Home Forums Community Discussions Backtesting Backtesting – initial testing

  • Backtesting – initial testing

     Solo updated 2 months ago 2 Members · 2 Posts
  • j4t1nd3r

    Member
    2021-03-04 at 2:47 pm
    MT Roids 14189 Roids
    Level 15: Lift-off

    Hello peeps,

    so I have done a like a quick 20 mins on the backtesting update. I plan to be using it more over the coming weeks… so I will have more in depth feedback.

    Good stuff:

    ) The backtesting system is laid out well, its easy to understand and works quickly.

    ) The information presented visually and table form are both excellent.

    ) It’s fast!

    ) You guys got vscode working in a browser, i don’t know how hard it is but that’s great, also the fact we got syntax highlighting is awesome.

    ) There is a history + rerunning the same parameters does not rerun the calculations again, it pulls the data from a cache making it even faster = wow

    Issues:

    On the history tab and clicking the triangle to expand one of the quick SMA backtests, the parameters on MAFast and MASlow are defaults instead of the correct values.

    This is the case on all Quick SMA backtests I ran. However the rest of the information on that tab looks correct.

    What I would like to see added:

    The page where to pick an algorithm, it would be nice to see a different kind of layout that explains the relationship between them visually.

    Maybe a tree based hierarchy that explains how each one was developed, how they factor in stuff the previous one didn’t or separate trees for different techniques.

    Right now I have no idea the relationship between all these algorithms what came first, what came last, which one is more complex etc (outside of quick SMA).

    Example:

    Family 1

    algorithm1 > algorithm2

    > algorithm3

    Family 2

    algorithm1 > algorithm2

    Family 3

    algorithm1

    Above, each technique that are completely different can be separated into families, we can see for family 1 that algorithm2 is kind of a successor to 1, algorithm3 may be an offshoot etc.

    This is just an idea, i don’t know if algorithms are actually related like this.

    It would be great if hovering over different ones in the same family gave information:

    like algorithm2 fixes some weaknesses in algorithm1 by the following attributes…. algorithm3 leans into a specific style of strategy loosely based on algorithm1… etc

    Forum related:

    Not sure if your aware… I seem to have dark mode on, not sure if its the default on the forum, creating a new discussion, the text is light grey on white background. I typed this post up in vscode and pasted it in to get around the visibility issue.

  • Solo

    Administrator
    2021-03-04 at 8:26 pm
    MT Roids 56590 Roids
    Level 57

    Hey j4t1nd3r

    thanks a lot for this extensive feedback of your first 20 minutes.

    First off, we know the version that is deployed right now has calculation bugs, they are fixed on the master and this should be pushed to your instances super soon, along with a couple of other bugfixes.

    Thanks for all the kind words, we are glad to hear that the new engine seems to do its job!

    besides the obvious bugs you mentioned (thanks, we’re getting them sorted), you are right and there will be many addons to the “library”, once we will activate our “marketplace”. We will have groups for strategies, dividing them into different “categories” as well as “risk”. Also, you will see historical performances, user ratings, comments, the author and much more. This, along with sorting filters should do the trick.

    We are hearing that you like a better strategy description with more details about the approach and we will incorporate this for sure.

    all the functionality of the social web are not yet bugfixed on the web api that delivers data to the MT instances, this is still in prototype mode.

    I’ll let you all know once we have pushed the next update!

    Looking forward to more feedback!

    Thank You!

Log in to reply.

Original Post
0 of 0 posts June 2018
Now