Going Automated

Discussion in 'Automated Trading' started by naifwonder, Aug 22, 2007.

  1. How would you determine, if a model can break down at any moment in the future? We all dont know the future, so..

    Why do you think pattern based models do just that? Is this some sort of theoretical reasoning, or the results of the patterns you tested?
     
    #11     Aug 23, 2007
  2. The software that I use has a special filter for blocking out bad prints. It works very well.
     
    #12     Aug 23, 2007
  3. maxpi

    maxpi

    People use the confirming tick rule. They don't get in until their signal is good for two ticks in a row.. very simple solution. Filtering bad ticks out of an indicator is similar, you have to write the tick filter into the indicator. Those solutions are much better than preprocessing data and trying to determine what is a bad tick and what is not..
     
    #13     Aug 23, 2007
  4. lindq

    lindq

    In my experience, you left out one of the most important potential problems: data spikes. Anyone who has closely monitored intraday data on any instrument, for any length of time, will tell you that while your data may be 99% reliable, there are often errant trades, or simply bad data that can come through your system. This can cause havoc with a completely automated system, and put you into or out of trades that are completely unexpected and out of sample.

    Better to let your system generate the trades with a manual confirmation on your part.

    My other feeling on the subject is that by cutting out your experience and "intuition", you're removing the most valuable mechanism involved in the trading process.

    IMO, a purely mechanical system is a crutch for someone who has not yet gained the experience to trust their intuition and overcome their fear and greed. It is a decent starting point, but it should be seen as a bridge to the time when you can feel comfortable making your own decisions.

    Let your computer present you with the option. But your intelligence and experience should be of value as to when to pull the trigger.
     
    #14     Aug 23, 2007
  5. I have tried this approach and have found it to be far to vulnerable to emotions. Often times I would ride out a loss to my stop loss and take early profits on the winning trades. In the long run, I would imagine this to be a hindrance to overall performance. As I see it, my system contains my knowledge of the market. What I know, it knows (well, almost). Therefore, any judgment made by me which contradicts what the system is telling me in real-time is most likely based on emotion rather then quantified logic.

    Regardless of all this, I have been employing your suggestion of having the system tell me the trade and awaiting my confirmation. I still end up following the system exactly with the only real benefit being that I can sometimes get a 1 to 2 tick better fill then the system by reading the print. Even that opportunity does not always present itself.

    However, college reopens next week and my class schedule overlaps with market hours. Added to that is the fact that my system trades after-hours as well and being confined to my house 6 days a week (e-mini market open on sundays as well) is not the kind of lifestyle that I want and I am willing to pay a "1 to 2 tick premium" for this. Those are the primary reasons why I want the system to be fully automated without any intervention on my part (except for freak errors/technical issues).
     
    #15     Aug 23, 2007
  6. DonKee

    DonKee

    I would be very cautious about using e signal or trade station into the ninja.

    I am testing this out with trade station right now. At least once per day the TS gets filled because of a price touch when the ninja has no execution because a buy order hit the bid, but never went below it, etc.

    I am trying my 10 tick scalping programs on the er and es and with this volatility it is not unusual to get filled on the exit, but not the entry or have a stop cancelled by TS thinking the trade was completed.

    You would be much better off programming the strategy directly into the ninja, however, at this time they seem to be in the "building stages" and are not equipped to implement certain conditions (the simplest is a move stop to break-even after a certain amount of profit).

    I think Ninja will probably be a great product down the line as they develop more conditions and structure to it. Right now, it seems to be best used in longer term trading or simply as a bracket trader.

    Review the ninja forum to get an idea on some of the problems others are encountering with it and how their problems are resolved.
     
    #16     Aug 23, 2007
  7. I never code systems directly into NinjaTrader since I have run into some bugs on their charting end. I have noticed that data is sometimes missing from their charts resulting in miscalculations. If it was not for that, I would have built the system entirely in NinjaTrader. I am really hoping that this gets fixed fast (or maybe it has and I don't know it yet) because I am a fan of NinjaScript.

    NinjaTrader does support moving to break-even after a certain amount of profit is triggered. Most order management strategies are codeable via Ninja's ATM system. In fact, Ninja has some of the best order management capabilities I have seen thusfar. Regardless, for the time being, my trades are bracketed with static profit targets and stop losses so this is not really a concern of mine. I do appreciate the heads up though.
     
    #17     Aug 23, 2007
  8. maxpi

    maxpi

    I noticed data missing from intraday Ninja charts if I close the chart window and reopen it with OpenTick as the provider. I don't think I would have this problem with any other provider. IB is my broker but they don't provide backfill for volume bars.

    What exactly is happening with your charts when you are missing data and data from what supplier?
     
    #18     Aug 23, 2007