Discussion re Pure Renko strategy

Forums ProRealTime English forum ProOrder support Discussion re Pure Renko strategy

Tagged: 

Viewing 15 posts - 241 through 255 (of 346 total)
  • #126579

    The screenshot was from Demo Live £1 lot size, however since this morning I am now running it “Real Live” with £0.2 lot size. So far 2 winning trades which matches my demo live test.

    I am also testing since this morning a daily optmisation but letting each strategy run for 1 week. So splitting the strategy into 5, mon-fri, and running daily optmisation the evening before the day it goes live. So Renko-Tuesday.itf will be optimised on Monday evening and reoptimised 7 days later. Tuesday evening I’ll optimise Renko-Wednesday.itf and so on. I’ll let you know how that goes..

    Daily optimsation has shown better results with different box sizes but the fixed settings has proven to work better for me (so far)

    I use box size 50 and renko type 1, run in 5 seconds (not 10s as mentioned above)

    No changes to the original code (I think) but posting my verison here for reference.

    4 users thanked author for this post.
    #126582

    @GraHal

    Yesterday wasn’t too bad 😉 but there was a big loss there too.

    #126590

    Applied machine learning on 2.3pR version, and results are AWESOME.
    I will try today to apply it on every version and on every timeframe in order to find the best one 🙂

    5 users thanked author for this post.
    #126592

    I’ll let you know how that goes..

    Yes please … that will be well interesting an useful.

    I got the S10 TF version going good now … backtest attached.

    PS

    Even better (2nd image) with the settings shown for Renko Type = 2 and Box size = 95.

    1 user thanked author for this post.
    #126602

    First good version, 10sec.

    Position size: 0.2

    It only works between 8.00 to 21.00 in order to use just the lowest spreads available for DOW.

    As you can see from the equity curve, the first day the system took a lot of trades in order to better optimize the machine learning algorithm values following my istructions; then, the system started to work normally with costant gains and trades.
    So in a live trading optic, should be better to run it before 8am and not during the day, that could bring to losses.

    I confirm that ML could be the best thing to fit into these short TF systems.

    2 users thanked author for this post.
    #126607

    In addiction, still thinking in a live trading optic, I’d run a new system every Sunday evening: “sacrifice” the Monday in order to let the system better optimize itself, and taking gains the rest of the week; closing it Friday evening and starting a new one on Sunday again.

    1 user thanked author for this post.
    #126611

    maybe simulated trading between certain times or the first day to let the heuristic engine determine the values and then go live.

    2 users thanked author for this post.
    #126614

    First good version, 10sec.

    Will you be sharing your first good version?

    Is there no way to avoid the ‘first day sacrifice’!? Poor System! 🙂

    Why not optimise the settings before you set it going?

    Above is what I did when I got a few ML Systems working. We could even do a walk forward and use the settings for starting value etc from the latest IS period … least then the System kicks off with a fighting chance?  Or even just use settings from a normal 10k bars optimisation?

    Is above what you do or do you guess / finger in the air for starting value etc?

    Just asking / just saying! 🙂

    #126617

    probably you had so many trades the first day, because the start values are way off the ideal ones.

    What about taking the last values of your backtest, and put them in as starting values, or only slightly less.

    #126632

    I just talked about some random ideas in order to stimulate brainstorming, sacrificing some systems 🙂

    Heres the valuex (renko type) and valuey (boxsize) that i’ve used on the 2.3pr 10sec.

    P.S.: Reset periods are just randoms

     

    1 user thanked author for this post.
    #126648

    Sorry for my newbie-question but how do i implement Francescos mashinelearning Valuex and ValueY in to the 2.3 version of the system?

    So exciting following these threads with your genius codes, hope to someday be able to contribute with something as well. Wish you all a nice weekend!

    #126651

    I’m everything but not a genius coder 🙂
    I’m still a newbie but reading everyday the forum since months and doing hundreds of tests on the platform i’m starting learing something. It becomes fun and fascinating especially when you are surrounded by the people of this section that works everyday for new challenges with strong motivations.

    Btw, if you want the answer to your question and start learning something, you can study this topic https://www.prorealcode.com/topic/machine-learning-in-proorder/ and you will easily understand in few pages how to implement the ML in the code.

    #126652

    Today “Real Live” results. Positionsize 0.2.

    1 user thanked author for this post.
    #126655

    nice! how many % or points did you set the stoploss?

    #126662

    Maxloss was set to 1 – does this part of the code act as a stop loss? I didn’t add a ‘set stop loss x’ code and wasn’t sure so I was watching the screen when the trades were on 🙂

    It seems to be the entry is not perfect, but it’s probably very difficult to achieve that on this timeframe. However, it has a high win rate so it often ends up in profit as can be seen in my previous demo forward-test.

    A tight stop loss will definatly see less wins. I don’t know how we can resolve this without adding too many parameters to the strategy. I was looking into oscillators on different timeframes, and voss predictive filter, but haven’t yet found anything that works well.

    Any ideas?

Viewing 15 posts - 241 through 255 (of 346 total)

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