• If this is your first visit, you will have to register before you can post. To view messages, please scroll below and select the forum that you would like to visits. Questions? Be sure to check out the Forum FAQ.

Announcement

Collapse
No announcement yet.

Partner 728x90

Collapse

Backtest filling question - prefer stoploss in case of uncertainty

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

    Backtest filling question - prefer stoploss in case of uncertainty

    I would like to ask how NT8 handles PT/SL fill in case there is a big doji bar,
    that hits both profit and stoploss and it is not clear from that bar, what was hit first in reality.

    Assume case, there is no other bar-series with more detailed resolution, that could handle this.

    What NT8 will produce as trade result - profit or stoploss ?
    (Sorry for asking directly, but I am busy now so do not have time to code myself.)

    Just this question came on my mind and I would expect, that that trade result should be stoploss - to make the backtest more realistic and prefer worse result in case of uncertainty)

    #2
    I remember I've read, that NT8 tries to simulate development of the bar according to the OHLC and bar body - and so trying to simulate the development of the bar in reality. This could be fine for fast backtesting.

    But we know, that in case of Doji bar - nothing can be simulated - we are fully uncertaint and nobody knows how to decompose such bar.

    But almost the same is true, when bar-body has only 1 tick. Such a bar is not doji, but there is almost the same 50/50 probability, we estimate it right/wrong. (Just image 40 ticks bar-height with 1tick bar-body - it is almost the same as the doji = almost max uncertainty)

    When thinking further we come to the pattern, that if the bar-body is less then X% of the bar-height, then we are getting close to full uncertainty about how to decompose such bar using OHLC for better filling.

    I would suggest, that all bars, that are below some defined treshold are simulated as with full uncertainty and as a result of this -> they should always prefer stoploss over profit, if bar touches both.

    If the bar-body is above that treshold - for example bar-body is bigger than at least 10% of the bar-height - then the simulation of bar development with OHLC could be used as it is now and allow case that profit could be hit as first one before stoploss.
    Last edited by misova; 08-05-2015, 02:58 AM.

    Comment


      #3
      Hello,

      In nt8 you can turn on High order fill resolution and this will act like live market data in the sense that either the target or stop could be filled before the other.

      You will always be able to use a lower granularity unless you are using a 1 tick series as your primary series. Doji could not form on a 1 tick series

      I look forward to being of further assistance.
      JesseNinjaTrader Customer Service

      Comment


        #4
        I am confused about the high-resolution filling feature with tick-data,
        because you are talking like it should work automatically.

        It sounds like:

        "Just turn on this feature and you are ready to go.
        Don't care, the setting handles tick-precise fill-resolution for you."


        But I assume, this is not so easy. There are no tick-data imported by default and most brokers also do not provide them (or max last few months).

        So I expect, that tick-level fill-resolution will not be applied in most cases (because of missing tick data) and so this feature will not work automatically.
        Am I missing something important here?
        Last edited by misova; 08-16-2015, 04:27 AM.

        Comment


          #5
          Hello,

          So I expect, that tick-level fill-resolution will not be applied in most cases (because of missing tick data) and so this feature will not work automatically.
          Am I missing something important here?
          If there is no tick data, the backtest would not be able to produce executions as there is no data loaded. You would still need tick data for the time period you are testing just as you would need any other timeframes data for the period you are testing.

          If tick data is not available for the time of testing, the fill resolution can be switched to a data type in which you do have available like 1 minute as an example.

          Please let me know if I may be of further assistance.
          JesseNinjaTrader Customer Service

          Comment

          Latest Posts

          Collapse

          Topics Statistics Last Post
          Started by swcooke, Today, 12:33 PM
          1 response
          7 views
          0 likes
          Last Post NinjaTrader_ChelseaB  
          Started by NinjaCustomer, Today, 12:10 PM
          2 responses
          6 views
          0 likes
          Last Post NinjaCustomer  
          Started by KINGKODA, Today, 11:48 AM
          0 responses
          3 views
          0 likes
          Last Post KINGKODA  
          Started by mlarocco, Today, 10:52 AM
          6 responses
          17 views
          0 likes
          Last Post mlarocco  
          Started by Eros Lima, Today, 09:45 AM
          2 responses
          5 views
          0 likes
          Last Post NinjaTrader_ChrisL  
          Working...
          X