Seeking Guidance for Low Frequency Trading

Discussion in 'Automated Trading' started by xtermz, Mar 21, 2013.

  1. xtermz


    Hello all,

    I've been manually trading stocks using daily/weekly bars for some years now and am looking into creating a blackbox trading system for some low frequency intraday trading (perhaps up to 10 trades in a day for a given instrument). I'm not trying to compete with HFT and would like to leave a generous margin for latency and slippage.

    I would like to take a very scientific approach to this as I research and come up with strategies to test. I have some questions:

    1) What is the ideal data resolution for developing a system like this? Do I really need tick data or is minute data enough? What are the pros/cons of one over the other?

    2) Based on (1) above, what is a reasonably priced historical data source and how many years of historical data is generally enough to thoroughly back-test a strategy on this scale?

    3) Should I stick with stocks or switch to futures (or both)? I would think futures would be easier to manage for tax purposes.

    Thank you!
  2. dom993


    IMO, 10 trades / day in a single instrument is *NOT* low frequency, especially if you are talking of a single setup - but if you are thinking about multiple setups, then your first step should be to consider each setup on its individual merits, not mix them in a single strategy.

    1) ideal resolution - there is no such thing, both have their pros & cons. It ultimately depends on what resolution you need to model the market. I am using primarily tick-data, even though my systems are (intraday) swing trading systems (up to a couple trades per day, on average). This allows me to be flexible in my definition of the patterns I am interested in (by flexible, I mean they don't have to be aligned with 1min bar boundaries). But many successful traders just use minute-charts - to each his own :)

    2) Use TickData - they are simply the best, at least for retail traders. For stock-index futures, I would suggest 10 years, for commodities futures, you might want to check when a specific product started trading electronically 24/7, the data before that doesn't have accurate volume information, and 1min data is the best you'll get from pit-sessions.

    Several thousand trades in backtesting over 10 years (on ONE market) will give you a reasonable starting point. Note that if your strategy averages 1 trade per day, you'll have 2500 trades in 10 years, so that level shouldn't be difficult to reach for intraday trading.

    3) Given all the crap with internalization / resell of order-flow / sub-pennying & darkpools / HFT, I doubt the stock market is a fair market these days, and for myself I stay away from it and just trade futures.

    Good luck in this endeavour, this is much more difficult that you can ever guess from the outside.

    I think the next poster should be EMG :)
  3. NoBias


    You are being trolled...
  4. gmst


    Thanks for all the research.

    Do you have a working hypothesis why anyone will go about creating new ids on any anonymous forums and asking legitimate looking questions?

    I mean what can such a person gain from such an exercise? Why do it in the first place? What kind of psychological reason can lead someone down this path? Any ideas...
  5. eurusdzn


    Regardless, Dom993 is always a good read......not a bad inquiry either.
  6. Sergio77


    You can't avoid that even if you do one trade per 10 years. HFT is your modern market maker .
  7. xtermz