Adventures in Automation

Discussion in 'Journals' started by fan27, Jan 21, 2019.

  1. fan27

    fan27

    Busy couple of days.

    Long 2 ES at 2964.75 for 15 min bar strategy.
    Trailing Stop at 2936.75
    Limit at 3015
    Exit after 80 bars
     
    #271     Aug 1, 2019
    shatteredx likes this.
  2. themickey

    themickey

    Ya, am wondering whether we are witnessing a market top. Often very high volatility at tops and bottoms.
    USD roaring up (bad for mkts)
    Gold and crude oil big swings atm, oil falling 8% atm.
    Yesterday on my algo saw numerous large red signals right across the board except oil but that's getting hammered right now.
    bear.jpg
     
    #272     Aug 1, 2019
    d08, nooby_mcnoob and fan27 like this.
  3. fan27

    fan27

    You could be right! Let's see how my strategies hold up. I already have done quite well for the year so a bit of heat is not bothering me.
     
    #273     Aug 1, 2019
    themickey likes this.
  4. shatteredx

    shatteredx

    I took a loss today on my NQ system too :( Could be a bumpy August with this new tariff.

    [​IMG]
     
    #274     Aug 1, 2019
    fan27 likes this.
  5. Daniel.a

    Daniel.a

    @fan27 Thank you for sharing your journey, i am also running algo based models, and at the moment i am looking into trying out a portfolio containing perhaps 30-50 different strategies on the same symbols but i would need to limit max protfolio size to for example 2x contracts per strategy and max 5 strategies as an example... however i am looking for best method to prioritising the signals and to avoid strategies that degrades.. what have you found in your research to give a best/good performance NP / DD... i am looking into Avr trade based on previous 10 trades as an example.. thoughts or ideas, better methods? Many thanks again for sharing
     
    #275     Aug 2, 2019
    alfa8 and fan27 like this.
  6. fan27

    fan27

    I am using the concept of max positions per strategy and max positions across a group of strategies to limit risk. As for strategy degradation, I reconstruct my strategy portfolio at least once a quarter and poorly performing strategies are weeded out automatically in favor of better performing ones. As for the ideas you mentioned, they sound worthy of investigation. As with anything, test and see what works.

    What platform are you using for trading?
     
    #276     Aug 2, 2019
    alfa8 likes this.
  7. fan27

    fan27

    Exit 2 ES at 2936.75
    -$2808.20

    Will post July results later today.
     
    #277     Aug 2, 2019
  8. Daniel.a

    Daniel.a

    Hi, Thanks for your feedback.
    I am using Multicharts. What i am after is to define"Best performing strategy" and "Poorly performing strategies" Preferably i would like to automate this by blocking signals from poorly performing strategies and prioritising signals from best strategies is signals are coming same time..

    I am trading intraday only.

    Anyway i just wanted to see what metrics you had found worked best when it came to define poor perfoming strategies and best ones, and what ranking metric you have found worked best.

    Cheers
     
    #278     Aug 2, 2019
  9. fan27

    fan27

    Here is a very brief summary of my process.
    1. Use an equal distance stop/limit exit and just look at win rate to test strategies. I identify a win rate that random entry (with same exit criteria) can only achieve 2% of the time via 1000 simulations on the same data set. Let's say it is 70%. Then I filter strategies based on that win rate and can say that the entry criteria beats random entry 98% of the time.
    2. With the strategies from step 1, try different exit criteria and filter the results. I usually save any strategy/exit criteria combination with at least a profit factor of 3.
    3. Next, run my portfolio optimization routine where the different strategies/exit criteria are backtested in different strategy portfolio combinations. The portfolios that have certain performance characteristics are saved and I can analyse the results to pick the one I want.

    That process is done at least once a quarter so poorly performing strategies are weeded out.

    I don't know the capabilities of MultiCharts but to do what you are suggesting, you would need to be able have a central component (C# extension) which has access into the performance characteristics of all of your strategies and can provide analytics based on that info. I am using a similar pattern in AlgoTerminal for managing risk.
     
    #279     Aug 2, 2019
    alfa8 and shatteredx like this.
  10. fan27

    fan27

    Down a bit in July and up 23% for the year. I am going to experiment with adding some sort of 10-year bond breakout strategies into the portfolio. In theory, those sort of strategies might smooth out my equity curve a bit and potentially improve net results.

    IbAccountEquity20190803YTD.JPG IB201907.JPG
     
    #280     Aug 3, 2019