ProOrder Exception Line Number

• 07/27/2020 at 4:30 PM #140221
  Ξ3

  Hi All,

  I have a fairly complex piece of automated code running on IG ProRealTime 10.3 that on the odd occasion comes up with a divide by zero error when using multi-timeframe… I suspect it might be due to a “phantom” bar but have gone through my code in detail yet I am unable to identify where the problem is sneaking in.

  Is there any way to get a rough idea of which line in the code causes this exception (or at least somehow narrow it down a bit)? I can’t seem to find any indication on any error message? Of course it works perfectly during back-testing 😉

  Any help or ideas would be greatly appreciated.

  

  Ξ3

  Participant

  Topics: 12

  Replies: 26

  Been thanked: 3 times
  07/27/2020 at 4:47 PM #140225
  Nicolas

  Check all divisions in your code, is there any possibility that the divisor can be equal to 0? If yes, you have your answer! 😉

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4742 times
  07/27/2020 at 6:08 PM #140228
  Ξ3

  Good advice Nicholas but I have gone through my code and put IF statements around all divisors to protect against zero (or at least that’s what I thought). I guess the answer is that there is no place find the actual offending code or line number in any error message? 🙁

   

  

  Ξ3

  Participant

  Topics: 12

  Replies: 26

  Been thanked: 3 times
  07/28/2020 at 7:01 AM #140251
  Dow Jones

  I do not know any other “exception” way to debug this issue. PRT is more like a simple scripting language thus it doesn’t offer many features from the common programming language (e.g. exception, function block, etc…) although it keeps improving.

  From  my experience on one occurrence for division zero issue (which for some reasons back test doesn’t catch it), you can pay attention to the timing when the zero division error message happened (PRT should send you a notification when it happen) and from there, you can observe from chart what information has gone wrong.

  Like below, I have to add condition “IF bolltrendUP = bolltrendDOWN THEN”, otherwise, they can end up with same value and when going to (bolltrendUP – bolltrendDOWN), it is zero. I only find it by focusing on the timing when the error happens and I put up the indicator and see for myself.

  It happened on MTF as well for me, the bolltrendperiod and strend were rather small thus resulted such issue occasionally.

  bolltrendMA = average[bolltrendperiod, 1](close)//50,1 STDDEVtrend = STD[bolltrendperiod] bollstrend = strend * STDDEVtrend bolltrendUP = bolltrendMA + bollstrend bolltrendDOWN = bolltrendMA - bollstrend IF bolltrendUP = bolltrendDOWN THEN bolltrendPercent = 50 ELSE bolltrendPercent = 100 * (close - bolltrendDOWN) / (bolltrendUP - bolltrendDOWN) ENDIF

  1 2 3 4 5 6 7 8 9 10

  bolltrendMA = average[bolltrendperiod, 1](close)//50,1 STDDEVtrend = STD[bolltrendperiod] bollstrend = strend * STDDEVtrend bolltrendUP = bolltrendMA + bollstrend bolltrendDOWN = bolltrendMA - bollstrend IF bolltrendUP = bolltrendDOWN THEN bolltrendPercent = 50 ELSE bolltrendPercent = 100 * (close - bolltrendDOWN) / (bolltrendUP - bolltrendDOWN) ENDIF

   

  

  Dow Jones

  Participant

  Topics: 10

  Replies: 135

  Been thanked: 75 times
  07/28/2020 at 7:18 AM #140253
  Dow Jones

  Out of curiosity, I tested with this simple code, the back test doesn’t give me any error. When 4 / 0 happens, C2 just equal to 0 and test continue to next trade.

  So back test cannot catch division zero…maybe can be an improvement to send to PRT

  // Conditions to enter long positions MA20 = ExponentialAverage[20](close) IF barindex > 0 AND barindex mod 2 = 0 THEN a = 2 ELSE a = 0 ENDIF C1 = close > MA20 C2 = 4 / a = 2 IF NOT LongOnMarket AND C1 AND C2 THEN BUY 1 CONTRACTS AT MARKET ENDIF // Conditions to exit long positions If LongOnMarket AND barindex - tradeindex = 70 THEN SELL AT MARKET SET STOP %LOSS 5 ENDIF GRAPH a GRAPH C1 GRAPH C2

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

  // Conditions to enter long positions MA20 = ExponentialAverage[20](close) IF barindex > 0 AND barindex mod 2 = 0 THEN a = 2 ELSE a = 0 ENDIF   C1 = close > MA20 C2 = 4 / a = 2 IF NOT LongOnMarket AND C1 AND C2 THEN BUY 1 CONTRACTS AT MARKET ENDIF   // Conditions to exit long positions If LongOnMarket AND barindex - tradeindex = 70 THEN SELL AT MARKET SET STOP %LOSS 5 ENDIF   GRAPH a GRAPH C1 GRAPH C2

   

  1 user thanked author for this post.

  Ξ3

  

  Dow Jones

  Participant

  Topics: 10

  Replies: 135

  Been thanked: 75 times
  07/28/2020 at 7:33 AM #140256
  Ξ3

  Dow Jones,

  Your idea of comparing the failure time with the various MTF chart timeframes used may just do the trick for me.

  Many thanks for the detailed response!

  

  Ξ3

  Participant

  Topics: 12

  Replies: 26

  Been thanked: 3 times
  07/28/2020 at 12:29 PM #140295
  juanj

  Have also had my fair share of this problem. Something I have picked up is that if division happens with a variable that is perhaps undefined at any particular stage you will also get that error.

  Can be mitigated by using prober initiation parameters i.e. once

  2 users thanked author for this post.

  GraHal, Ξ3

  

  juanj

  Participant

  Topics: 51

  Replies: 505

  Been thanked: 170 times
  07/28/2020 at 12:34 PM #140296
  Ξ3

  Worth looking into… Thanks JuanJ

  

  Ξ3

  Participant

  Topics: 12

  Replies: 26

  Been thanked: 3 times
• Author
  Posts