Backtest shows wrong results

• 03/10/2023 at 5:53 PM #211325
  ProRealP

  Hello,

  I’m running a system with a partial close code that shows that it’s closing positions in backtesting, but it doesn’t happen when live. How could this happen? Spread is the same, or even lower live, but this keeps happening. The stoploss also resets itself at completely the wrong level for no apparent reason. It’s quite maddening!

  This is a revised version of my code.

  defparam cumulateorders = false DEFPARAM CumulateOrders = false DEFPARAM Preloadbars = 10000 OK = time >= 153000 AND time <= 220000 // if OK and onmarket then set stop loss sl endif indicator3 = AVERAGE[50] indicator4 = AVERAGE[200] c5 = indicator3 crosses over indicator4 c6 = indicator3 crosses under indicator4 //B/S If not onmarket and c5 then buy positionsize contract at market set stop loss sl endif If not onmarket and c6 then sellshort positionsize contract at market set stop loss sl endif //SL TIMEFRAME(1hour) sl = PositionPrice + AverageTrueRange[14](close)*pipsize timeframe(default) //MaTrailing minDistance = 12 If LongOnMarket then TP = Close + average[450](close) If TP > 0 then SELL AT min(close-minDistance*pointsize,average[450]) stop Endif Endif If ShortOnMarket then TP = Close - average[450](close) If TP > 0 then EXITSHORT AT min(close-minDistance*pointsize,average[450]) stop Endif Endif //PartialClose timeframe(2hour) // partial close ONCE partialclose = 1 ONCE PerCent = 0.25 //25% = positions to close ONCE PerCentGain = 0.005 //0.5% increase ONCE MinLotSize = 0.5 //0.5 lots minimum ONCE Increments = 1 // if partialclose and OnMarket then ExitQuantity = max(ExitQuantity[1],abs(CountOfPosition) * PerCent) LeftQty = max(MinLotSize,abs(CountOfPosition) - ExitQuantity) CloseQuantity = abs(CountOfPosition) - LeftQty CloseQuantity2 = abs(CountOfPosition) - LeftQty else CloseQuantity = 0 Increments = 1 endif // IF LongOnMarket and close >= (PositionPrice + AverageTrueRange[14](close)*pipsize) AND Flag THEN SELL CloseQuantity Contracts AT Market Flag = 0 Increments = Increments + 1 endif IF ShortOnMarket and close <= (PositionPrice - AverageTrueRange[14](close)*pipsize) AND Flag2 THEN exitshort CloseQuantity2 Contracts AT Market Flag2 = 0 Increments = Increments + 1 endif IF Not OnMarket THEN Flag = 1 Flag2 = 1 ENDIF TIMEFRAME(default) //BE if flag = 0 or flag2 = 0 then set stop breakeven endif

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101

  defparam cumulateorders = false DEFPARAM CumulateOrders = false DEFPARAM Preloadbars = 10000   OK = time >= 153000 AND time <= 220000   //   if OK and onmarket then set stop loss sl endif   indicator3 = AVERAGE[50] indicator4 = AVERAGE[200]   c5 = indicator3 crosses over indicator4 c6 = indicator3 crosses under indicator4   //B/S   If not onmarket and c5 then buy positionsize contract at market set stop loss sl endif   If not onmarket and c6 then sellshort positionsize contract at market set stop loss sl endif       //SL   TIMEFRAME(1hour) sl = PositionPrice + AverageTrueRange[14](close)*pipsize timeframe(default) //MaTrailing   minDistance = 12 If LongOnMarket then TP = Close + average[450](close) If TP > 0 then SELL AT min(close-minDistance*pointsize,average[450]) stop Endif Endif If ShortOnMarket then TP = Close - average[450](close) If TP > 0 then EXITSHORT AT min(close-minDistance*pointsize,average[450]) stop Endif Endif   //PartialClose   timeframe(2hour) // partial close ONCE partialclose = 1 ONCE PerCent      = 0.25             //25%  = positions to close ONCE PerCentGain  = 0.005            //0.5% increase ONCE MinLotSize   = 0.5              //0.5  lots minimum ONCE Increments   = 1 // if partialclose and OnMarket then ExitQuantity   = max(ExitQuantity[1],abs(CountOfPosition) * PerCent) LeftQty        = max(MinLotSize,abs(CountOfPosition) - ExitQuantity) CloseQuantity  = abs(CountOfPosition) - LeftQty CloseQuantity2  = abs(CountOfPosition) - LeftQty else CloseQuantity  = 0 Increments     = 1 endif //     IF LongOnMarket and close >= (PositionPrice + AverageTrueRange[14](close)*pipsize)  AND Flag THEN SELL CloseQuantity Contracts AT Market Flag = 0 Increments = Increments + 1 endif   IF ShortOnMarket and close <= (PositionPrice - AverageTrueRange[14](close)*pipsize)  AND Flag2 THEN exitshort CloseQuantity2 Contracts AT Market Flag2 = 0 Increments = Increments + 1 endif   IF Not OnMarket THEN Flag = 1 Flag2 = 1 ENDIF   TIMEFRAME(default)   //BE   if flag = 0 or flag2 = 0 then set stop breakeven endif

  I’ve attached a couple of pictures that show this issue when the system is live. As you can see in the first picture the system should’ve sold a portion of the position and then moved the stop to breakeven. Instead I logged in to see it had set the stop loss to a completely different, made up level faaar away. I closed it manually. In the second picture you can see that when live the system doesn’t close the positions at nearly the same level as in back test. How is this possible?

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/10/2023 at 6:04 PM #211328
  ProRealP

  I found another example. Again the system should’ve partially closed the open position in the afternoon, but didn’t, and instead closed the entire position when it hit the ma 450. Again, costing me money.

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/10/2023 at 8:10 PM #211332
  phoentzs

  Isn’t the MA450 in the TF 1hour completely somewhere else at the moment or am I misunderstanding your code?

  

  phoentzs

  Participant

  Topics: 86

  Replies: 962

  Been thanked: 200 times
  03/10/2023 at 9:00 PM #211333
  ProRealP

  The system runs on 1m TF, on us tech 100(nasdaq). Not sure of the relevance though tbh. The question is, why does the backtest act completely differently to live, when spread etc. are the same.

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/10/2023 at 9:07 PM #211334
  ProRealP

  Misread your reply. Yes, it’s probably somewhere else at the moment. Again, that’s not relevant?

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/10/2023 at 9:13 PM #211336
  fifi743

  Je voudrais le tester : indice ou forex et quelle unité ?

  

  fifi743

  Participant

  Topics: 10

  Replies: 940

  Been thanked: 444 times
  03/10/2023 at 9:19 PM #211337
  ProRealP

  I don’t speak french.

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/10/2023 at 9:46 PM #211338
  fifi743

  I would like to test it: indice or forex and which unit?

  

  fifi743

  Participant

  Topics: 10

  Replies: 940

  Been thanked: 444 times
  03/11/2023 at 7:50 AM #211348
  ProRealP

  Nasdaq 1m

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/11/2023 at 10:50 AM #211351
  JS

  Hi @ProRealPierre,

  I’ve only looked at your code superficially but what strikes me is that your Stop Loss is reposted every time by:

  If OK and OnMarket then

  Set Stop Loss SL

  EndIf

  The actual SL is calculated with:

  TimeFrame(1 Hour)

  SL = PositionPrice + AverageTrueRange[14](Close) * PipSize

  So every hour your SL will change and be reinserted… (is this the intention?)

  On the partial closure of positions:

  I don’t know any better that partial closure only works in “backtesting” and not in “live” at IG… but maybe this has changed…???

  

  JS

  Participant

  Topics: 27

  Replies: 1815

  Been thanked: 885 times
  03/11/2023 at 10:57 AM #211352
  ProRealP

  Hello @JS,

  Yes, that’s the intention, though I don’t mind changing it if it fixes the problem with the SL. See the first picture I’ve attached to this post. It show’s the SL yesterday being put at 21082, though according to the backtest it should’ve been set to roughly 12010.

  That’s changed then, yes. See the second attached picture which shows another system of mine partially closing the position as it should.

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/11/2023 at 11:42 AM #211363
  PeterSt

  Glancing over your code … it is full with ambiguities.  This really can’t work. Like JS said, reposting the SL is a first thing I would stay far from.

  Then, you can’t have several SL commands in there in the first place because only the last one executed will be THE one and only which is active.

  Next you have pending Stops which also don’t combine with SL. Oh, it can, if you have very explicit mutual exclusive pieces of code for it. This is quite hard in itself (to make decent) and you seem not to know about it. And thus nothing works.

  Next, what you see from the Backtest, never comprises reality when MTF is in order *and* when pending orders are used (Limit, Stop). Those orders only fall on the main TF limit. However, in practice it will just work fine (in Live). So this is “interpretation” as such (the gain reported by the backtest will act upon the correct price levels of exit (and entry))

  Lastly – and a positive one – you can partially close orders as much as you want. But, do it the correct way (I have no judgment on your code regarding this – I just did not look for it).

  …

  Now tell me this is on PRT-IB and someone might delete my post and I’d have to create a new one. ‘-)

  

  PeterSt

  Participant

  Topics: 73

  Replies: 2555

  Been thanked: 837 times
  03/11/2023 at 12:43 PM #211366
  ProRealP

  Thanks for your reply, @PeterSt!

  I see your point. I’ll change the stop to sell at at stop instead of setting the stop loss, and see if that works better.

  Appreciate the help!

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/12/2023 at 12:11 PM #211409
  ProRealP

  @robertogozzi, or @Nicolas, do you have any idea as to why my system won’t partially close positions? I could understand slippage being the reason in cases where it just touches the level, but that doesn’t look to always be the case(look at the picture “ex3”, for example). Sometimes it just doesn’t partially close. There is seemingly nothing wrong with the code, right? I mean, sometimes it works..

  

  ProRealP

  Participant

  Topics: 7

  Replies: 47

  Been thanked: 2 times
  03/12/2023 at 12:53 PM #211412
  phoentzs

  Would you trust your money to a code that sometimes works? 😉

  

  phoentzs

  Participant

  Topics: 86

  Replies: 962

  Been thanked: 200 times
• Author
  Posts