5-10 years time. Does this make sense?

This is a general question - anyone feel free to answer.

save your money, just go to MIT/YALE/UC/Princeton/Harvard, etc... sites and you will find most of the classes online and for free in video format, just get the books and that is it...

to be honest, what you are referring to will not come from a MsC, you will gain the knowledge only via the experience of creating, failing and succeeding, so if you are not going to get the job, just take the classes without paying, same thing...

you will also not be able to put into practice what you learn at an enterprise scale, but it can easily be done at a home level... most of the things can be found at stores or online if you have the budget and are not focused on having the latest and greatest...

personally, I dont have any intention of ever creating a trading firm, etc... I trade on the side for my own account to create capital to diversify... yes, eventually I wouldnt mind not having to do IT work, but at the same time, I get bored quickly, and trading is boring... so i need IT work to keep me entertained... last thing I would need is to churn and trade to keep entertained... so I cant tell you if the skills you seek will help you at all... as I see it, you surround yourself with those that know and it becomes a team effort... like a pack of wolfes hunting the markets...

so, those are my 2 cents, for whatever is worth...

Allistah · May 4, 2013

Quote from QuantWizard:

As written in another post, Tradestation works for simple indicators, however these are all virtually obsolete these days. You also need more control to test thoroughly, which can only be done in a real programming language. The sooner you get past that learning curve, the better.

So are you saying that nobody can make simple indicators like a moving average into a profitable system? You have to have some unknown to the world custom indicator to make any money?

buzzie77 · May 4, 2013

Two sites that are growing on me for automated trading. ET is good for chit chat, but when I want real questions answers I post at the following.

How to secure your automated strategy?
http://investfeed.com/question2answer/index.php?qa=182

http://quant.stackexchange.com/questions/4238/how-to-calculate-equally-weighted-market-portfolio

Allistah · May 4, 2013

Quote from buzzie77:

Two sites that are growing on me for automated trading. ET is good for chit chat, but when I want real questions answers I post at the following.

How to secure your automated strategy?
http://investfeed.com/question2answer/index.php?qa=182

http://quant.stackexchange.com/questions/4238/how-to-calculate-equally-weighted-market-portfolio

Awesome, thanks for the tips. I'll take a look at those sites.

Mike805 · May 4, 2013

NetTecture,

While I appreciate your IT know-how, your posts are heavy on tech-speak and very light on actual trading know-how.

Put another way; what good is a really-super-awesome backtesting automated execution engine if your core strategy logic can't make a dime because its rules are over-optimized junk?

You're hinting at Genetic optimization??? Why even go there, you're building in a very heavy historical bias every-time you do an optimization.

Some of the most consistent profitable automated strategies I've come across have very few lines of code, need no optimization, and can run with any retail level platform. Markets aren't complicated... why make this so complicated?

None of what you describe is necessary to making money. What is necessary is a good understanding of fundamental price behavior first, and, some basic programming skills second.

Mike

Allistah · May 4, 2013

Quote from Mike805:

NetTecture,

While I appreciate your IT know-how, your posts are heavy on tech-speak and very light on actual trading know-how.

Put another way; what good is a really-super-awesome backtesting automated execution engine if your core strategy logic can't make a dime because its rules are over-optimized junk?

You're hinting at Genetic optimization??? Why even go there, you're building in a very heavy historical bias every-time you do an optimization.

Some of the most consistent profitable automated strategies I've come across have very few lines of code, need no optimization, and can run with any retail level platform. Markets aren't complicated... why make this so complicated?

None of what you describe is necessary to making money. What is necessary is a good understanding of fundamental price behavior first, and, some basic programming skills second.

Mike

Thanks for the reply Mike. Now if I understand what you've said is that there are plenty of strategies that don't need any indicators at all or maybe one. I guess that just leaves price action.

I've written a fee strategies in the past that were only based on price action. For example, I would look for patterns that seem to repeat and then program the code to look for that exact pattern. Nothing to optimize. For example, look for 5 down bars and one up bar or something. Is this the sort of thing you're referring to?

vicirek · May 4, 2013

Quote from Allistah:

For example, I would look for patterns that seem to repeat and then program the code to look for that exact pattern. Nothing to optimize. For example, look for 5 down bars and one up bar or something. Is this the sort of thing you're referring to?

dr. Evil it has been already done!

Markets are being actively randomized on top of normal market noise just to prevent any pattern formation giving consistent results. If you run this strategy I would suspect that it will give you 10-20% success ratio at most which is not sufficient to make money in automated fashion.

Likewise I do not understand testing of terabytes of data when statistics would prove that much smaller sample is sufficient to reproduce market noise. Signals in the market are rare and elusive.

This also applies to idea of seeking perfect back testing data when in reality it is just a noisy set of the same data no matter how clean they are.

If you have never traded then try small sizes to get the idea what trading is in psychological and technical terms. At the same time you can test your ideas for fully automated or assisted trading using commercial platforms but do not expect much or any breakthrough using them. I would rather invest time to learn and implement your own simple system using mainstream programming platform. Once you find your niche then you have the tools ready to make progress.

gmst · May 6, 2013

Quote from ofthomas:

save your money, just go to MIT/YALE/UC/Princeton/Harvard, etc... sites and you will find most of the classes online and for free in video format, just get the books and that is it...

to be honest, what you are referring to will not come from a MsC, you will gain the knowledge only via the experience of creating, failing and succeeding, so if you are not going to get the job, just take the classes without paying, same thing...

Thanks for your perspective, makes sense. If I can get all the lectures of MIT MS in CS online, then there is very little need for me to go physically to the campus to get a degree. I can watch and understand all the lectures on my own - just needs dedication. I like to ask lots of questions to the faculty - I think that is the only thing I will miss by being not in a classroom.

Quote from ofthomas:

you will also not be able to put into practice what you learn at an enterprise scale, but it can easily be done at a home level... most of the things can be found at stores or online if you have the budget and are not focused on having the latest and greatest...

Why do you say so it is not possible to implement it on enterprise scale, given say 1 million dollars to spend?

gmst · May 6, 2013

So, I looked at electrical engineering & computer science page at mit opencourseware website. It lists all the courses available on ocw. Link:

http://ocw.mit.edu/courses/electrical-engineering-and-computer-science/

My aim is following:
1) Substantially increase my practical and theoretical understanding of computer science, programming languages, parallelism, architectures, databases etc.
2) Be able to participate in technical discussions with MS holders in CS, HFT practitioners
3) Technically, be in a position to design and implement infrastructure for a systematic (high frequency) firm in 1-2 years.

I expect to be able to devote 20-25 hours per week to the reading and taking classes.

Does this approach of taking ocw classes to fulfill above 3 goals sound reasonable/valid? Or, Am I setting myself on a path which will lead to frustration and I will not achieve my 3 goals above?

My background is: Graduate technical degree from an Ivy

gmst · May 6, 2013