Pathfinder _ zilliq research

• 03/03/2017 at 8:42 PM #27095
  zilliq

  Hi Arno,

  For me, it’s against my religion to average my losses, sorry.

  I think it can work ONLY if you have a very high % of winning trade (in the past, but who can predict the future)

  The problem is what I said to Reiner before on my previous post : Pathfinder take more position than it could  (depending on, the risk). If it works, no problem, you’re rich, but if you lose you cry

  I think (it’s my opinion, not a critical), that this booster to 10 positions (if I remember) is too dangerous and against the Van Tharp theory.

  This afternoon I had a little bit time to code different suggestions I spoke to Reiner

  1/ Add position ONLY if you are after the trailing stop = pyramid trading

  I code a simple strategy DAX ut15 Spread=1 and test it

  As you will see on the pictures

  With this system you have higher winning trade, higher ratio and you’re less in the market (and so less expensive) ! +8.34 % vs +7.91 %

  It need to be confirm but seems promising

  2/ Add an Alembert Martingale (to the basic code) (with maxposition depending on the risk)

  It seems you win not more and with higher drawdown and you’re more in the market

  I think it can be optimize (Trailing stop and so on)…But not sure it could be a good option +7.02 %

  I will test other suggestions when I will have time and say to you the results

  Have fun

  Zilliq

  6 users thanked author for this post.

  gforton, larouedegann, traderfred, Alco, Jesús, Robin

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 10:05 AM #27291
  gforton

  @Reiner,

  Really sorry to see you go, but I understand and respect your decision. Thank you for all the work you have done for this forum. I hope we will hear from you soon again on this forum! Take care!

   

  
  @Zilliq:

  I understand your comments and agree with your reasoning. How would you implement this in Reiner’s code (latest version) ? Can you show us a script that actually does what you propose ?

  Thanks for your efforts & greetings.

  

  gforton

  Participant

  Topics: 2

  Replies: 11

  Been thanked: 2 times
  03/06/2017 at 10:11 AM #27296
  gforton

  @Zilliq:

   

  Especially the first results you show ‘avec’ look very promising indeed. A d’Alembert can b!e very rewarding, but as you correctly point out, is you loose, you can loose BIG

  Thanks for sharing;

   

   

  

  gforton

  Participant

  Topics: 2

  Replies: 11

  Been thanked: 2 times
  03/06/2017 at 11:42 AM #27312
  xadami

  No matter, which system of risk management you choose – never forget:

  pathfinder works only with a DESASTER-STOP, which was NEVER needed during the backtest period of the 4H-systems. This does not mean, that there will not be a desaster for example like Lehman (2008) in the future. And image, what will happen to your account, if your system is trading several postions on several stock Indices at such a moment – About 15 open positions will kill your account !

  1 user thanked author for this post.

  manel

  

  xadami

  Participant

  Topics: 0

  Replies: 2

  Been thanked: 1 time
  03/06/2017 at 12:54 PM #27320
  zilliq

  @gforton

  Sorry I’m at work without PRT

  I will show you later

  I think the Martingale are very dangerous, like a weapon, but if you know use it it can be very useful

  Notably, I think they can be used ONLY with high %winning trades. Probably >70 %. And you need to manage the risk as Van Tharp do. So you can lose ONLY what you wanted (always the same amount by trade)

  You have the code of Alembert by Nicolas on the blog

  https://www.prorealcode.com/blog/practical-martingale-code-automated-trading/

  Bye

   

  Zilliq

  Ps: I don’t think I will have time as Reiner to implement all ideas on Pathfinder, because it’s not my code notably, but we can speak about Money Management that I love

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 1:01 PM #27321
  zilliq

  Other ideas to explore:

  1/ As we pyramid our positions, I will see if it is interesting to pyramid the Take profit. I don’t remember someone try this. May be I’m wrong

  2/ Other ideas: I think Saisonnality is a good thing, but very dangerous. See the DAX, if we see what happened in the 30 past years, and what happened last year, is very different

  http://www.seasonalcharts.com/future_index_dax.html

  I think that “May be” it could be more interesting to follow the %winning/losing trades and ponderate the position with it

  3/ I think a lot about MAE/MFE and there is probably something to do with is to protect our capital. I will see…

  I will try and I will say to you

  If you have other suggestions you can post here

  Bye

  Zilliq

   

  https://www.zilliqtradingresearch.fr/

  Coding is not a crime 😉

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 1:08 PM #27323
  zilliq

  Today is a bad day !

  We all know that there is some day, all goes wrong …:-(

  I think it could be interesting too to ad the code of Nicolas to stop to lose after some amount

  https://www.prorealcode.com/blog/learning/max-profit-loss-day-trading-strategy/#more-14476

  May be not for pathfinder because there is only few trades but perhaps for some intraday strategy…

  To try…

   

  https://www.zilliqtradingresearch.fr/

  Coding is not a crime 😉

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 1:38 PM #27324
  Joachim Nilsson

  I agree with you zilliq. Pathfinder is over optimized acording too me. Have someone tested it on 200k bars?

  

  Joachim Nilsson

  Participant

  Topics: 4

  Replies: 62

  Been thanked: 18 times
  03/06/2017 at 2:21 PM #27328
  zilliq

  !!!! Not a critical !!!

  The problem is, as with ALL algo that you always wanted to OverOptimizd, and so, the results are beautiful for one time, not another

  And one more time, Pathfinder open a lot of positions more than what is was defined at the beginning (5 % if I remember)

  It’s like a linear regression of the results

  I hope that the new possibility with Walk Forward will resolve a part of the problem In/Out Sample.

  The problem with the WF of PRT is double :

  1/ We have no indication to how use it…How to interpret the results, what variable do we need to use at the end …We wait for the video/post of Nicolas

  2/ It’s very very very slow and I can’t imagine to do a WF with >4 variables for 20 years (PathFinder has more than 10-15 variables…). Last week (see n other post), I do a WF with 5 variables a simple code 7000 comb  and it takes many hours…

  Zilliq

   

   

  https://www.zilliqtradingresearch.fr/

  Coding is not a crime 😉

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 3:48 PM #27336
  Paris

  hi zilliq , in your site https://www.zilliqtradingresearch.fr/

  i can see what should be very good strategies , why dont you share these with us ?

  The same way reiner give here his code and ideas for free and to keep the spirit of sharing of this

  site , i think all of us would be interressed to improve your strategies .

  sharing is not a crime …..

  

  Paris

  Participant

  Topics: 6

  Replies: 145

  Been thanked: 11 times
  03/06/2017 at 3:57 PM #27337
  zilliq

  That’s what I ask to Mark Jurik

  http://www.jurikres.com/

  But he doesn’t want to 😉

  Thanks to keep on the discussion and share our ideas on Money Management

  Bye

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 4:05 PM #27338
  Paris

  Very nice !

  ok keep sharing with yourself  !

  

  Paris

  Participant

  Topics: 6

  Replies: 145

  Been thanked: 11 times
  03/06/2017 at 5:03 PM #27352
  zilliq

  Well, I do the first tests…

  It seems that pyramid take profit doesn’t bring more…It brings more winning trade, but it cuts some, and the gain is less

  BUT it seems that it is interesting to play MAE, because it cuts big losing trade and increase the VQN

  I will post graph later

  See U

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 8:34 PM #27478
  zilliq

  Well it seems my idea was not so bad …

  Instead to ponderate the number of position by saisonnality I wanted to ponderate the number of position with the number of call or short trade when there is a loosing trade (=less lag)

  ZRatio :-)=winning/losing WHEN the trade is loosing

  The purpose is to ponderate the winning or loosing trade, with this ratio. Ponderate the call trade when short trade are more often losing and vice versa

  If there is less call trade who are loosing I will ponderate them and same with short

  For this I create a very simple code (MM crossing) UT15 Spread 1 Capital 10000

  DEFPARAM CumulateOrders=true DEFPARAM PRELOADBARS=2000 REM Achat once ordersize=1 once exitindex=-2 indicator1 = ExponentialAverage[a](close) indicator2 = Average[b](close) c1 = (indicator1 CROSSES OVER indicator2) IF c1 THEN BUY OrderSize contract ROUNDEDUP AT MARKET ENDIF REM Vente indicator3 = ExponentialAverage[c](close) indicator4 = Average[d](close) c2 = (indicator3 CROSSES UNDER indicator4) IF c2 THEN SELL AT MARKET ExitIndex=BarIndex ENDIF c3=(indicator1 CROSSES under indicator2) IF c3 THEN SELLSHORT OrderSize contract AT MARKET ENDIF c4=(indicator1 CROSSES over indicator2) IF c4 THEN EXITSHORT AT MARKET ExitIndex=BarIndex ENDIF // positions sizing IF Barindex= ExitIndex+1 THEN ExitIndex =0 IF PositionPerf(1)<0 THEN if countofposition[1]>0 then w=w+1 elsif countofposition[1]<0 then l=l+1 endif OrderSize = 1 //OrderSize+1 ELSIF PositionPerf(1)>=0 THEN OrderSize =1 //MAX(OrderSize-1,1) ENDIF ENDIF graph w coloured (10,204,10) graph l

  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

  DEFPARAM CumulateOrders=true DEFPARAM PRELOADBARS=2000 REM Achat once ordersize=1 once exitindex=-2   indicator1 = ExponentialAverage[a](close) indicator2 = Average[b](close) c1 = (indicator1 CROSSES OVER indicator2)   IF c1 THEN BUY OrderSize contract ROUNDEDUP AT MARKET ENDIF     REM Vente   indicator3 = ExponentialAverage[c](close) indicator4 = Average[d](close) c2 = (indicator3 CROSSES UNDER indicator4)   IF c2 THEN SELL  AT MARKET ExitIndex=BarIndex ENDIF     c3=(indicator1 CROSSES under indicator2)   IF c3 THEN SELLSHORT OrderSize contract  AT MARKET ENDIF   c4=(indicator1 CROSSES over indicator2)   IF c4 THEN EXITSHORT AT MARKET ExitIndex=BarIndex ENDIF   // positions sizing IF Barindex= ExitIndex+1 THEN ExitIndex =0 IF PositionPerf(1)<0 THEN if countofposition[1]>0 then w=w+1 elsif countofposition[1]<0 then l=l+1 endif   OrderSize = 1 //OrderSize+1 ELSIF PositionPerf(1)>=0 THEN OrderSize =1 //MAX(OrderSize-1,1) ENDIF ENDIF   graph w coloured (10,204,10) graph l

  Without the ponderation and the ratio we obtain with this simple strategy Gain +5.02 % VQN 1.06

  With the ponderation AND the ratio we obtain Gain +20.56 % (X4!)  VQN 1.56. Who is much better 🙂

  The code with the “Zilliq Ponderation” 🙂

   

  DEFPARAM CumulateOrders=true DEFPARAM PRELOADBARS=2000 REM Achat once ordersize=1 once exitindex=-2 if w<>0 and l<>0 then pondc=round(l/w) ponds=round(w/l) else pondc=0 ponds=0 endif indicator1 = ExponentialAverage[a](close) indicator2 = Average[b](close) c1 = (indicator1 CROSSES OVER indicator2) IF c1 THEN BUY OrderSize+pondc contract ROUNDEDUP AT MARKET ENDIF REM Vente indicator3 = ExponentialAverage[c](close) indicator4 = Average[d](close) c2 = (indicator3 CROSSES UNDER indicator4) IF c2 THEN SELL AT MARKET ExitIndex=BarIndex ENDIF c3=(indicator1 CROSSES under indicator2) IF c3 THEN SELLSHORT OrderSize+ponds contract AT MARKET ENDIF c4=(indicator1 CROSSES over indicator2) IF c4 THEN EXITSHORT AT MARKET ExitIndex=BarIndex ENDIF // positions sizing IF Barindex= ExitIndex+1 THEN ExitIndex =0 IF PositionPerf(1)<0 THEN if countofposition[1]>0 then w=w+1 elsif countofposition[1]<0 then l=l+1 endif OrderSize = 1 ELSIF PositionPerf(1)>=0 THEN OrderSize =1 ENDIF ENDIF graph w graph l

  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

  DEFPARAM CumulateOrders=true DEFPARAM PRELOADBARS=2000 REM Achat once ordersize=1 once exitindex=-2   if w<>0 and l<>0 then pondc=round(l/w) ponds=round(w/l) else pondc=0 ponds=0 endif     indicator1 = ExponentialAverage[a](close) indicator2 = Average[b](close) c1 = (indicator1 CROSSES OVER indicator2)   IF c1 THEN BUY OrderSize+pondc contract ROUNDEDUP AT MARKET ENDIF     REM Vente   indicator3 = ExponentialAverage[c](close) indicator4 = Average[d](close) c2 = (indicator3 CROSSES UNDER indicator4)   IF c2 THEN SELL  AT MARKET ExitIndex=BarIndex ENDIF     c3=(indicator1 CROSSES under indicator2)   IF c3 THEN SELLSHORT OrderSize+ponds contract  AT MARKET ENDIF   c4=(indicator1 CROSSES over indicator2)   IF c4 THEN EXITSHORT AT MARKET ExitIndex=BarIndex ENDIF   // positions sizing IF Barindex= ExitIndex+1 THEN ExitIndex =0 IF PositionPerf(1)<0 THEN   if countofposition[1]>0 then w=w+1 elsif countofposition[1]<0 then l=l+1 endif   OrderSize = 1 ELSIF PositionPerf(1)>=0 THEN OrderSize =1 ENDIF ENDIF   graph w graph l

  After small Walk Forward Optimisation Gain +39.08 % VQN 1.02 / Bad VQN but not bad for a “simple” strategy

  This is the previous results I will work on it to optimize and try the other ideas

  Cheers

  Zilliq

   

  https://www.zilliqtradingresearch.fr/

  Coding is not a crime 😉

   

  1 user thanked author for this post.

  Robin

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
  03/06/2017 at 8:46 PM #27482
  zilliq

  Just for fun, if you do a Alembert Martingale (I use the code of Nicolas )

  We obtain +49.42 % VQN 1.26

  Remember it’s just a crossing average and a little bit of Magic

  🙂

  Zilliq

  1 user thanked author for this post.

  CN

  

  zilliq

  Participant

  Topics: 58

  Replies: 683

  Been thanked: 152 times
• Author
  Posts