Backtest RHS shows false sell signal

• 08/15/2018 at 2:19 AM #78214
  capngoggles

  Hi all,

  I am not sure if I have my thinking wrong or not.

  I have a simple backtest built off an indicator and it shows the black program termination mark on the very last bar (RHS) and a sell (X), whether the indicator

  is in sell or not.

  To me this is a false and misleading signal, and I have to look at the indicators to determine if it hasnt just flagged a real sell or not. I dont why it would do that.

  Any assistance would be greatly appreciated.

  Although the following code uses a CALL, the problem exists on other backtests as well using standard indicators.

  Thanks for any assistance you can provide

  John

  // Definition of code parameters DEFPARAM CumulateOrders = False // Cumulating positions deactivated // Conditions to enter long positions ignored, indicator1 = CALL JBWaveTrendDiff c1 = (indicator1 > 0) IF c1 THEN BUY 5000 CASH AT MARKET ENDIF // Conditions to exit long positions ignored, indicator2 = CALL JBWaveTrendDiff c2 = (indicator2 < 0) IF c2 THEN SELL AT MARKET ENDIF

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

  // Definition of code parameters DEFPARAM CumulateOrders = False // Cumulating positions deactivated   // Conditions to enter long positions ignored, indicator1 = CALL JBWaveTrendDiff c1 = (indicator1 > 0)   IF c1 THEN BUY 5000 CASH AT MARKET ENDIF   // Conditions to exit long positions ignored, indicator2 = CALL JBWaveTrendDiff c2 = (indicator2 < 0)   IF c2 THEN SELL AT MARKET ENDIF

  

  capngoggles

  Participant

  Topics: 2

  Replies: 5

  Been thanked: 2 times
  08/15/2018 at 3:33 AM #78218
  robertogozzi

  To write code, please use the <> “insert PRT code” button, to make code easier to read. Thank you.

  For anyone to be of help, please post the code for the CALLed indicator or modify the code using a standard indicator.

   

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 16469

  Been thanked: 3643 times
  08/15/2018 at 5:50 AM #78220
  capngoggles

  Hi Roberto,

  sorry about the code.

  Indicator code is below:

  Wavetrend indicator code

  n1 = 10 //Channel Length n2 = 21 //Average Length ap = medianprice esa = exponentialaverage[n1](ap) apesa = abs(ap - esa) d = exponentialaverage[n1](apesa) ci = (ap - esa) / (0.015 * d) tci = exponentialaverage[n2](ci) wt1 = tci wt2 = average[4](wt1) RETURN 0 as "0", (wt1-wt2)*20 coloured(0,0,200) as "difference"

  1 2 3 4 5 6 7 8 9 10 11 12 13 14

  n1 = 10 //Channel Length n2 = 21 //Average Length   ap = medianprice esa = exponentialaverage[n1](ap) apesa = abs(ap - esa) d = exponentialaverage[n1](apesa) ci = (ap - esa) / (0.015 * d) tci = exponentialaverage[n2](ci)   wt1 = tci wt2 = average[4](wt1)   RETURN 0 as "0", (wt1-wt2)*20 coloured(0,0,200) as "difference"

   

  

  capngoggles

  Participant

  Topics: 2

  Replies: 5

  Been thanked: 2 times
  08/15/2018 at 10:06 AM #78241
  robertogozzi

  It works, as far as I understand what it does, I tested it on a Daily TF on a coulple of stocks as from attached pics. I embedded the indicator in the strategy, but this shouldn’t make any difference. I used GRAPH to debug the strategy:

  DEFPARAM CumulateOrders = False // Cumulating positions deactivated n1 = 10 //Channel Length n2 = 21 //Average Length ap = medianprice esa = exponentialaverage[n1](ap) apesa = abs(ap - esa) d = exponentialaverage[n1](apesa) ci = (ap - esa) / (0.015 * d) tci = exponentialaverage[n2](ci) wt1 = tci wt2 = average[4](wt1) //RETURN 0 as "0", (wt1-wt2)*20 coloured(0,0,200) as "difference" // Conditions to enter long positions indicator1 = (wt1-wt2)*20 c1 = (indicator1 > 0) c2 = (indicator1 < 0) IF c1 THEN BUY 5000 CASH AT MARKET ENDIF // Conditions to exit long positions IF c2 THEN SELL AT MARKET ENDIF graph indicator1

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

  DEFPARAM CumulateOrders = False // Cumulating positions deactivated n1 = 10 //Channel Length n2 = 21 //Average Length ap = medianprice esa = exponentialaverage[n1](ap) apesa = abs(ap - esa) d = exponentialaverage[n1](apesa) ci = (ap - esa) / (0.015 * d) tci = exponentialaverage[n2](ci) wt1 = tci wt2 = average[4](wt1) //RETURN 0 as "0", (wt1-wt2)*20 coloured(0,0,200) as "difference" // Conditions to enter long positions indicator1 = (wt1-wt2)*20 c1 = (indicator1 > 0) c2 = (indicator1 < 0) IF c1 THEN    BUY 5000 CASH AT MARKET ENDIF // Conditions to exit long positions IF c2 THEN    SELL AT MARKET ENDIF graph indicator1

  Did you set your initial capital large enough to accomodate all losses?

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 16469

  Been thanked: 3643 times
  08/15/2018 at 5:06 PM #78295
  Nicolas

  This is a “normal” behavior if you backtest a strategy on a stock you haven’t the real time subscription. Is it the case? 😉

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4739 times
  08/15/2018 at 11:37 PM #78330
  capngoggles

  Hi Nicolas and thanks,

  I dont recall it being normal some time ago with the end of day version on other backtests I had. I think I only started seeing it with 10.3 perhaps?

  I wouldnt mind the black stop line signalling the end of run, its the sell signal that is just plain incorrect logic. It should not flag a sell unless the logic says to sell, not the end of program.

  Even if it isnt real time, it shouldnt give misleading signals which are effectively telling you to close out each day. I am trying to use the backtest to work out what I do on the next day, not much use if it constantly tells me every day to close out open positions.

  I dont need real time, since I work the ASX and its not available real time through PRT but I shouldnt expect different behavoiur in the two either.

  Thanks for your help

  John

  

  capngoggles

  Participant

  Topics: 2

  Replies: 5

  Been thanked: 2 times
  08/15/2018 at 11:49 PM #78331
  capngoggles

  Thank you Roberto for your effort in setting it up and testing it. The stock I used as an example the capital went up over the period, not down so running out of funds was not an issue.

  As per Nicolas post, It seems realtime works differently to end of day. As I said in response to Nicolas, it would be better with the sell signal not being there.

  End of run line OK, but an end of run shouldnt force a close out of position. I think that is an error.

  Thanks again.

   

  1 user thanked author for this post.

  robertogozzi

  

  capngoggles

  Participant

  Topics: 2

  Replies: 5

  Been thanked: 2 times
• Author
  Posts