How to tell when you trully dont curve fit

Forums › ProRealTime English forum › General trading discussions › How to tell when you trully dont curve fit

  • This topic has 12 replies, 5 voices, and was last updated 6 years ago by avatarLeo.
Viewing 13 posts - 1 through 13 (of 13 total)
  • 11/02/2018 at 12:05 PM #84036
    Nobody

    When you build an algo that works on any Indice correlated or not . Curve fitting is an economical virus , will make your finances sick  .

    11/02/2018 at 12:54 PM #84037
    Leo

    Is this a question?

    a small talk?

     

    11/05/2018 at 10:53 AM #84161
    Nobody
    Leo wrote:

    small talk

    Pretty well .   Although to do have some solid ideas i might lay down

     

    Adhere to a philosophy of quantitative integrity, whereby an idea must not only be supported by empirical data, but must also be grounded in sound theory

     

    This might lead somewhere and it might not

    11/05/2018 at 11:08 AM #84162
    GraHal
    Brisvegas wrote:

    solid ideas i might lay down

    We look forward to reading your ideas!

    11/05/2018 at 3:03 PM #84176
    jebus89

    This is my approach to avoid curvefitting, pls let me know what u think.

     

    https://www.prorealcode.com/topic/how-i-create-strategies-and-optimize/

    1 user thanked author for this post.
    avatar Leo
    11/05/2018 at 3:19 PM #84179
    Vonasi

    What is a ‘trully’ as it is not a word I can find in any dictionary?

    Is it ‘truly’ but said with an Australian accent? 🙂

    11/05/2018 at 4:40 PM #84189
    Nobody
    Brisvegas wrote:

    and it might not

    and it wont

    11/05/2018 at 5:33 PM #84194
    Vonasi
    Brisvegas wrote:

    and it wont

    It won’t or it trully wont?

    Every strategy we create is curve fitted due to the fact that we only have history to base our theories on – so we fit our ideas to that history. If we have created a strategy that works on various indices in back testing then that fact tells us absolutely nothing about whether it is curve fitted or not. Forward testing in the real world with real spreads and real world market moving events is the only way to establish how curve fitted a strategy is. The longer the test the better but at some point we have to decide it is or isn’t working. The very next day the market could decide to prove the opposite but it is called spread ‘betting’ for a reason so we have to accept that risk as part of the game.

    11/05/2018 at 8:58 PM #84206
    Nobody

    Hey NIcholas do you really think this is how a moderator should operate and if so delete my account please  , i cant be part of this place if your moderator here is a troll , thank you  .

    11/05/2018 at 9:06 PM #84208
    Leo

    Hey Brisvegas, calm down. We just read his opinion. The opinion of Vonasi is: everything is curve fitting.

    I still thinking about the topic. My approach is to change to artificial intelligent that look back and learn continuously but of course that is super-curve fitting. Curve fitting that evolve in real time though.

     

     

    1 user thanked author for this post.
    avatar Vonasi
    11/05/2018 at 10:22 PM #84213
    Vonasi
    Brisvegas wrote:

    When you build an algo that works on any Indice correlated or not . Curve fitting is an economical virus , will make your finances sick  .

    You started a topic with the above two random sentences and when someone bothers to reply with an opinion or a thought to try to get the discussion going for you you then call it trolling.

    It seems that you have failed to spot the smiley face after my ‘trully’ question.

    I will highlight your request regarding your account to Nicolas.

    11/05/2018 at 10:43 PM #84215
    Vonasi
    Leo wrote:

    My approach is to change to artificial intelligent that look back and learn continuously but of course that is super-curve fitting. Curve fitting that evolve in real time though.

    It certainly is curve fitting – not sure about the super though. A strategy that adapts is still adapting based on history. It is trying to fit itself to history. It does not know how important recent history is compared to much older history. It could put the emphasis of importance on very recent history and then suddenly there is a market crash similar to one twenty years ago. Suddenly really old history is more relevant than the recent history – but if we let the strategy give equal importance to all history then it is just doing what we do when we write a basic strategy and back test it on all the data we have. So just deciding on the importance of the age of history that we use for our adaptive strategy to use for its decision making is curve fitting. Not so super – but definitely curve fitted as is every strategy. As I have always said – our job is to curve fit as lightly as possible – but we will always curve fit.

    11/05/2018 at 11:36 PM #84218
    Leo

    Machine learning will take more relevance in recent data because new data is feeding the algorithm constantly.

    Which I like, by the way, because I think the way of trading has changed so much compare to several years ago.

  • Author
    Posts
Viewing 13 posts - 1 through 13 (of 13 total)

Find exclusive trading pro-tools on ProRealCode Marketplace

Similar topics:
you don’t know what you don’t know!
attached orders not closing when order bar is closed
auto find time when exchange open and close?
avoid return of a value when “0”
backtests not saving results when platform closed and reopened
buy /sell when the price crosses over/under a value and equals a different value
can you add compounding to a strategy?
cann’t you use a ttrailingstop and a stoploss together?
capital curve as an indicator?
charts look terrible when zoomed out

Create your free account now and post your request to benefit from the help of the community
Register or Login