Error in back test product

• 04/29/2017 at 11:44 AM #33939
  njames879

  Hi, I think that my back test product is showing an error.  I’d be interested in other people’s views as to whether I’m right, or whether I’m making some fundamental error.

  I’ve got a system that is really simple, though it works off a fairly complicated indicator.  Here’s the code

  Indicator1 = CALL "Basic system 1 + Bollinger"(close) // Conditions to enter long positions IF Indicator1 = 1 THEN BUY 1 CONTRACTS AT MARKET SET STOP pLOSS 27 SET TARGET pPROFIT 30 ENDIF // Conditions to enter short positions IF Indicator1 = -1 THEN SELLSHORT 1 CONTRACTS AT MARKET SET STOP pLOSS 27 SET TARGET pPROFIT 30 ENDIF

  1 2 3 4 5 6 7 8 9 10 11 12 13

  Indicator1 = CALL "Basic system 1 + Bollinger"(close) // Conditions to enter long positions IF Indicator1 = 1 THEN BUY 1 CONTRACTS AT MARKET SET STOP pLOSS 27 SET TARGET pPROFIT 30 ENDIF // Conditions to enter short positions IF Indicator1 = -1 THEN SELLSHORT 1 CONTRACTS AT MARKET SET STOP pLOSS 27 SET TARGET pPROFIT 30 ENDIF

  So, the issue arises when an existing position is in place, and a signal is generated in the opposite direction to the existing position, (ie, i’ve got a long position and a signal is generated which says I should open a short position.   My intention was that, in this case, the system should simply close out the existing position and wait for the next opportunity.  However, the live system and pro-back test treat this differently.  Here’s a  case, (AUD/USD 5 minutes) where this happened. (first attachment).  When I ran this in live against an IG spreadbetting account, the system behaved as I would expect, the system opened a position at 11:05 and closed it out at 15:40, I’ve attached the transaction history, and you can see that the sell order is opened but immediately nets out against the long position and both then close.

  However!  when I run a pro-back test over the same period, it not only closes the long position, but it also opens a new short position, giving entirely different results to the live situation.  I’ve attached the picture of the pro-back test, and also the detailed results.

  Am I being stupid, or is this actually an error in pro-back test.

  Any input very gratefully received.

  

  njames879

  Participant

  Topics: 5

  Replies: 7

  Been thanked: 2 times
  04/29/2017 at 1:03 PM #33947
  Maz

  Hiya. Without seeing the indicator code I can’t help. But one or two things are springing to mind.

  1. Make the logic robust by testing if on market / if longOnMarket / if not shortOnMarkrt etc just to avoid any ambiguity. Explicitly say when it is and when it is not allowed to flip direction. Try elsif where appluckable
  2. Avoid call. Import your indicator to the system code.

   

  

  Maz

  Participant

  Topics: 14

  Replies: 121

  Been thanked: 139 times
  04/29/2017 at 2:23 PM #33967
  Nicolas

  Agree with Maz, you could also GRAPH Indicator1 in backtest to see how really the signals are read by your code.

  My first idea would be to completely include your indicator into your automatic strategy code and make new comparison.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4775 times
  04/29/2017 at 3:50 PM #33975
  njames879

  thanks both.  I’ll try the ideas out.  much obliged for the input

  

  njames879

  Participant

  Topics: 5

  Replies: 7

  Been thanked: 2 times
• Author
  Posts