Pathfinder Trading System

• 03/29/2018 at 2:25 PM #66697
  GraHal

  Vonasi wrote:

  Writing codes and testing them is a very good educational tool

  I agree, a good point well made (as usual) but I do think that newbie (or not so newbie coders can get sucked in by the optimiser and it can take a long long time coding, backtesting, demo running to understand the clear signs of curve fitting. Maybe the  Topic Aloysius proposed might shorten the learning curve.

  I’ll put my hand up and say the Optimiser sucked me in for long enough … and I understood about curve fitting and the signs … hell it still sucks me in especially after many hours coding! I think because I want to see results / some £££s for my time and so I launch a System (sometimes even in Live) and then wish I never had! 🙂

  Hey we should start new Threads on these interesting Topics under Trading Discussions?? It would mean we don’t hijack Threads on the coding Support Forum? Apologies readers!

  1 user thanked author for this post.

  robertogozzi

  

  GraHal

  Participant

  Topics: 211

  Replies: 8774

  Been thanked: 2180 times
  03/29/2018 at 2:32 PM #66699
  Despair

  I think the main problem of most of the strategies here is that they are optimized over the whole available data. One must understand that this is just producing fantasy results. Either WFA or at least an OOS test is absolutely mandatory for having a result that is somehow useful. I pointed out so several times to Reiner but he always replied that he had no time to do so.

  BTW I agree that we should start a new thread. This discussion is not really a pathfinder discussion anymore.

  

  Despair

  Blocked

  Topics: 97

  Replies: 1034

  Been thanked: 218 times
  03/29/2018 at 2:43 PM #66700
  Vonasi

  GraHal wrote:

  and so I launch a System (sometimes even in Live) and then wish I never had!

  I never put a strategy live straight away – I need to build confidence in it first. I am very very cautious about putting any code live (at the moment I have exactly zero codes running live while I wait for the markets to sort themselves out and get rid of all the noise we currently have).

  I am actually very happy if a strategy fails in demo. First of all I have lost no money (half the battle is won) and secondly I can re-analyse the strategy and work out why it went wrong  – which is a free lesson in how not to do it.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1811 times
  03/29/2018 at 4:54 PM #66712
  Aloysius

  I have create a new topic to talk about codes “one year after”.

  

  Aloysius

  Participant

  Topics: 1

  Replies: 44

  Been thanked: 16 times
  04/03/2018 at 11:12 PM #67078
  Gianluca

  Hello guys, maybe i’ve found a new way to filter the strategy, here is the result for now, what do you think?

  (the multiplier and the position size are the same)

   

  

  1 user thanked author for this post.

  JohnScher

  

  Gianluca

  Participant

  Topics: 64

  Replies: 417

  Been thanked: 55 times
  04/04/2018 at 12:13 PM #67132
  JohnScher

  Pls, can you put the code in here or give as attachment?

  

  JohnScher

  Participant

  Topics: 44

  Replies: 336

  Been thanked: 85 times
  04/08/2018 at 10:06 PM #67565
  Gianluca

  JohnScher wrote:

  Pls, can you put the code in here or give as attachment?

  Sure, at the beginning i was jelous of it, but in this forum i found so much help and new idea that i want to share mine too!

  I hope we can work on it ad see if these could help us!

   

  FILTERS

  ////FILTRI long ONCE LL = -160 ONCE SS = 26 //once mioL = var //filtri short once AA = -140 ONCE BB = 55 mac= MACDline[12,26,9](close) ONCE P = 5 ONCE COEFF = 0.7 ///codici filtri REM Calcolo della Media Mobile MME3 = ExponentialAverage[P](ExponentialAverage[P](ExponentialAverage[P](close))) MME4 = ExponentialAverage[P](MME3) MME5 = ExponentialAverage[P](MME4) MME6 = ExponentialAverage[P](MME5) REM Calcolo dei coefficienti ONCE c1 = SQUARE(coeff)*coeff ONCE c1 = -c1 ONCE c2 = 3*SQUARE(coeff)-3*c1 ONCE c3 = -2*c2 -3*c1 -3*coeff ONCE c4 = 1 + 3*coeff + c2 +2*c1 Tilson = c1*MME6 + c2*MME5 + c3*MME4 + c4*MME3//////////// c02=R2[50](Tilson)-r2[50](average[50]) filtrolong= (mac > LL) and (mac < SS) and (c02 < 0.03)and (c02>-0.49)//and c1//AND ff2 and c1 //AND ff3 filtroshort= (mac > AA) and (mac < BB) //and c2// and ss1 and c2// AND ss2

  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

  ////FILTRI long ONCE LL               = -160 ONCE SS               = 26 //once mioL             = var //filtri short once AA               = -140 ONCE BB               = 55 mac= MACDline[12,26,9](close) ONCE P            = 5 ONCE COEFF        = 0.7 ///codici filtri REM Calcolo della Media Mobile MME3 = ExponentialAverage[P](ExponentialAverage[P](ExponentialAverage[P](close))) MME4 = ExponentialAverage[P](MME3) MME5 = ExponentialAverage[P](MME4) MME6 = ExponentialAverage[P](MME5)   REM Calcolo dei coefficienti ONCE c1 = SQUARE(coeff)*coeff ONCE c1 = -c1 ONCE c2 = 3*SQUARE(coeff)-3*c1 ONCE c3 = -2*c2 -3*c1 -3*coeff ONCE c4 = 1 + 3*coeff + c2 +2*c1 Tilson = c1*MME6 + c2*MME5 + c3*MME4 + c4*MME3//////////// c02=R2[50](Tilson)-r2[50](average[50]) filtrolong= (mac > LL) and (mac < SS) and (c02 < 0.03)and (c02>-0.49)//and c1//AND ff2 and c1 //AND ff3 filtroshort= (mac > AA) and (mac < BB) //and c2// and ss1 and c2// AND ss2

   

  order section need to be changed in

  // long entry with order cumulation IF ( l1 OR l4 OR l2 OR (l3 AND f2) ) AND NOT alreadyReducedLongPosition THEN // check saisonal booster setup and max position size IF saisonalPatternMultiplier > 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize * saisonalPatternMultiplier) * scaleFactor) IF (COUNTOFPOSITION + numberContracts) <= maxPositionSizeLong * scaleFactor THEN IF SHORTONMARKET THEN EXITSHORT AT MARKET ENDIF if filtrolong then BUY numberContracts CONTRACT AT MARKET ENDIF endif ELSIF saisonalPatternMultiplier <> 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize) * scaleFactor) IF (COUNTOFPOSITION + numberContracts) <= maxPositionSizeLong * scaleFactor THEN IF SHORTONMARKET THEN EXITSHORT AT MARKET ENDIF if filtrolong then BUY numberContracts CONTRACT AT MARKET ENDIF ENDIF ENDIF stopLoss = stopLossLong takeProfit = takeProfitLong ENDIF // short entry with order cumulation IF ( (s1 AND f3) OR (s2 AND f1) OR (s3 AND f3) ) AND NOT alreadyReducedShortPosition THEN // check saisonal booster setup and max position size IF saisonalPatternMultiplier < 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize * ABS(saisonalPatternMultiplier)) * scaleFactor) IF (ABS(COUNTOFPOSITION) + numberContracts) <= maxPositionSizeShort * scaleFactor THEN IF LONGONMARKET THEN SELL AT MARKET ENDIF IF filtroshort THEN SELLSHORT numberContracts CONTRACT AT MARKET ENDIF ENDIF ELSIF saisonalPatternMultiplier <> 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize) * scaleFactor) IF (ABS(COUNTOFPOSITION) + numberContracts) <= maxPositionSizeShort * scaleFactor THEN IF LONGONMARKET THEN SELL AT MARKET ENDIF IF filtroshort THEN SELLSHORT numberContracts CONTRACT AT MARKET ENDIF ENDIF ENDIF stopLoss = stopLossShort takeProfit = takeProfitShort 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

  // long entry with order cumulation IF ( l1 OR l4 OR l2 OR (l3 AND f2) ) AND NOT alreadyReducedLongPosition THEN   // check saisonal booster setup and max position size IF saisonalPatternMultiplier > 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize * saisonalPatternMultiplier) * scaleFactor) IF (COUNTOFPOSITION + numberContracts) <= maxPositionSizeLong * scaleFactor THEN IF SHORTONMARKET THEN EXITSHORT AT MARKET ENDIF if filtrolong then BUY numberContracts CONTRACT AT MARKET ENDIF endif ELSIF saisonalPatternMultiplier <> 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize) * scaleFactor) IF (COUNTOFPOSITION + numberContracts) <= maxPositionSizeLong * scaleFactor THEN IF SHORTONMARKET THEN EXITSHORT AT MARKET ENDIF if filtrolong then BUY numberContracts CONTRACT AT MARKET ENDIF ENDIF ENDIF   stopLoss = stopLossLong takeProfit = takeProfitLong   ENDIF   // short entry with order cumulation IF ( (s1 AND f3) OR (s2 AND f1) OR (s3 AND f3) ) AND NOT alreadyReducedShortPosition THEN   // check saisonal booster setup and max position size IF saisonalPatternMultiplier < 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize * ABS(saisonalPatternMultiplier)) * scaleFactor) IF (ABS(COUNTOFPOSITION) + numberContracts) <= maxPositionSizeShort * scaleFactor THEN IF LONGONMARKET THEN SELL AT MARKET ENDIF IF filtroshort THEN SELLSHORT numberContracts CONTRACT AT MARKET ENDIF ENDIF ELSIF saisonalPatternMultiplier <> 0 THEN numberContracts = MAX(1, MIN(maxPositionSizePerTrade, positionSize) * scaleFactor) IF (ABS(COUNTOFPOSITION) + numberContracts) <= maxPositionSizeShort * scaleFactor THEN IF LONGONMARKET THEN SELL AT MARKET ENDIF IF filtroshort THEN SELLSHORT numberContracts CONTRACT AT MARKET ENDIF ENDIF ENDIF   stopLoss = stopLossShort takeProfit = takeProfitShort   ENDIF

  
  @reiner what do you think?

   

  2 users thanked author for this post.

  Aloysius, JohnScher

  

  Gianluca

  Participant

  Topics: 64

  Replies: 417

  Been thanked: 55 times
  04/09/2018 at 6:30 AM #67576
  JohnScher

  Thank you very much.

  

  JohnScher

  Participant

  Topics: 44

  Replies: 336

  Been thanked: 85 times
  04/09/2018 at 8:58 AM #67582
  GraHal

  Would be great if there could be a ‘switch’ that makes max numbercontracts = 1 or 2 or 4 or user choice?

  Reason is I always Backtest at Lot Size = 1 and so to compare with other Systems I need Lot size = 1.

  I have made changes as below in 4 places and still it is sellingshort at more than 1 on some trades.

  Also if I make changes as below then I have to remember where I made the changes if I want it back again as original 🙂

  numberContracts = 1 or MAX(1, MIN(maxPositionSizePerTrade, positionSize) * scaleFactor)

  1	numberContracts = 1 or MAX(1, MIN(maxPositionSizePerTrade, positionSize) * scaleFactor)

   

   

  

  GraHal

  Participant

  Topics: 211

  Replies: 8774

  Been thanked: 2180 times
  04/09/2018 at 9:27 AM #67587
  Aloysius

  @Gianluca, be careful because your code cause positions of 60 lots when winning in the backtest, but in real life it could be very dangerous. Filtering past bad trades does not mean that it will filter future bad ones. It would be wise to watch in demo before live trading.

  

  Aloysius

  Participant

  Topics: 1

  Replies: 44

  Been thanked: 16 times
  04/09/2018 at 9:33 AM #67589
  Gianluca

  Aloysius wrote:

  @gianluca, be careful because your code cause positions of 60 lots when winning in the backtest, but in real life it could be very dangerous. Filtering past bad trades does not mean that it will filter future bad ones. It would be wise to watch in demo before live trading.

  I know, but if you can see the 60 lots are after thousand and thousand of gain. In fact the basic size is 1, and the increase is not so fast.

  Of course the risk could be changed upon the own risk.

   

  Of course it doesn’t mean that in future will not be bad trades, but we are trying to find a statistics advantage, and we can find it only working on the past.

  

  Gianluca

  Participant

  Topics: 64

  Replies: 417

  Been thanked: 55 times
  04/09/2018 at 9:51 AM #67590
  Aloysius

  Yes so if i look at the curve of the backtest, the first time it takes 60 lots (june 27, 2013), the system has won 23000€ : but if the market had unfortunatly  fallen for 350 points (which is not exceptional), all the gains were gone. But if this had happen, when backtesting now, optimization would have found other filter variable to avoid this bad trade.

  

  Aloysius

  Participant

  Topics: 1

  Replies: 44

  Been thanked: 16 times
  04/09/2018 at 11:17 AM #67683
  Gianluca

  Aloysius wrote:

  Yes so if i look at the curve of the backtest, the first time it takes 60 lots (june 27, 2013), the system has won 23000€ : but if the market had unfortunatly fallen for 350 points (which is not exceptional), all the gains were gone. But if this had happen, when backtesting now, optimization would have found other filter variable to avoid this bad trade.

  DAX

  The first big position is on march 28 2012 and was 14 size, in that date the system had also a gain of 3000€ maybe yes is too much, and in a real account is better to put less risk on, i am using it on DEMO, and there there’s no problem with size 😀

  

  Gianluca

  Participant

  Topics: 64

  Replies: 417

  Been thanked: 55 times
  04/12/2018 at 3:31 PM #68181
  JohnScher

  PATH 4H DAX V7 (filt)  Demo-Mode took yesterday a short postion at 12.284 SL 12.683 TP 11.946

  maybe the first trade will be a loss?

   

  

  JohnScher

  Participant

  Topics: 44

  Replies: 336

  Been thanked: 85 times
  04/12/2018 at 7:39 PM #68209
  Gianluca

  JohnScher wrote:

  Demo-Mode took yesterday a short postion at 12.284 SL 12.683 TP 11.946

  We shall see! 😀

  

  Gianluca

  Participant

  Topics: 64

  Replies: 417

  Been thanked: 55 times
• Author
  Posts