Mother of Dragons trading strategy…

Forums ProRealTime English forum ProOrder support Mother of Dragons trading strategy…

Viewing 15 posts - 361 through 375 (of 523 total)
  • #146111

    Ok.. I’ve had a similar issue on another instrument but had some values as low as 3 in the smoothK/smoothD variables. I increased them above 5 and optimised again, haven’t had an issue again yet but it does not fire as often as say the DJI algo. I’d be curious if you increase those to a min of 5 if you continue getting the same error.. I was suspecting that area to be my lines of code failing.. as obviously this can only occur in the code where a division is carried out where the denominator is getting quite small in comparison to the numerator… unless I have missed something elsewhere.

    #146117

    Brent but realized I had to scrap it because of constant zero division error.

    Didn’t experience this before. Thanks for sharing.


    @christofferrydberg
    , in any case, I suspect more on this few lines,

    I noticed you set the lengthStocha to be 3. In case the market very low volatility, and RSI doesn’t move for 3 bars, then you might result a zero for MaxRSIa-MinRSIa

    Maybe you can protect it by something like this,

    If you really interested to confirm this issue, then you can check the time stamp when the issue is reported to you. Put the indicator on chart, then pay attention to the time stamp so it narrow down your debugging range.

    If not above issue, you can still do the same for other division function, by doing a graph of the division.

    2 users thanked author for this post.
    #146692

    My version of the system (little changed from the V4a I copied from here) took a long on 1/10/2020 at 12:00:01 and was stopped out for a big loss on 2/10/2020 at 05:55:58. Today I ran the backtest to look at something else and discovered that the backtest does not take this trade and I cannot see why. Even if I copy the code from the live system into a new system and run it it doesn’t take the trade no matter how much I change the spread value (can’t think what else could account for it). Anyone else had this happen?

    #146831

    Hi Autostrategist, yes, unfortunately I experienced the same (currently running version DJ 5m 4.7.2a).

    Live entry on 1/10/20 12:00:00 and stop loss hit 2/10/20 05:55:56 (UK times). Doesn’t show up on my backtest.

    I’ve been running a version of this since 29 March 2020 and usually real results are very similar to backtest (although 2-3 June and 20 and 24 Aug I found were less good replication). So not just you, but not sure why it happened! Very grateful though for the sharing of this code, as still in profit over that period, despite this 🙂

    #146838

    I added a rounding up for the order sizes, to full numbers, just in case someone has problems orders smaller than 1.

     

    1 user thanked author for this post.
    #147233

    I have tried to run Mother of dragons v4.5 live for a while. The only Changes I have made is my own optimization of the values. Time after time the algo is stopped because “division by zero”.

    Is there a general way to eliminate that or how do a newbie in PRT coding proceed to fix this?

    1 user thanked author for this post.
    #147339

    hi, which version do you suggest to me to trade in real ? I’m interested on NQ, DJ and DAX.

    Thanks

    #147680

    @OboeOpt Might be stating the obvious but a division by zero error occurs when the variable to the right of the division operator i.e. ‘/’ is zero.  Take this line from MOD code as an example (myRSIaMinRSIa) / (MaxRSIaMinRSIa),  if ever the values to the right of the ‘/’ were to equal zero (not saying they can) then the error would occur. I have been running MOD and a DAX version for months without seeing this error.  I have seen the error in the past with other systems and it is usually caused by a situation in the market that had not been accounted for like when the DJI was paused on limit up or down or a market  is unusually quiet and fails to make one or more bars for a while etc.  The only way to track it down is to work through all the possible divisions and see which one is causing the error.

    1 user thanked author for this post.
    #147681

    @Andrew7

    Normally I can figure out why a trade was taken even though I wasn’t expecting it e.g. a difference in the spread from live Versus backtest or insufficient backtest data for the moving average but in this case I couldn’t work it out so I raised the issue with PRT, see below:

    The referenced strategy took a long on 1/10/2020 at 12:00:01 and exited at a loss on 2/10/2020 at 05:55:58. The backtest does not take this trade and I cannot see why. Can you explain please.
    Concerned strategy: (required) : DJ 5m Mother of Dragons v4.9 – Wall Street (DFB) – 28-Jun-2020 16:51:22

    To my amazement I got this response:

    We have identified the problem you reported about your backtest and our technical team is currently working on a correction, which will be implemented as soon as possible. Please note that depending on the complexity of the correction needed, it may take up to several weeks for the problem to be corrected.

    That errant trade cost me several hundred pounds but I don’t expect they will be compensating me 🙁 .

    2 users thanked author for this post.
    #147797

    Mother of Dragons made in one week its 2 worst losses since Feb 2018. In both cases, it first closed one long position (in profit) and took again a long position a few minutes later.

    Then this position would stay opened several hours / days until it hits stoploss. I was wondering whether it would be great to prevent the robot to open a new position if another one has been closed a few minutes ago.

    What are your thoughts about that ?

    1 user thanked author for this post.
    #147864

    You can use something like the below, I got this from one of the PRC gurus.  It will force the strategy to wait N bars.  However, the reality is that the returns are such that you sometimes just take the rough with the smooth, or you risk missing out on the next big move.

    2 users thanked author for this post.
    #147956

    You can use something like the below, I got this from one of the PRC gurus. It will force the strategy to wait N bars. However, the reality is that the returns are such that you sometimes just take the rough with the smooth, or you risk missing out on the next big move.

    Thank you for your answer.  I will look into it.

    But I agree, with this code implemented, it could miss the next big move…

    #147962

    It’s worth noting, with elections in the US soon it’s worth de-risking for the next few weeks.

    #148151

    Hola he puesto lo que le dices a la Madre de Dragones y da un error en mincount. Can you help me please? I’m trying to learn, thank you very much

    #148388

    @josef1604

    In PRT unlike some other programming languages if a variable is declared it has to be used, either use Mincount or comment it out/delete it.

Viewing 15 posts - 361 through 375 (of 523 total)

Create your free account now and post your request to benefit from the help of the community
Register or Login