Customized automated trading software

Discussion in 'Automated Trading' started by ESolutions, Sep 5, 2009.

  1. Hi everybody. I'm interested what kind of trading software would be a big improvement comparing to actual solutions.

    Let's say there is a software built up from the scratch, solid and sophisticated engineering, fast and very stable. The program has a datafeed (Fix and Fast), certified by a big market gateway, a connection to a broker and a extensive monitoring infrastructure for simulations on historical data and backtesting of strategies. The strategies are written in mathlab, and the software can backtest and trade them.

    This is a very basic setup and works for example for daytrading S&P 500. But given the idea of a totally modular built of the software, basically every marketgateway and broker connections are possible, also the change from mathlab to C or Excel.

    As a person with only very limited knowledge of trading, I'm interested what kind of software features people in this business are interested in. What would be a big technical challenge, what would be very interesting but doesn't work yet because of problems with connectivity or other issues?

    At last some very basic questions, but they might help me to understand the needs of traders a bit better:

    *Which market gateways are the most important?
    *How interesting is speed, f.e. real-time trading?
    *Which language? C, Pascal, R, Mathlab or Excel?
    *What kind of graphical user interface would be interesting?
    *Is an encryption of the strategies as a safety feature interesting?
  2. Eight


    I'd like to see software that worked like Tradestation's charts, had encrypted strategies [if the encryption was impossible to break and I could keep the non-encrypted versions of the strategies completely off the computer somehow] and maybe all built upon the Tickzoom core... with a backtester that always tested to the tick level so that it was impossible to access any data ahead of time and tested accurately so that backtested results always matched reality...

    Most of this stuff we have available to us has these compromises regarding how they backtest.. they are left over from the 80's when computers were slow and you had to compromise by making guesses as to whether the high or the low came first... there are all sorts of compromises and ambiguity... drop all that shit, make all tests to the tick level. One is slow in backtesting because it implements the entire fix protocol for every transaction...

    Some well written software would be a welcome thing too... some of this crapware we have available won't run without admin privileges, is buggy, is slow, some won't run in Vista natively yet after Vista is out for how long now? 4-5 years...... I found a way to crash one of them so bad I had to nearly wipe the hard drive to get going again, it required a rebuild of the .net 3.5 and complete wipe and reinstall of the software... and all I did was use an odd naming convention for a script!!

    Provide some intelligent backtesting feature like the simulated anneal tests that are available in Openquant.. a development environment that was like Openquant, with a faster backtester, based on Tickzoom core, with charts like Tradestations... that might be trader nirvana for yours truly...
  3. All the "enhancements" you mentioned are already parts of any good automated trading system today.

    Let's talk about some possible technical challenge, if that's what you are interested in. I will give you two, one very technical, one very analytical. Technical, an universal Exchange / Broker connectivity "adapter", such that FIX specifications be fed into the the "universal handler", without coding requirements, and perform periodical and on-demand automated interoperability with all brokers and exchanges? This includes any future "dialect" of FIX, since the "adapter" would take in the original XML specifications. The integration customer would have minimal coding requirements (scripting language level).

    Analytical, an automated tool that can realistically "simulate" the market (for any arbitrary exchange listed product), using historical tick data, providing simulated executions, slippage, etc, the tool will also take in actual executions, to automatically "tune" itself to be as realistic as possible (matrices such correlation, regression, need to be provided).
  4. Nice discussion here... Does anyone know a vendor like Opentick (now they are gone)? They used to provide API to pull in data and do whatever you want to it. I used their free data but haven't used their API. However, with something like that one could write an automated system from scratch...

    In short, I am looking for a data vendor that provides tick by tick data, and has open API to access it. I wish Interactive Brokers sent out all the ticks. With their open API, it would have been the ultimate...

  5. Occam


    I think NxCore is pretty good, if you don't mind paying for a feed of the whole market; minimum is several hundred $ per month.

  6. Thank you very much, I wasn't aware of them.