Machine Learning in ProOrder ProRealTime

Forums ProRealTime English forum ProOrder support Machine Learning in ProOrder ProRealTime

Viewing 15 posts - 61 through 75 (of 455 total)
  • #121413

    Good evening,
    Unable to initialize variables
    Monthinit
    Dayinit
    removing some of the code works.

    1 user thanked author for this post.
    #121416

    SHOULD we initialize it is variable?

     

    #121417

    Valuex :
    MOVE

    #121420
    add

     

    1 user thanked author for this post.
    #121422

    @fifi743 … got ValueY working so my (now deleted) comment is no longer needed.

    #121432

    only a single increment is made either up or down

    Is the 1 x increment of the value set as increment?

    So if I set increment = 2 then in any one bar, ValueX or Value Y should not change by more than ‘2’ (up or down)?

    #121435

    only a single increment is made either up or down

    Is the 1 x increment of the value set as increment?

    So if I set increment = 2 then in any one bar, ValueX or Value Y should not change by more than ‘2’ (up or down)?

    Correct, unless the increment limit value has been reached, after which ValueX will revert to the new Average best performing value

    2 users thanked author for this post.
    #121436

    for test :
    Depending on the month it changes from HeuristicsAlgo1 to HeuristicsAlgo2
    I moved valuey to line 40

     

    1 user thanked author for this post.
    #121443

    I just had a little play with the first code that Juanj posted and I noticed that I could sometimes get the ‘Periods’ variable to go negative. You might want to add a minimum value allowed to prevent this as it could cause a strategy to be stopped.

    #121480

    @Vonasi, if you consider the latest version of the algorithm (not the one from a few years back), there is a minimum value and as long as MinValue is specified as a positive value a negative value shouldn’t occur

    #121546

    Attached results using Paul’s System.

    • Top Curve WITHOUT Heuristics (code is bare System / Paul’s original code only).
    • Bottom Curve WITH Heuristics.

    Bottom Curve … Heuristic Algo1 and Algo2 are both working at the same time.

    I found that it is not a case of adding Heuristics on top of a best effort bare code.

    I am going to start my version on Forward Test now to establish that my version is not just a massive curve fit of the Heuristics!? 🙂

    I am posting my results as encouragement / to show that JuanJ code does work.

    Maybe you can get JuanJ code to work with even better results on Paul’s System than attached?

    3 users thanked author for this post.
    #121716
    Jan

    Hi Juan,

    I try to understand the concepts of what you are trying to obtain with this machine learning, could you explain this ?

    Normally you develop a strategy with fixed settings (eg like a Macd crossing strategy) based upon past data, and assuming the future movements of price are relatively similar, and you let run your trading-algo.

    With this machine learning concept, are you trying to further optimize the following of past changes in price movements, then with those optimized findings approaching the future price movements with the new strategy ?

    Or when running a algo-strategy, with this machine learning concept, it adjust the setting of the variables automatically given the structure of price movements ?

    (I hope my question is clear,  in this case it would be great to discuss this kind of topics face to face in a meeting with others, to avoid misunderstandings)

    #121819
    Jan

    Hi Grahal or others,

    Could you be so kind to spend a few words to explain the concepts of the “Heuristics” ?   Is it daily optimising of a fixed setting, and yes, what is the base of the adjustment, how many bars backwards ?

    I am sorry, but I missed that out of lots of interesting code.

    Thanks in advance for your help !

    Your comment “Attached results using Paul’s System.

    • Top Curve WITHOUT Heuristics (code is bare System / Paul’s original code only).
    • Bottom Curve WITH Heuristics. “
    #121909

    when running a algo-strategy, with this machine learning concept, it adjust the setting of the variables automatically given the structure of price movements ?

    It is above. (I am commenting as JuanJ is probably busy trading or mentoring etc)

    I have found though that the starting value for ValueX is important to get good results.

    Also a few other variables in the Heuristics are important re initialisation value. This is understandable else the Heuristics would be the panacea for all!? 🙂

    #121910

    how many bars backwards ?

    This is a setting in the Heuristics Algo (Halgo), but it is set by number of Trades backwards.

    So to use an example … the results of the last 3 Trades Trades 1, 2 and 3 are used to vary the Halgo variables in order to produce an optimum ValueX to go forward with for the next Trade (4).

    Then it is my understanding that … Trade4 would be used together with Trade3 and Trade 2 to repeat above.

    I will try and decide … from the code … if it is above or is it … the Halgo variables are next optimised after Trade 6 using Trade 4, 5 and 6??


    @Jan
    you are far better at coding than I so maybe you can read JuanJ code and decide??

    Whoever finds out first, post on here, unless JuanJ passes by and tells us first?

     

Viewing 15 posts - 61 through 75 (of 455 total)

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