automation based on figures

Discussion in 'Automated Trading' started by sarcastic, Sep 8, 2008.

  1. sarcastic


    I want to automate some of my strategies based on some events and figures released. I heard about one software NTKN(news to know services) which provides figures in a digital format in ur computer and you can react faster than others on that by automation based on that data.does anybody have experience with that or any other services similar to that.if yes then please share your experience.
    1.whats the total time lag between data released and you receive the data(in milseconds if possible)
    2.what are the technical difficulties using that software(my firm is not so keen in doing that becuase they think they dont want to give any third party excess to their network)
  2. Pippi436


    Check out Reuters Newscope Realtime - they got an API for news-related algorithmic trading.
  3. You'll probably never have the edge you desire due to latency issues. But even if you were collocated in the same room as the CME or whatever, you'd still have problems. Most of these economic numbers are released in advance and embargoed until a certain time. Clearly, there are some out there who abuse this embargoed data and coordinate huge trades a fraction of a second prior to the release time. Happens all the time. It's fucking crooked. I would suspect that some have even opened news services specifically to exploit this. This is why all news releases should be done via conference call.
  4. sarcastic,

    when I was working with designing professional-use forecasting models we successfully used figures and even legislation changes as input for forecasting models. This was through the learning and back-propagation of feedforward neural networks for non-linear regression analysis on timeseries data. It was most successful on spot-market energy forecasting.

    However, you must consider the timeframe effects on including numbers. The effect is not linear at all, and there are several things to consider - such as delayed reactions.

    It is not as simple as just "throwing in input data" to model a complex system, you need a thorough model which you are reflecting. A simplistic approach to this is using sliding averages and other aggregated values. I used that for the most successful forecasting models to market data predictions. The success of the models was such that one large company saved more than average US$100 000 per day on the improved forecasts - continuously beating all their professional analysts.

    Look into systems sciences and systems theory for more clues towards modelling complex adaptive systems like financial markets.