Machine Learning in ProOrder ProRealTime

Forums ProRealTime English forum ProOrder support Machine Learning in ProOrder ProRealTime

Viewing 15 posts - 31 through 45 (of 455 total)
  • #121130

    @GraHal try this 🙂 The HeuristicsCycleLimit parameter will determine how many ‘cycles’ each algorithm gets. If someone is up to it I also think we should look at introducing different performance metrics like Drawdown vs Profit Ratio to the existing WinRate (WinCount) and AvgProfit (StratAvg) metrics

     

    2 users thanked author for this post.
    #121137

    Here is the version of the algo outputting ‘ValueY

    Referring to Line numbers of the version2 / ValueY posted with the link above, the ‘2’ is missed off  …

    Line 123 –  increment
    Line 157 & 160 – i
    Line 180 & 188 –  Mode
    Line 201 – BestA

    I’m not trying to be clever, I ran the ValueY version on its own as part of a trial of my ‘barindex idea’ and the syntax checker told me the above were not defined! 🙂

    I’ll now have a go with the HeuristicsCycleLimit.

     

    As an aside … anybody got any ideas / code to identify even & odd number barindex ?

    #121143

    identify even & odd number barindex

    Good how ideas come when away from the screen!? 🙂

    I’m not there yet, but I’m thinking around ?? …

     

     

    #121159

    Below is the HeuristicsCycleLimit code with the corrections mentioned 2 posts above … 2 was missing here and there in Algo 2.

    Least I can do to save JuanJ some time for letting us in on his coding gems! 🙂

     

    #121168

    JuanJ you sure below code works please?

    I can’t get any results using below … despite trying what I thought was logical re changes etc! 🙂

    Maybe you try it if you get chance please?

    Or anybody else get below to work (using the full code above)?

     

     

    #121171

    Ah! I forgot to initialize the variables:

     

    2 users thanked author for this post.
    #121186

    I should be able to GRAPH ValueX and ValueY as normal?

    If yes then mine isn’t working with the 2 Algos as ValueX and ValueY GRAPH as constant starting values.

    I’m going to start again with one Algo! 🙂

    #121221

    I give up! I can’t get it to work, it doesnt give bad results … it doesn’t even even start.

    I’m sure it is something simple, although I’m sure I’ve tried everything simple!? 🙂

    I thought I was getting 1 x Algo to work. but I realise now that I was optimising the starting value and the starting value  continued to be used for all trades!

    Seems to have gone quiet after the first few flurries of excitement?

    Anybody else got it to work?


    @Juanj
    I have changed your values to my respective values etc, but how are you defining … Best Avg Strategy Profit?

     

    #121222

    I feel somewhat stupid  but I don’t have a clue what this heuristic algo code does ! Maybe I’m not the only one…can anybody explain in a few sentences how does it work?

    Thanks

    #121224

    can anybody explain in a few sentences how does it work

    It optimises 1 x variable value up / down  1 x pre-set increment depending on the performance of the previous 3 (pre-set number) trades.

    I am reading between the lines of the code (as I can’t follow it all) but what I’m not clear about is … how does the heuristics decide that the one or all of the past 3 trades were not just bad trades anyway?? As we all know we can get  a bad run for while.

    I guess the heuristics must be comparing against an average profit … which is why I asked JuanJ how  Best Avg Strategy Profit is defined as this must be critical.

    But at this stage I’d be happy if I could see the GRAPH of ValueX moving from it’s starting value! 🙂

     

     

    2 users thanked author for this post.
    #121229

    @GraHal Please share the ITF file of the strategy you wish to implement the algorithm to and I will have a look to see where the problem is.

    Then to further elaborate on GraHal’s explanation:

    The algorithm also keeps track of the following:

    • Best average performance (Profit) measured over all the previous evaluation periods (StratAvgB)
    • Best win rate (Win %) measured over a all the previous evaluation periods (WinCountB)

    It then measures and saves the same metrics but only for the last evaluation period (StratAvgA and WinCountA) and compares them to the above.

    It then starts by systematically incrementing the target variables both up and down to see what appears to improve performance most

    There are also some other calculations going on but that is mostly to achieve the above.

     

    5 users thanked author for this post.
    #121230

    I just added the machine learning into a system, here’s the results (before and after).

    Very impressive, and also the robustness test went up in total average!

    4 users thanked author for this post.
    #121274

    I will have a look to see where the problem is.

    Thank you, I may take you up on the offer! 🙂

    I am learning stuff by me keep asking myself  … “why the eff isn’t it working?” and as your code is applicable to any strategy it is time well sweated over! But it sure don’t feel like it yet!  🙂

    I got it going on another strategy.

    Should your heuristic code work on strategies that switch position from  Long to Short with no bars not onmarket ?

    #121286

    Should your heuristic code work on strategies that switch position from  Long to Short with no bars not onmarket ?

    Yip that is the problem. Solution;

    Replace this:

    With This:

     

    2 users thanked author for this post.
    #121318

    hi juanj

    one algo works, but 2 fails somewhere.

    I created a basic strategy on a second timeframe for testing. Can you see what is wrong please below?

     

     

Viewing 15 posts - 31 through 45 (of 455 total)

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