How much backtesting?

Discussion in 'Strategy Building' started by prox, May 11, 2003.

  1. prox


    How many backtest trades might be considered adequate for a 1 minute scalp system? I am just eyeballing it over past charts and just wondering what others think is enough to validate results.
  2. DT-waw


    I like your signature "The only way to make a small fortune in trading is to begin with a large fortune".
  3. Foz


    At least a few dozen trades are necessary to get enough statistics to validate a system. And note that these are objective, non-optimized trades with transaction costs applied

    With just eyeballing charts your brain is going to have a preference for picking out the profitable trades. Try to backtest your system mechanically.

    Also, if you keep varying your parameters or setup pattern until you get a system that looks good historically, be sure to test it on out of sample data to avoid curve fitting.
  4. At least 100.

    Also check out the streaks, i mean the winner/losers in a row. Personally, a trading system with 25 losers in a row would scare me.
  5. Yes, at least 100...
  6. prox


    Thanks a lot. :D
  7. For a short term scalping system, 100 may be enough,
    but not if they all occur within a single week :D

    I would trade several hundred, but takes samples from
    several years and different seasons, etc.


  8. What if 25 in a row was "normal" for the system and it
    had a 500% return a year? :D



  9. prox


    Then you'd have all the money in the market soon enough..

    or until your method gets shared and exploited where your edge gets neutralized.
  10. Yes yes yes.... but my point is, systems come in all kinds
    of flavors, and a system with a 25 in a row losing streak
    should not automatically be disqualified. Money management
    can sometimes solve issues like this.

    I currently am working on a system with ONE big draw down
    dip which im trying real hard to get rid of. Although not
    acceptable in its current form, its totally acceptable if
    I only risked 1/4 of my capital.



    #10     May 12, 2003