Not logged in
Viewing 4 posts - 1 through 4 (of 4 total)
  • #114085 |
    Customer
    32 Posts

    Hi,

    I think it’s a amazing feature of the current version of the program. However what worries me that the selection is done on the complete dataperiod available. Why is it not possible to user select a data range so that the selection can be tested out of sample? Otherwise I think that curvefitting is a big problem with this method despite the monte carlo.

    Regards,
    Dennis

    #132166
    Moderator
    1568 Posts

    When I paste strategies reports into Quant Analyzer, those results already contain enough out-of-sample performance (never seen before) data. This way I don’t worry too much about curve fitting …

    #132191
    Customer
    32 Posts

    I don’t mean curvefitting for the system itself. I mean curve fitting for the portfolio combination. So in current version you would find a combination which looks great in a portfolio because of correlation. However this combination might be pure luck or random. And when trading in realtime this might not be a perfect portfolio at all!

    #132346
    Moderator
    1568 Posts

    Sure, this could be considered as “curve-fitting” – selecting system to get the best portfolio equity but it’s all we got. So I said, I only want to add systems that are proven not to be “curve-fitted” on that specific dataperiod

Viewing 4 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic.