I did backtest my formula

Discussion in 'Trading' started by yinyang, Jul 26, 2021.

  1. taowave

    taowave

    Whatever you want it to be:)


     
    #21     Jul 28, 2021
  2. SunTrader

    SunTrader

    Is backtesting based on tick data? If not it is making all kinds of assumptions.
     
    #22     Jul 28, 2021
  3. ROFL ! :D
     
    #23     Jul 28, 2021
    tayte likes this.
  4. yinyang

    yinyang

    signals are made by close price of 4 hour candle, then wait another 4 hours for next signal. Pretty straightforward. No assumptions involved. I either open or close positions every 4 hours, or do nothing if signal says so
     
    Last edited: Jul 28, 2021
    #24     Jul 28, 2021
    janbillian likes this.
  5. ph1l

    ph1l

    Aren't you assuming your orders would have been filled and the fill prices in your backtest?
     
    Last edited: Jul 28, 2021
    #25     Jul 28, 2021
  6. SunTrader

    SunTrader

    So you are taking the exact price at the end of 4 hour bar both for your entries and exits?

    Realize even doing that you are making assumptions about fills.
     
    #26     Jul 28, 2021
  7. yinyang

    yinyang

    for the last 1 minute of 4hr candle, there are thousands of contracts filled. assumption that a couple of my contracts are filled is not so wild assumption.
     
    #27     Jul 29, 2021
  8. SunTrader

    SunTrader

    If you think so.

    Many others before thought as well until .......

    But before you shoot the messenger I've unsubscribed to save you the trouble.
     
    #28     Jul 29, 2021
    Van_der_Voort_4 likes this.
  9. yinyang, backtesting can be valuable if you don't fall into the trap over curvefitting, proabably the #1 error. A dataset for a give timeperiod gets an indicator and / or candelstick overlay, and than depending on adjustment possibilites, the setups caught the past timeperiod perfectly.

    #2 error is being too conservative on slippage assumptions. Maybe the fills would have been worse.

    To show the value of your system, you should try an out of sample test with your system and the given parameters. How would it have performed between say 2006 - 2013 ?

    On paper, it looks perfect, but you'll be hard pressed to find traders who can generate 30% per year or more in real life.
     
    #29     Jul 29, 2021
    yc47ib likes this.
  10. gkishot

    gkishot

    I disagree. There is no way around backtesting your strategy before going live.
     
    #30     Jul 29, 2021