JSystemTrader (Java/IB)

Discussion in 'Automated Trading' started by nonlinear5, Sep 26, 2006.

  1. What about Reuters?
    #31     Nov 22, 2006
  2. squeeze


    Reuters is fine but RMDS via Radianz costs a lot of money.
    #32     Nov 22, 2006
  3. You're perhaps making assumptions with respect to time frames used for the ATS. A strategy that operates on EOD data doesn't require a $2-3K/month data feed.

    Furthermore, one can make the assertion that if the data feed is good enough for manual trading, then it is good enough for replicating the manual trading strategy in an ATS as long as there are enough mechanisms to deal with faults and failures.

    There are hundreds of ATS that operate successfully with the IB data feed and broker interfaces for example.

    Granted, it would be nice to have the best data and reliability available and certain trading systems will simply not function without that but there are many systems that can operate with much less onerous requirements.

    Lastly, there seems to be some merit in prototyping systems with lesser data feeds and having a framework or infrastructure to do so rapidly. At least, that is the intent of my development.

    Do you consider data feeds from the following providers to be up to scratch:

    #33     Nov 22, 2006
  4. squeeze


    A strategy that runs on end of daya data doesn't need an ATS at all. It can be traded easily by hand.

    Feeds for screen based trading and automated trading have quite different requirements. It is not realistic to replicate all of the intelligence and filtering that a trader using a screen applies when entering orders. ATS's are basically stupid and need to be fed with good quality reliable data for them to give reliable results.

    I have not used feeds from Lime or Redsky so cannot comment but both have very good reputations and I would expect using a point-point connection to their trading servers would provide a professional trading set-up.
    #34     Nov 22, 2006
  5. If the ATS was handling a large numbr of stocks ( say > 20), it might be much better to automate.
    #35     Nov 22, 2006
  6. squeeze


    You could probably just use a basket order to gets fills on a lot of stocks.
    #36     Nov 22, 2006
  7. Yeah, I was going to say if it operates on 500 symbols then doing so by hand is not necessarily feasible!

    No matter. Squeeze has made his point with respect to data feed quality requirements. I get the impression that he would not under any circumstances undertake building an ATS with any retail data feeds.
    #37     Nov 22, 2006
  8. squeeze


    I have been running multi-strategy, multi-market ATS platform for over two years. If there is one thing I have learnt it is that a good datafeed is vital. Depending what you are doing, you can cut corners in the infrastructure used to send out orders but having absolutely reliable and accurate data is just a requirement. At the moment I just cannot see any retail feeds that really do the job.

    For example how many retail feed providers would be willing to guarantee you 99.999% or better feed resiliency?
    How many are quoting figures on latency?
    #38     Nov 22, 2006
  9. So what is it that you consider most important ?

    - accuracy

    - latency

    - availability (up time)

    And what do you consider the biggest failing of feeds such as IB ?
    #39     Nov 22, 2006
  10. squeeze


    In this order

    A feed that that isn't totally reliable is no use for automated trading.

    Accuracy is very important but there is scope for implementing filtering to mitigate some of the problems of bad data.

    Latency issues can be side-stepped by placing contingent orders at the exchange or with your broker.
    However, it becomes a much bigger issue when trading in large size or running arb strats.
    #40     Nov 22, 2006