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

Forums>StrategyQuant>General Discussion>Robustness tests for what?

  • #245556 |
    Customer
    64 Posts

    Hi Friends!

    I made an interesting experiment.

    I had 120 strategies, that seemed to be very good from 2004.01.01. – 2017.01.01.
    From these 120 strategies:

    – 20 were nice between 2017.01.01. – 2019.01.01. too
    – 100 have been failed. The equity curve went down.

    What I wanted to know: how do look the robustness tests at the 20 good and the 100 bad strategies? (For me are the DIFFERENT MARKET and the RANDOMIZE PARAMETERS the most important. I made now just these 2 tests.)

    The result is very-very interesting!

    I found, that:

    – none of the 20 good strategies passed the diff. market and rand. parameters
    – 6 from the 100 bad strategies passed both the diff. market and rand. parameters

    So I would have taken just loser strategies if I decide based on the robustness tests.

    Question: ROBUSTENSS TEST for what?

    #245565
    Customer
    399 Posts

    I believe using robustness tests is only away to sort out strategies based on their worst historical performance hoping that the future result will not be worse then that. It cant in anyway tell you what next year expected performance is it can only simulate what it can be in worse or best case and that is long term short term you have to look at things as max draw down. So if the max simulated draw down is 100% you can expect 100% draw down in worst case when you start trading it immediately  :). So to counter this you have to build a diversified portfolio.

    #245581
    Customer
    61 Posts

    Hi Friends! I made an interesting experiment. I had 120 strategies, that seemed to be very good from 2004.01.01. – 2017.01.01. From these 120 strategies: – 20 were nice between 2017.01.01. – 2019.01.01. too – 100 have been failed. The equity curve went down. What I wanted to know: how do look the robustness tests at the 20 good and the 100 bad strategies? (For me are the DIFFERENT MARKET and the RANDOMIZE PARAMETERS the most important. I made now just these 2 tests.) The result is very-very interesting! I found, that: – none of the 20 good strategies passed the diff. market and rand. parameters – 6 from the 100 bad strategies passed both the diff. market and rand. parameters So I would have taken just loser strategies if I decide based on the robustness tests. Question: ROBUSTENSS TEST for what?

     

    Here it would be interesting to know what kind of robustness tests you mean or did.

    If you assume that the robustness tests are possibly unnecessary, you also have to tell us what kind of test you are talking about.

     

    Hello! I'm looking for people who want to earn some money on the side! The entry is simple, just install the browser https://get.cryptobrowser.site/4117939 and use it daily. It's fast, easy to find and practical to use - you'll love it! But the main thing is that you can earn Bitcoins directly in it! Does that sound good? Don't think long and join in!

    #245584
    tomas262
    Administrator
    1180 Posts

    If the robustness test has negative impact on strategy performance it does not have to necessarily mean the strategy is bad and must be thrown away intermediately.

    Does the logic behind what makes the strategy work make sense? If so and if the strategy continues to work after many years of backtest data you could consider keeping it. But it also depends which of RT tests went bad whether it is the slippage test, parameters change test, just trade order modification etc. So you need to investigate what caused the strategy RTest to fail and how likely is that the worst case scenario shown by the robustness test comes true in live trading

    #245587
    Customer
    64 Posts

    Hello Everybody!

    Thank you all for the answers. You asked, what kind of RT I made / mean. I wrote in my post: “For me are the DIFFERENT MARKET and the RANDOMIZE PARAMETERS the most important. I made now just these 2 tests.”

    Of course I make also slippage, randomize orders tests, etc., but the above mentioned 2 tests are for me really important. And for this little experiment I made just diff market and rand. parameters. For me these 2 tests are the most crucial.

    And for you?

    #245588
    Customer
    61 Posts

    For me are the most important tests:

     

    1.  Slippage

    2. Different Timeframe

    3. Different market

     

    Hello! I'm looking for people who want to earn some money on the side! The entry is simple, just install the browser https://get.cryptobrowser.site/4117939 and use it daily. It's fast, easy to find and practical to use - you'll love it! But the main thing is that you can earn Bitcoins directly in it! Does that sound good? Don't think long and join in!

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

You must be logged in to reply to this topic.