What does it take to build an automated trading system?

Discussion in 'Automated Trading' started by dpeck, Jan 25, 2006.

  1. dpeck


    Ever wonder what it takes to be a “professional” trading system designer? Here is an article written for Dice.com by a system designer. For those of you who are not familiar with Dice.com they are a job board, similar to Monster.com.


    I am posting this article as an informative resource for the new/non programmers. My intention is not to discourage anyone but rather inform the group. Every programmer has to start somewhere. For those of you who do take the plunge and learn programming you can be comfortable in the knowledge that if trading does not work out you can always get a job coding, which by the way pays very well.

    I do believe that an automated system can be created with much less experience than he describes but it certainly will be more difficult and less robust. As always I am curious to see how people react.

    Do you think he is correct?
    Have you actually built a fully automated system, NOT in TradeStation or others, with less then 50,000 lines of code?
    If you use TradeStation or others like that have you been able to successfully fully automate you programs?
  2. You can build a system in about 100 lines of code. Automating to a broker, connecting to an api would be where most work is done, but you could hire an expert in api's for that.
  3. Choad


    It is damn near trivial to build an automated trading system that will connect to IB's API, safely make trades, and is profitable.

    Or at least, it is after you've done one and have lots of experience in the market...
  4. It also depends on what kind of automated system are we talking about. For instance, one can code up a fairly simple system that pulls data down from Yahoo, and buys stocks based on P/E ratios (I know, I know, but it *is* an automated system), say, buy the 3 stocks with lowest P/E while have average dividend payout, I believe such a system can be coded up in under 50k lines.

    But if we are talking about sophisticated automated systems (say one that does high frequency statistical pairs-trading), then what the article describes are basic knowledge necessary. It is somewhat like ability to read financial statements (and understand what a debt ratio is) is basic to an equity analyst, but just be able to read quarterly income statement doesn't mean that one would be a good research analyst. In fact, complex computation and performance related stuff in sophisticated automated trading systems goes quite a bit beyond what the article describes.

    While people have described ATS developered as well-paid (typically start at 120-150k base + typically 40-100% in bonus), which is a lot of money to 99% of "normal people". But once a technologist gets in that situation, and seeing that the quant strategist sitting next to you get a 500-700k package, one couldn't help but feel underpaid. I did exactly that, I was a sr tech mgr at an ibank, before saying to myself "hey, why don't I make 50-60% more over on the business side?", so I did.
  5. Thats true. I would never assume that a systems designer makes big big money, its all in the execution, manager, trader, dealmaker, these guys pay designers to sit down for 6 months and code the firms strategies. The only real secret is to be at a place where they code strategies and trade them. Some places are too conservative for anything more than arbitrage.
  6. Problem is that cross over between trader and developer is getting blurred. Since it is beneficial for a prop desk or a hedge fund for as few people to have access to the code base as possible so that the strategy doesn't leak out too much, so a lot of the quant traders are coding the system directly without hiring a developer. A lot of the ibank prop traders and quant hedge fund people I know came out of academia anyways where they are used to rollup their own sleeves, so they are used to coding. In these cases, the technologist (if there is one) is virtually doing the same thing as the traders, hence they are being paid out of the desk p&l, hence also the feeling of "underpaid". These are especially true in an automated environment, since it is really the system that is doing the trading.

    For instance, I know a guy who got his phd last year, and got hired at ibank prop desk, the desk heads won't even let him *see* the code base until 4+ months later. So for 4 months he literally coded up simple little prototypes to show the desk heads that he has the ability. Those 2 desk heads coded up the system by themselves over 18 months or so.