System any good?

Discussion in 'Strategy Building' started by pistolpt, May 25, 2012.

  1. What do you mean by that? Every system developed using historical data is an artifact of data mining. Did you mean maybe "curve-fitted"? That is a different issue. Not all systems obtained through data mining are curve-fitted.

    Your system equity is an illusion of compounding fitted profits. A high school student using Ninjatrader or Multichart, you name it, can get thousands of similar equity curves on hindsight. Just post the live trading statistics hare and we will let you know what lies ahead for you.
     
    #11     May 26, 2012
  2. jcl

    jcl

    Optimizing a strategy means selecting its rules and adjusting its parameters for maximum profit with historical price data. When you do that, no matter if manually or automatically, any backtest with the same price data that you used for optimization will give unrealistic, too positive results. This is not a negligible effect - the test results can be totally wrong. Here's a page explaining the matter:

    There are several methods for getting realistic results. The best method is Walk Forward Optimization, also explained under the link above. All serious trade platforms support it.

    For comparing strategies, use equity curves without reinvestment. Otherwise they show more the performance of your portfolio management than the performance of the strategy. That your strategy also works in life trading is good, but 7 months are not enough. A 5 years backtest, when done right, can more precisely determine a strategy performance than life trading.
     
    #12     May 27, 2012
  3. post up the p and l blotter.

    s
     
    #13     May 27, 2012
  4. Monthly Returns. Working on the Log Graph...
     
    #14     May 29, 2012
  5. =========
    Sounds like the Pistol Pete ''4 simple part ..''.may be right; but none of this is a prediction.

    A good reason why JC points tend to be right;
    the future may repeat, but not exactly.

    Its strange, but watch how often people get much better[positive] results in 5 months, rather than 5 years. Anyone is likely to get more whipsaws/sideways slop/slippage in 5 +years than 5 months.
     
    #15     May 29, 2012
  6. Ok, here is the chart in log scale.

    $10,000 initial investment.
     
    #16     Jun 13, 2012
  7. promagma

    promagma

    Are you backtesting on real VIX futures data?
     
    #17     Jun 13, 2012
  8. Yes.
     
    #18     Jun 13, 2012
  9. By the long periods of inactivity and relatively large jumps, I conclude that the number of trades is small. Too small a sample size to conclude whether the system is any good or not.

    Also, were the drawdowns accomplished by just a few trades gone bad? Then be prepared for a world of hurt...
     
    #19     Jun 14, 2012
  10. vinc

    vinc

    very good system , if it ever stops working go to the thread ' Making money with a losing
    strategy' for some answers .. either way you are a winner :)
     
    #20     Jun 14, 2012