Linear Regression Universal Strategy, how to improve it?

• 08/25/2022 at 4:36 PM #199634
  JS

  A typical case of selective reading and reverse psychology… 😉

   

  

  JS

  Participant

  Topics: 22

  Replies: 1443

  Been thanked: 566 times
  08/25/2022 at 9:25 PM #199639
  MaoRai54

  @  JS

  what do you mean ?

  

  MaoRai54

  Participant

  Topics: 28

  Replies: 80

  Been thanked: 9 times
  08/25/2022 at 9:37 PM #199640
  JS

  I mean, there are a lot of questions about the system and I had hoped to get an answer to that… 😉

  

  JS

  Participant

  Topics: 22

  Replies: 1443

  Been thanked: 566 times
  08/25/2022 at 9:43 PM #199643
  MaoRai54

  me too …………………   🙂

  in fact, no one is writing any comment about my .itf  I’ve uploaded.

  well, let’s see !!!

  

  MaoRai54

  Participant

  Topics: 28

  Replies: 80

  Been thanked: 9 times
  08/25/2022 at 10:04 PM #199644
  JS

  Sorry MaoRai54, I thought that comment came from davidelaferia, is not about your system, was about the “Linear regression universel strategy” system.

  

  JS

  Participant

  Topics: 22

  Replies: 1443

  Been thanked: 566 times
  08/25/2022 at 10:50 PM #199645
  JS

  Hi @MaoRai54

  I’ve downloaded your itf -file now, thanks for sharing, but not sure which time frame to use?

  You say “5M as basic timeframe” but in the itf -file it says 15 minutes?

  

  JS

  Participant

  Topics: 22

  Replies: 1443

  Been thanked: 566 times
  08/26/2022 at 8:52 AM #199657
  MaoRai54

  @ JS

  you’re right, sorry.

  the TF ‘ve tested better is 15 M not 5M

  1 user thanked author for this post.

  JS

  

  MaoRai54

  Participant

  Topics: 28

  Replies: 80

  Been thanked: 9 times
  08/26/2022 at 9:26 AM #199658
  GraHal

  For now, I have reluctantly given up on David’s original TS as it not work without ’round’ and with ’round’ it gets Rejected for not enough historical data.

  I have a version of MaoRai54 running in Live Demo on 2 mins TF on EURUSD for the last 2 days, but it has yet to take a trade.

  The original MaoRai54 posted version took too few trades for my trading profile. If I recall correctly there were many days / weeks even (?) where no trades were taken … due I think to too many restrictive conditions?  I’d love to be corrected so if anybody wants to post results?

  

  GraHal

  Participant

  Topics: 204

  Replies: 8445

  Been thanked: 2052 times
  08/26/2022 at 9:51 AM #199662
  JS

  The “linear regression universal system” which I started yesterday in demo (time frame 1 day) is still running and has made two transactions…

  Probably something else going on yesterday, that’s why it rained error messages…

  1 user thanked author for this post.

  GraHal

  

  JS

  Participant

  Topics: 22

  Replies: 1443

  Been thanked: 566 times
  08/26/2022 at 10:08 AM #199663
  JS

  Hi @MaoRai54

  I’m going to study your system and put it in demo.

  What strikes me is that there are many optimization parameters, can you tell us more about this?

  How did you envision how the system should work?

  

  JS

  Participant

  Topics: 22

  Replies: 1443

  Been thanked: 566 times
  08/26/2022 at 10:33 AM #199668
  GraHal

  JS wrote:

  I started yesterday in demo (time frame 1 day) is still running

  On the strength of above, I just started my M5 version of David TS.
  First full 5 min bar it got Rejected for
  not enough historical data to calculate at least one indicator etc

  
  Seems Rejections are linked to TF’s at < 1 day fo some weird reason?

  Anybody got any ideas (Roberto?) … here’s the code (optimised for 10K bars)  that gets Rejected at first full 5 min bar,

  //------------------------------------------------------------------------- // Main code : David LinReg DJI M5 10K v1gh //------------------------------------------------------------------------- //https://www.prorealcode.com/topic/linear-regression-universal-strategy-how-to-improve-it/#post-199478//Linear Regression Universal Strategy //DJI // timeframe : M5 // Spread: 5 // created and coded by davidelaferla //————————————————————————- defparam cumulateorders=false DEFPARAM PRELOADBARS = 10000 //*********************************************************************************************************** ONCE A15 = 25.0 ONCE A16 = 10.0 ONCE A17 = 0.8 N = 1 //------------------ VARIABILI DEL SISTEMA--------------------------------------- //CNHJPY Values: -------------------------------------------- Ottimization info Volatilityperiod=A15 //15// different fixed value for each currency pair: range=3-20, with step=1 Regression1period=A16 //18// Linear Regression Imput Signal: range=6-100 with step=1 Regression2period=A17 //1.3//Linear Regression Output Signal: range=1-2, with step=0.01 //*********************************************************************************************** //------------------ INDICATOR --------------------------------------- hv=HistoricVolatility[VolatilityPeriod](close) hv2=HistoricVolatility[VolatilityPeriod](typicalprice) st=linearregression[ROUND(Regression1Period*hv)](close) lr=linearregression[ROUND(Regression1Period*Regression2Period*hv2)](typicalprice) // Condizioni per uscire da posizioni short e entrare su posizioni long IF st<lr THEN BUY n contract AT lr stop ENDIF // Condizioni per uscire da posizioni long e entrare su posizioni short IF st>lr THEN SELLshort n contract At st stop 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

  //------------------------------------------------------------------------- // Main code : David LinReg DJI M5 10K v1gh //------------------------------------------------------------------------- //https://www.prorealcode.com/topic/linear-regression-universal-strategy-how-to-improve-it/#post-199478//Linear Regression Universal Strategy //DJI // timeframe : M5 // Spread: 5 // created and coded by davidelaferla //————————————————————————- defparam cumulateorders=false DEFPARAM PRELOADBARS = 10000 //*********************************************************************************************************** ONCE A15 = 25.0 ONCE A16 = 10.0 ONCE A17 = 0.8 N = 1 //------------------ VARIABILI DEL SISTEMA--------------------------------------- //CNHJPY Values:      -------------------------------------------- Ottimization info Volatilityperiod=A15 //15// different fixed value for each currency pair: range=3-20, with step=1 Regression1period=A16 //18// Linear Regression Imput Signal:              range=6-100 with step=1 Regression2period=A17 //1.3//Linear Regression Output Signal:             range=1-2, with step=0.01 //*********************************************************************************************** //------------------ INDICATOR --------------------------------------- hv=HistoricVolatility[VolatilityPeriod](close) hv2=HistoricVolatility[VolatilityPeriod](typicalprice) st=linearregression[ROUND(Regression1Period*hv)](close) lr=linearregression[ROUND(Regression1Period*Regression2Period*hv2)](typicalprice) // Condizioni per uscire da posizioni short e entrare su posizioni long IF st<lr THEN BUY n contract AT lr stop ENDIF // Condizioni per uscire da posizioni long e entrare su posizioni short IF st>lr THEN SELLshort n contract At st stop ENDIF

   

  

  GraHal

  Participant

  Topics: 204

  Replies: 8445

  Been thanked: 2052 times
  08/26/2022 at 11:10 AM #199670
  MaoRai54

  @ JS

  @GraHal

  Comments on my itf :

  • I started from the version of Davide then I applied different TF to try to get more triggers
  • I added the long part at the beginning to have increase/decrease of number of contracts based on profits. that’s why you’ll see many variables
  • in addition H1 and H2 are the best hour to operate
  • furthermore, to reduce the amount of losses based on contract q.ty I divide the SL by the number of contracts. So in case of multiple q.ty the maximum SL amount remains the one of 1 contract
  • finally, I’m running it with basic TF  15M and, at the moment, only for Eur/$. I’ll try other later.

  The result of backtest are quite poor but constant:

  • Eur/$    period  July 2018 – Aug 2022
  • winning rate  71%
  • 1897 trades
  • +0,90 $ average per trade

   

   

  

  MaoRai54

  Participant

  Topics: 28

  Replies: 80

  Been thanked: 9 times
  08/26/2022 at 12:22 PM #199672
  JC_Bywan

  Hi GraHal, hi all,

  maybe worth trying these 2 modifications together for lines 29-30 of your #199668?

  (a) change the period to max(1, ROUND(Regression1Period*Regression2Period*hv2)) in order to avoid a zero period in case of low hv or hv2, and still keep the “round” spirit (read: not use ceil when not necessary above 1)

  (b) add an “if” statement on barindex to make sure hv and hv2 actually exist when calculting these periods for lr and st: if barindex>=VolatilityPeriod then…

  All together:

  if barindex>=VolatilityPeriod then lr=linearregression[max(1,ROUND(Regression1Period*Regression2Period*hv2))](typicalprice) st=linearregression[max(1,ROUND(Regression1Period*hv))](close) endif

  1 2 3 4

  if barindex>=VolatilityPeriod then lr=linearregression[max(1,ROUND(Regression1Period*Regression2Period*hv2))](typicalprice) st=linearregression[max(1,ROUND(Regression1Period*hv))](close) endif

  If it doesn’t work I can try to have another look, but it’s all I can think of for now…

   

   

  1 user thanked author for this post.

  GraHal

  

  JC_Bywan

  Moderator

  Topics: 20

  Replies: 2563

  Been thanked: 845 times
  08/26/2022 at 12:33 PM #199673
  GraHal

  Thanks so much JC, unfortunately the TS – including your suggestions above – got Rejected (not enough historic data to calc an indicator) at the first full 5 minute bar … weird eh?

   

  

  GraHal

  Participant

  Topics: 204

  Replies: 8445

  Been thanked: 2052 times
  08/26/2022 at 1:16 PM #199675
  JC_Bywan

  ok, as it solved display issues with returned lr and st in probuilder, I was hoping to avoid the pain of adding another system in proorder… but I tried and got it rejected too.

  With it working in probuilder and not proorder it made me look at orders code lines, two thoughts:  (1) I never tried sending pending order « from the wrong side », and position of price wasn’t checked to decide if pending order at stop or at limit is more appropriate, so I modified code as follows in case wrong side pending order was messing things up :

  IF st<lr THEN
  if close<lr then
  BUY n contract AT lr stop
  elsif close>lr then
  BUY n contract AT lr limit
  ENDIF
  endif
  //
  IF st>lr THEN
  if close>st then
  SELLshort n contract At st stop
  elsif close<st then
  SELLshort n contract At st limit
  endif
  ENDIF

  and (2) st and lr are highly likely to be within minimum stop distance and without stop readjustment box checked system is likely to get stopped fast, so I tried again checking the box…

  But even with those further 2 changes, got it rejected again… sorry…

  1 user thanked author for this post.

  GraHal

  

  JC_Bywan

  Moderator

  Topics: 20

  Replies: 2563

  Been thanked: 845 times
• Author
  Posts