Trial to Transform a Donkey in to a RaceHorse !

Forums › ProRealTime English forum › ProOrder support › Trial to Transform a Donkey in to a RaceHorse !

Viewing 15 posts - 1 through 15 (of 24 total)
1 2
  • 09/09/2020 at 3:05 PM #143758
    zilliq

    Hi Guys,

    As Promise I open a post to explain how we can transform a stup strategy=Donkey in a racehorse=Good OOS

    We will try on EUR/USD as I love Forex on 1 mn to have more trades on V11

    The purpose is to confirm too that the strategy means nothing in an algo and the most important is the code and the money management = We can transform a Donkey in a RaceHorse ! In any case we’re going to try 🙂

    First we need a Donkey. I choose CCI that many people use and the stupid signal is crossover with -100/100 (One variable)

    I think that we are all agree that we can’t do more stupid

    Market structure with ADX in this case (One variable) and money management with stop loss (One variable)

    So we have only 3 variables (Generally I don’t use no more that 4 variables). Less is potentially better, but very difficult to obtain good results

    Now backtest on 100 000 units with a large step to find the best variables

    And we have the results below : +75 % not bad, but not excellent 69 trades = begin to be significative +405 % gain

    2 users thanked author for this post.
    avatar paisantrader
    09/09/2020 at 3:11 PM #143761
    zilliq

    Now I see the 2D results and choose the variables with better Gain as on the picture

    As I said, results are not fantastic but it’s a Donkey !

    To do a second pass backtest with a step of 1

    I see that the better range is

    A 17-50

    b 60-80

    c 30-45

     

    09/09/2020 at 3:24 PM #143763
    zilliq

    I change the range A 22-50 because there was more than 10 000 combs

    We can adjust the range with the 3D view

    And now we wait for the second and last pass

    09/09/2020 at 4:56 PM #143770
    zilliq

    Sorry the backtest on 10 000 combs was long

    And now we select our “best” variables according on the 2D results

    a 27

    b 68

    c 28

    What is interesting is that is not the best results ! That’s why I’m very doubtful on Walk Forward actually because we can’t select the filter we want to (not on sme indicators)

    The result on “Detailed results” are not very good too : 88.46 % winners, Win/Lose 3.11 Sharpe ratio 0.55

    And now we begin the OOS and see what it appends on the next days/weeks

    Does the Donkey become a racehorse ? 🙂

    09/09/2020 at 5:02 PM #143774
    zilliq

    Let’s start playing OOS with the last Algos on incubator 🙂

    1 user thanked author for this post.
    avatar Paul
    09/09/2020 at 6:01 PM #143786
    April O’Neil

     

    Hello Zilliq,

    how can you get only one variable for all your money management ? i mean, there is many things to eventually optimise : Trailing steps, trailing start, ATR period, , SL, Take Profit.

    The only possibility i see it’s you using ATR Trailing stop with defined period, start and step, without TP. You only optimise %SL. Am i wrong ?

    Thank you for your awarness about strategies simplicity.

    09/09/2020 at 6:45 PM #143788
    zilliq

    Hi @April O’Neil

    Remember that more you optimize, more you overfit, worst will be the OOS 😉

    Keep always simple and avoid to double some code (Stop loss and Trailing stop for example).

    And sometimes we don’t need Target as they stop our gain. With some Algos it could be better to use trailing stop (Not all algos)

    Personnaly I never use a variable for Stop loss, target, trailing stop, trailing step and so on…The Equity curve on backtest will be perfect but OOS very bad

    And the trailing stop could be fixe, or dynamic Var*ATR

    In conclusion one variable is necessary (I never find how to have 0 variable)

    Bye

    09/10/2020 at 1:11 AM #143806
    Finning

    Hi Zilliq,

    I like your approach, and I have been doing something similar over the last couple of weeks, with things seeming to work at this stage.

    I have given up on the set and forget forever bot – tried to make, love to have, but I can’t see it happening.

    If I have to revise parameters every couple of days, or at the end of the week – then so be it – I’ll do what I have to to stay in profit.

    I do agree though, it is a fine line between making something that works in the future, and having an oh so splendid looking backtest that doesn’t give you anything when you trade it.

    The 3 or so parameters seems to be the number that works for me too at this stage – 15min or so to set up something simple.

    Just an idea though – voicing something that you’re probably already doing. Just another form of risk management, but on an account level. How much do you want to risk on your new set of numbers, and when do you decide that your new set of numbers isn’t working anymore after you’re in profit. Drop in at the end of a code.

    Thanks for your posts, they are interesting reading, and food for thought.

     

    09/10/2020 at 9:55 AM #143834
    zilliq

    First day of the bad Donkey : 2 wins, goal of the day reach + 20,99 %

    Good start for a Donkey 🙂


    @Finning     Not sure to understand your question. If I understand, no I don’t use max drawdown because it exits too soon

    In this algo capital = 10000, I risk 10 % on each trade to determine my position size (Van tharp / Vince method)

    09/10/2020 at 10:15 AM #143844
    Finning

    Hi Zilliq,

    “The How much do you want to risk on your new set of numbers” wasn’t necessarily a question to you, but I was trying to define a possible answer to that question with the supplied code. Yes i was just throwing an idea about – not a recommendation, as I’m sure you know what you’re doing, and you didn’t ask. It is just something that I use with a strategy when I’m not sure how long it’s going to last/be valid for, so I know where I stand in terms of account risk if the strategy decides to go bad.

    When you say 10% risk, is that 10% of your capital for use as margin, or is that 10% of your capital as a stop loss risk? I use 2% normally for margin risk, and if a strategy lives with 10% margin risk it may possibly be a good sign that it’s robust.

    I also try and stay with a 2% of account stop loss risk, though this always varies a little bit.

    The idea behind my above post, as, since it’s strategyprofit based, was to 1:- limit account drawdown from the trading method, and, 2: – try to protect made profits. It doesn’t take into account the amount of margin that is used by the strategy.

    I do have a question though, with your ATR money management:

    Do you do something of the sorts like:

    keeping a 1:1 risk:reward ratio, or is it something more elaborate than that? I know you can’t give away too much, but any hints are appreciated.

    I do fully agree that money management is a centrepiece of any strategy, and it’s something that I focus a lot on.

    Also, I noted your use of ADX to help define the market – do you have any other things that you use for the same purpose/any other hints??

    Thanks,

    Finning

     

    09/10/2020 at 2:28 PM #143876
    zilliq

    Hi @Finning

    10 % is to define the size of my position (It could be 2, 6, 8 %) with a demo capital of 10 000

    It could be 1 % in a demo capital of 100 000 whatever …

    The amount must be enough to win enough.

    The % and amount doesn’t change a lot in the money management in fact (It depends of others factors like Win/Loss and so on..for example you can have only 1 % risk with a very good Win/Loss ratio)

    Yes for the ATR (In fact it’s more complicated (Something like a supertrend on ATR) but the idea is here)

    I don’t use Risk Ratio but try not to have less than 1

    Have a nice day

    09/12/2020 at 10:10 AM #144005
    zilliq

    Yesterday wasn’t a good day for our Donkey, not (actually) a problem, we will see next week

     

    09/14/2020 at 6:45 PM #144245
    zilliq

    Not bad, but not very good today on Bad CCi Donkey, only half the goal reach

    +5 % today, +18 % since 3 days, 46.73 % in the market = bad for me

    We will see tomorrow

     

    09/15/2020 at 8:35 AM #144290
    zilliq

    Bad CCI Donkey system already stop ! It’s 9:33 AM. I love it 🙂 Goal reach 1000 USD/day

    +28% in 4 days Win/loss 2.24 Sharpe R 0.92

    Remember the purpose is top see if after one week a bead Donkey can be profitable

    Bye

     

    09/16/2020 at 11:15 AM #144411
    Gubben

    Following this just because I would love to have a strategy called RaceDonkey running in the portfolio.

    1 user thanked author for this post.
    avatar zilliq
  • Author
    Posts
Viewing 15 posts - 1 through 15 (of 24 total)
1 2

Find exclusive trading pro-tools on ProRealCode Marketplace

Similar topics:
fourier transform and moving averages
how can trial strategies purchased from marketplace be deleted

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