Not logged in
Viewing 6 posts - 1 through 6 (of 6 total)

Forums>StrategyQuant>General Discussion>A simple way to know and calculate how strategies will profit in the future

  • #233082 |
    Customer
    28 Posts

    There is a very simple way to know how our portfolio will profit in the next year without waiting a year, of course!

    1. Standardize your working method at max

    2. Create the strategies you want to test in real using only data prior 1 year ago. Try to imagine that today is (today-1year). At least 100 strategies multi timeframe and multi instrument.

    3. Once that you are satisfied with your basket (all robustness test that you want, exetera…) test all passed strategies in the last year.

    4. Create a portfolio with the results. Done

    Now in your portfolio you will have some strategies in profit and some not.

    a. Just count them and you will know how many strategies will profit in one year after all your work.

    b. Then, if the portfolio in in profit, this means that your method is good and when you will repet the same workflow using also the last year and you will put the strategies in real, you will have positive return in one year.

    c. You can also evaluate if the curve of the portfolio at some point will flatter or starting to decrease. In this way you can also know how many time you can trust in your strategies before creating an other portfolio.

    With this simple (and long, I know) procedure, you can have a real and very accurate test of your job. If the portfolio is in loss, create a new procedure to find and test the strategies.

    What do you think?

    Are you able to have a positive portfolio ested in the last year?

    #233302
    Participant
    21 Posts

    Sounds like a good idea but can you share screenshots of how this process is done?A picture is worth a 1000 words.Please share screenshots for better understanding.Thank you and green pips to you Good Sir.

    #233671
    Customer
    247 Posts

    @roberto,

    yes ist is a good way. This is my way how I generate Strategies and test the filtering process.

    But somtimes some Strategies are working good in backtest and not so good on demoaccount.

    Somtimes strategies are good in backtests and good on demoaccount but not good on real account.

    It is a long way to find good strategies.

    You can record the tickdata from different brokers and make backtests with your eas and compare the results.

     

     

     

     

    https://monitortool.jimdofree.com/

    #233672
    Customer
    28 Posts

    @roberto, yes ist is a good way. This is my way how I generate Strategies and test the filtering process. But somtimes some Strategies are working good in backtest and not so good on demoaccount. Somtimes strategies are good in backtests and good on demoaccount but not good on real account. It is a long way to find good strategies. You can record the tickdata from different brokers and make backtests with your eas and compare the results.

    Personally, the only good strategies I’ve found are only buy stop, so if the broker is not good you suffer big slippage, because usually these orders re executed on an explosion of volatility. Besides this, I get in real the same that in backtest.

     

    #234623
    Participant
    18 Posts

    Personally, the only good strategies I’ve found are only buy stop, so if the broker is not good you suffer big slippage, because usually these orders re executed on an explosion of volatility. Besides this, I get in real the same that in backtest.

     

    Exactly, after years of searching all my strategies are buy stop/sell stop

    #238309
    Customer
    100 Posts

    There is a very simple way to know how our portfolio will profit in the next year without waiting a year, of course!

    I have been thinking about something similar for a while. The new SQ-X version gives us the possibility of creating our own ‘custom project’ workflows. Now we can validate the whole design flow from A to Z, (from ‘strategy generation’=A to ‘final results’=Z).

    Assuming we have already verified the correlation between SQ backtest trades and the real trades on our real broker, then yes, we could ‘reserve’ one or two most recent years to perform the final OOS tests. This way we could see if our workflow, is able to produce profitable strategies on new ‘unseen’ data.

    Ideally we could validate ALL resulting strategies in this last OOS period (last few years of data) and compare the results to the behavior as observed during IS and OSS (used during strategy generation and validation tests). I wrote ALL in capital letters because, we need to prevent manual ‘cherry-picking’ of the final strategies, since this would destroy our ability to objectively compare different workflows.

    HOWEVER! This does not necessary mean that our portfolio needs to be profitable in this last OSS period. Note that any of generated strategies that have passed all validation test, can easily start with a period of drawdown or a long stagnation. So we should only look, if generated strategies do not perform differently (worse) in this last OOS than in IS/OOS used during strategy design.

    In a perfect world, for every tested workflow we would have a automatically calculated percentage of strategies that are still OK in this last OOS period.

    Those are just my thoughts..

    Any one has some ideas how we could automate this?

    Gr

    Chris

     

     

     

    Do you want to become a profitable trader? Join our free community to learn more about algo-trading:forum.coensio.com

Viewing 6 posts - 1 through 6 (of 6 total)

You must be logged in to reply to this topic.