Ask HN: How to learn about automated trading systems?
4 comments
Well, it really depends on your experience and available capital. All major exchanges these days are electronic, so you don't have to worry about which provide APIs, because it's highly unlikely that you'll working directly with them (at least until you're bankrolling some decent change). Interactive Brokers is the preferred brokerage for professional retail and small-midsized prop firms and hedge funds, specifically because they offer a decent API (cross-platform, which is rare since most non-fix providers target Windows) for free and FIX connectivity if you're doing more volume. IB gives you access to trade on virtually all large international markets with very competitive commissions. For about $10/month (fee waived if you have a few trades/month), they'll also provide an acceptable data feed. If you're wanting to focus on specific instrument types, such as only futures, there are alternatives.
There are various open source and cheap commercial platforms you can take a look at. The simple open source trading platforms are ActiveQuant (java) and TradeLink (c#). There is also MarketCetera (java), which is a professional platform, but most retailers don't find it very attractive, because they only provide a generic FIX adapter and a base market data feed adapter which needs to be extended for your specific needs (there is also no backtesting, proper data storage). There are a few commercial solutions targeted at retailers, such as NinjaTrader (free for sim-mode with free futures tick/historical data), OpenQuant, and AmiBroker. These aren't appropriate for complex high frequency strategies, but you'll get some good experience coding strategies immediately to gain some experience instead of having to tinker with the infrastructure.
All that being said, it's difficult to break into the high frequency and statarb space if you're just beginning. People always assume it's just a technology problem, but you'll soon enough find out it's not ;) So, I'd suggest that you take it a bit slower and work your way up from daily data trading to intraday to high frequency, a journey that will surely take some time and keep you challenged along the way.
I'm a CS guy by schooling and got started in the industry a few years ago. For me, trading is a very challenging, but extremely fun and rewarding. The markets are so dynamic that you'll constantly be going through the R&D cycle and you get to use all that CS/Math theory you learned, instead of hammering out generic run of the mill code like you do at a lot of jobs these days.
There are various open source and cheap commercial platforms you can take a look at. The simple open source trading platforms are ActiveQuant (java) and TradeLink (c#). There is also MarketCetera (java), which is a professional platform, but most retailers don't find it very attractive, because they only provide a generic FIX adapter and a base market data feed adapter which needs to be extended for your specific needs (there is also no backtesting, proper data storage). There are a few commercial solutions targeted at retailers, such as NinjaTrader (free for sim-mode with free futures tick/historical data), OpenQuant, and AmiBroker. These aren't appropriate for complex high frequency strategies, but you'll get some good experience coding strategies immediately to gain some experience instead of having to tinker with the infrastructure.
All that being said, it's difficult to break into the high frequency and statarb space if you're just beginning. People always assume it's just a technology problem, but you'll soon enough find out it's not ;) So, I'd suggest that you take it a bit slower and work your way up from daily data trading to intraday to high frequency, a journey that will surely take some time and keep you challenged along the way.
I'm a CS guy by schooling and got started in the industry a few years ago. For me, trading is a very challenging, but extremely fun and rewarding. The markets are so dynamic that you'll constantly be going through the R&D cycle and you get to use all that CS/Math theory you learned, instead of hammering out generic run of the mill code like you do at a lot of jobs these days.
Out of curiosity, what else is involved in HFT besides tech?
A significant amount of domain knowledge. You need a solid foundation in higher mathematics, statistics, and economics. This industry is extremely competitive and you'll quickly notice the manipulative games that are being played. If it were a simple technology game, a lot more market participants would be successful. That being said, as a retail trader, you have significant advantages, as far as capital allocation and risk tolerance. The best advice I can give you, look for edges were it would be impractical for the big boys to do so. Out of curiosity, what do you think is actually involved in quantitative trading? How do you think models are created and applied?
If you want to chat a bit more, let me know.
If you want to chat a bit more, let me know.
Hi, I'm actually the OP -- sorry for taking so long to reply, but my laptop went KABOOM a little whlie back, taking my saved hn password with it. I'd love to chat more if you've the time, my AIM is thelemonkiller -- if you prefer another means of communication, let me know!
One of the big bottlenecks here is access to the low latency network connections & data streams. Successful HFT systems integrate information from ultra low latency news wires, trading activity in microsecond timescales, and are constantly testing things like dark pools for private market pricing information which isn't publicly accessible. Even if you'd written the ultimate trading app, you still wouldn't have that kind of access -- which takes millions in infrastructure and business partnerships you don't have, and no consumer brokerage would give away practically no matter how much you wanted to pay for it -- making the game unwinnable for anyone but institutions and highly specialized proprietary trading firms.
Foyle's bookstore in London has a financial section larger than a lot of bookstores, including a wide variety of books on statistical modelling for financial instruments.
In terms of keeping up with the community, zerohedge.com provides valuable advice and recently provided the full PDF of 'an introduction to high-frequency finance'.
In terms of low latency architecture and platforms, check the docs for Red Hat's MRG, SuSE's SLIRT, and highscalability.com. You'll find most of your Unix knowledge is tied towards performance rather than determinism, and you'll need to re-learn a lot of your knowledge and delve deeper into strace and systemtap.
In terms of keeping up with the community, zerohedge.com provides valuable advice and recently provided the full PDF of 'an introduction to high-frequency finance'.
In terms of low latency architecture and platforms, check the docs for Red Hat's MRG, SuSE's SLIRT, and highscalability.com. You'll find most of your Unix knowledge is tied towards performance rather than determinism, and you'll need to re-learn a lot of your knowledge and delve deeper into strace and systemtap.
Ameritrade has an api.
I've also been looking at the BATS exchange: http://www.batstrading.com/home/ . I don't know how feasible it is to get an account with them. http://www.batstrading.com/membership/
I've also been looking at the BATS exchange: http://www.batstrading.com/home/ . I don't know how feasible it is to get an account with them. http://www.batstrading.com/membership/
Thanks!