If you have account already, Log in here first

Log In
  • en
    Menu

    OOS Data – How to build Reliable Strategies

    3 Jan 2017,

    Finding a profitable strategy is not difficult, but that is not enough. Not only must be the strategy profitable, it must also bring the profit in the future.

    Step by step we will have a look on how to build trustable strategies. In the first part we will focus on OOS data, which means Out of Sample data. No problem if it does not known to you, I’ll explain you everything :).

    What are OOS data?

    Let´s imagine you have historical data for 10 years period, since 2006. You have two options if you want to start to build the strategies. The first and worse is to take all the data and do backtests on it. Maybe you can see yourself thinking about doing it in this way, after all this mistake is done by many people. If we build a strategy on all data we have, there is a big chance that we will find something that works randomly and it will not work in the future.

    It is better to divide the data into more parts. For example we can use the 2006-2013 data for the strategy building and on the other part of the data (2014-2016) we will let the strategy run and we will see how it behaves outside the data area where the strategy has been build and optimized. The strategy will be more robust and we have proof that it is not just something that works by chance. Of course, this is not the only test that needs to be done, but it already means something.

    The strategy then looks like this (the blue part is OOS):

    How do I work with OOS?

    I am personally using more OOS, precisely said three of them, which I am involving in various stages of development. It allows me to test whether the other quality tests of the strategy are still conclusive and the strategy is really as good as I think. Finally, I build fewer strategies but better ones. You can find more information about my development method here.

    Option report

    Get your report from Jan Kaska's trading every week!

    02-2018-05-11-account-overview

    Leave a Reply

    Your email address will not be published. Required fields are marked *