Mean Reverting Strategy on AUDCAD on h1 timeframe

• 03/05/2018 at 11:27 AM #64053
  Francesco78

  This is a mean reverting strategy that I currently run live on AUDCAD.

  Enjoy, comment and improve 🙂

  //------------------------------------------------------------------------- // Main code : audcad_meanrevert_v2 //------------------------------------------------------------------------- defparam cumulateorders = false n = 1 m = 3 l = 90 cs= summation[n](close>open) = n cs = cs and close>bollingerup[20](close) and close>dlow(2) cl= summation[m](close<open) = m cl = cl and close<bollingerdown[20](close) and close<Dhigh(2) size = 2 //entry conditions if cs then sellshort size contract at market endif if cl then buy size contract at market endif //exit conditions if shortonmarket and close<average[l](close) then exitshort at market endif if longonmarket and close>average[l](close) then sell at market 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

  //------------------------------------------------------------------------- // Main code : audcad_meanrevert_v2 //------------------------------------------------------------------------- defparam cumulateorders = false   n = 1 m = 3 l = 90     cs= summation[n](close>open) = n cs = cs  and close>bollingerup[20](close)  and close>dlow(2) cl= summation[m](close<open) = m cl = cl   and close<bollingerdown[20](close) and close<Dhigh(2)     size = 2 //entry conditions if cs   then sellshort size contract at market endif   if cl  then buy size contract at market endif   //exit conditions if shortonmarket and close<average[l](close) then exitshort at market endif     if longonmarket and close>average[l](close) then sell at market endif

   

  

  Francesco78

  Participant

  Topics: 43

  Replies: 242

  Been thanked: 77 times
  03/05/2018 at 11:37 AM #64416
  Nicolas

  Thank you Francesco, going back to mid 2010, the strategy is still quite good in tick by tick backtesting. Did you try some WalkForward analysis since you seem to have optimized the variables? I know there are only 3 of them, but should be interesting to know how it performs with WF. I launched one just now, I’ll get back with results.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  03/05/2018 at 12:10 PM #64424
  juanj

  I threw my Heuristics Algorithm in there to optimize the average[l] variable;

  Now results are looking good on other AUD pairs like AUDUSD as well.

  //------------------------------------------------------------------------- //------------------------------------------------------------------------- // Main code : audcad_meanrevert_v2 //------------------------------------------------------------------------- defparam cumulateorders = false // Heuristics Algorithm Start once OPT = 1 If OPT < 2 and onmarket then OPT = 0 ElsIf countofposition = 0 Then OPT = 1 EndIF If OPT = 0 Then optimize = optimize + 1 OPT = 2 EndIf //////////////////////////////////////// once periods = 100 Increment = 14 MaxIncrement = 4 Reps = 20 //Number of bars to use for analysis once PIncPos = 1 //Positive Increment Position once NIncPos = 1 //Neative Increment Position once Optimize = 0 ////Initialize Heuristicks Engine Counter (Must be Incremented at Position Start or Exit) once Mode = 1 //Switches between negative and positive increments once WinCountB = 0 //Initialize Best Win Count once StratAvgB = 0 //Initialize Best Avg Strategy Profit If Optimize = Reps Then WinCountA = 0 //Initialize current Win Count StratAvgA = 0 //Initialize current Avg Strategy Profit For i = 1 to Reps Do If positionperf(i) > 0 Then WinCountA = WinCountA + 1 //Increment Current WinCount EndIf StratAvgA = StratAvgA + positionperf(i) Next StratAvgA = StratAvgA/Reps //Calculate Current Avg Strategy Profit If StratAvgA >= StratAvgB Then StratAvgB = StratAvgA //Update Best Strategy Profit BestA = Periods EndIf If WinCountA >= WinCountB Then WinCountB = WinCountA //Update Best Win Count BestB = Periods EndIf If WinCountA > WinCountB and StratAvgA > StratAvgB Then Mode = 0 ElsIf WinCountA < WinCountB and StratAvgA < StratAvgB and Mode = 1 Then Periods = Periods - (Increment*NIncPos) NIncPos = NIncPos + 1 Mode = 2 ElsIf WinCountA >= WinCountB or StratAvgA >= StratAvgB and Mode = 1 Then Periods = Periods + (Increment*PIncPos) PIncPos = PIncPos + 1 Mode = 1 ElsIf WinCountA < WinCountB and StratAvgA < StratAvgB and Mode = 2 Then Periods = Periods + (Increment*PIncPos) PIncPos = PIncPos + 1 Mode = 1 ElsIf WinCountA > WinCountB or StratAvgA > StratAvgB and Mode = 2 Then Periods = Periods - (Increment*NIncPos) NIncPos = NIncPos + 1 Mode = 2 EndIf If NIncPos > MaxIncrement or PIncPos > MaxIncrement Then If BestA = BestB Then Periods = BestA Else Periods = (BestA+BestB)/2 EndIf NIncPos = 1 PIncPos = 1 EndIF Optimize = 0 EndIf GRAPH Periods AS "Periods" // Heuristics Algorithm End n = 1 m = 3 l = periods//90 cs= summation[n](close>open) = n cs = cs and close>bollingerup[20](close) and close>dlow(2) cl= summation[m](close<open) = m cl = cl and close<bollingerdown[20](close) and close<Dhigh(2) size = 2 //entry conditions if cs then sellshort size contract at market endif if cl then buy size contract at market endif //exit conditions if shortonmarket and close<average[l](close) then exitshort at market endif if longonmarket and close>average[l](close) then sell at market 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121

  //------------------------------------------------------------------------- //------------------------------------------------------------------------- // Main code : audcad_meanrevert_v2 //------------------------------------------------------------------------- defparam cumulateorders = false   // Heuristics Algorithm Start   once OPT = 1   If OPT < 2 and onmarket then OPT = 0 ElsIf countofposition = 0 Then OPT = 1 EndIF   If OPT = 0 Then optimize = optimize + 1 OPT = 2 EndIf   ////////////////////////////////////////   once periods = 100 Increment = 14 MaxIncrement = 4 Reps = 20 //Number of bars to use for analysis once PIncPos = 1 //Positive Increment Position once NIncPos = 1 //Neative Increment Position once Optimize = 0 ////Initialize Heuristicks Engine Counter (Must be Incremented at Position Start or Exit) once Mode = 1 //Switches between negative and positive increments once WinCountB = 0 //Initialize Best Win Count once StratAvgB = 0 //Initialize Best Avg Strategy Profit   If Optimize = Reps Then WinCountA = 0 //Initialize current Win Count StratAvgA = 0 //Initialize current Avg Strategy Profit   For i = 1 to Reps Do If positionperf(i) > 0 Then WinCountA = WinCountA + 1 //Increment Current WinCount EndIf StratAvgA = StratAvgA + positionperf(i) Next StratAvgA = StratAvgA/Reps //Calculate Current Avg Strategy Profit   If StratAvgA >= StratAvgB Then StratAvgB = StratAvgA //Update Best Strategy Profit BestA = Periods EndIf   If WinCountA >= WinCountB Then WinCountB = WinCountA  //Update Best Win Count BestB = Periods EndIf   If WinCountA > WinCountB and StratAvgA > StratAvgB Then Mode = 0 ElsIf WinCountA < WinCountB and StratAvgA < StratAvgB and Mode = 1 Then Periods = Periods - (Increment*NIncPos) NIncPos = NIncPos + 1 Mode = 2 ElsIf WinCountA >= WinCountB or StratAvgA >= StratAvgB and Mode = 1 Then Periods = Periods + (Increment*PIncPos) PIncPos = PIncPos + 1 Mode = 1 ElsIf WinCountA < WinCountB and StratAvgA < StratAvgB and Mode = 2 Then Periods = Periods + (Increment*PIncPos) PIncPos = PIncPos + 1 Mode = 1 ElsIf WinCountA > WinCountB or StratAvgA > StratAvgB and Mode = 2 Then Periods = Periods - (Increment*NIncPos) NIncPos = NIncPos + 1 Mode = 2 EndIf   If NIncPos > MaxIncrement or PIncPos > MaxIncrement Then If BestA = BestB Then Periods = BestA Else Periods = (BestA+BestB)/2 EndIf NIncPos = 1 PIncPos = 1 EndIF   Optimize = 0 EndIf   GRAPH Periods AS "Periods"   // Heuristics Algorithm End   n = 1 m = 3 l = periods//90 cs= summation[n](close>open) = n cs = cs  and close>bollingerup[20](close)  and close>dlow(2) cl= summation[m](close<open) = m cl = cl   and close<bollingerdown[20](close) and close<Dhigh(2) size = 2 //entry conditions if cs   then sellshort size contract at market endif if cl  then buy size contract at market endif //exit conditions if shortonmarket and close<average[l](close) then exitshort at market endif   if longonmarket and close>average[l](close) then sell at market endif

   

  4 users thanked author for this post.

  GraHal, manel, Nicolas, Francesco78

  

  juanj

  Participant

  Topics: 51

  Replies: 505

  Been thanked: 170 times
  03/05/2018 at 12:11 PM #64425
  Francesco78

  Hi Nicolas, yes I did WF test of course.

  Ill be curious to see the results since 2010. Many thanks

  Francesco

  

  Francesco78

  Participant

  Topics: 43

  Replies: 242

  Been thanked: 77 times
  03/05/2018 at 1:50 PM #64435
  Nicolas

  Sorry, I forgot to attach the result in my first post.

  You’ll find attached 3 different pictures, first with WF non anchored, second one with WF anchored and the last one with the variables you are using but with 200k bars backtest.

  As you can see all the OOS periods make huge gain while the IS period is not performing so well, it is a sign of overfit. It doesn’t mean that the strategy is bad, it means that it should need refining with some extra filters or more OOS periods to optimize more often.

  
  @juanj Thank you, will study this version.

   

  1 user thanked author for this post.

  Francesco78

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  03/05/2018 at 9:31 PM #64503
  Francesco78

  Thank you Nicolas, can you claryfy for my own benefit why you think that adding filter in this case could be effecting on reducing overfitting?

  Thanks!

  

  Francesco78

  Participant

  Topics: 43

  Replies: 242

  Been thanked: 77 times
  03/06/2018 at 10:45 AM #64559
  Nicolas

  The question is, why does the strategy is not performing as well in first IS as in OOS? 2012 crisis? I meant to study entries depending of the market behavior in this case, of course, more variables mean more overfit… But as I said, maybe it could be solved by adding more IS+OOS repetitions.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  03/15/2018 at 8:45 AM #65330
  Nicolas

  Well, I think it could be now added to the library, with the WFA attached, what do you think @Francesco78 ?

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  03/15/2018 at 11:13 AM #65358
  Francesco78

  @Nicolas

  Thank you, I am fine with it, let me know if you want me to make some amendments on my original post.

  

  Francesco78

  Participant

  Topics: 43

  Replies: 242

  Been thanked: 77 times
  03/17/2018 at 12:35 PM #65490
  Francesco78

  @Nicolas

  What shoud I do to put it on the library?

  Thank you

  

  Francesco78

  Participant

  Topics: 43

  Replies: 242

  Been thanked: 77 times
  03/17/2018 at 6:03 PM #65528
  Nicolas

  Hey Francesco, I’ll do it for you on Monday 😉

  1 user thanked author for this post.

  Francesco78

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20863

  Been thanked: 4774 times
  04/23/2018 at 8:54 PM #69022
  Dymjohn

  For what it is worth I tried tackling the large value of drawdowns compared to winning trades. I noticed the losing trades tend to have a larger number of bars so I optimized bars to < 80 bars which on 100000 bars reduced profit slightly 11384.4 down to 11015.20.  It did reduce the maximum loss to 415.2o from 761.00.

  The code for this change is attached

  defparam cumulateorders = false n = 1 m = 3 l = 90 once Trade = 1 cs= summation[n](close>open) = n cs = cs and close>bollingerup[20](close) and close>dlow(2) cl= summation[m](close<open) = m cl = cl and close<bollingerdown[20](close) and close<Dhigh(2) size = 2 //entry conditions if cs then sellshort size contract at market endif if cl then buy size contract at market endif if Barindex - Tradeindex < 80 then Trade =1 else Trade = 0 endif //exit conditions if shortonmarket and close<average[l](close) or Trade = 0 then exitshort at market endif if longonmarket and close>average[l](close)or Trade = 0 then sell at market 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

  defparam cumulateorders = false   n = 1 m = 3 l = 90 once Trade = 1   cs= summation[n](close>open) = n cs = cs  and close>bollingerup[20](close)  and close>dlow(2) cl= summation[m](close<open) = m cl = cl   and close<bollingerdown[20](close) and close<Dhigh(2)     size = 2 //entry conditions if cs   then sellshort size contract at market endif   if cl  then buy size contract at market endif if Barindex - Tradeindex < 80 then Trade =1 else Trade = 0 endif   //exit conditions if shortonmarket and close<average[l](close) or Trade = 0 then exitshort at market endif     if longonmarket and close>average[l](close)or Trade = 0 then sell at market endif

   

  1 user thanked author for this post.

  Francesco78

  

  Dymjohn

  Participant

  Topics: 16

  Replies: 43

  Been thanked: 4 times
  05/04/2018 at 8:19 PM #69796
  Jan

  Thank you very much,  Francesco, for the strategy, which seems solid and robust !

  I tried this on AUD CAD 30 minutes and it works also quite well.

  Extended variables :   S for short average close

  With 15.000 bars from 7 February till 4 May 2018 it gives a robust profit of 2.945 CAD in 3 months (I did no IS/OOS testing) with M=3, N=2, L = 50 and S = 40, in 398 trades, spread 1,7 pips calculated

  I also tried a maximum of one trade per day, with adding the condition ” OneTradePerDay = IntradayBarIndex < (Barindex – TradeIndex(1)) ”

  This reduces the amount of trades with more then half, 180 trades, giving a robust profit of 2.354 CAD, with  M=3, N=5, L = 50 and S = 90  (also I did no IS/OOS testing so far)

  ========================================================

  For other currencies and indices I am still testing

  So far for the FTSE with 2 contracts, exposure more than 30.000 GBP (!), it gives interesting profit of 3.900 GBP on the 15 minutes bar chart with 15.000 bars, from an a maximum of one trade per day, from 6 september 2017 till 4 May 2018,  with  M=1, N=3, L = 50 and S = 40  (also I did no IS/OOS testing so far !!)

  The risk of overfitting for this kind of short periods is very big !! When testing for 100.000 bars the profit especially in the beginning, completely deminish !!!!!!!, but the endresult is quiet the same !

  When testing the FTSE for 100.000 15 minutes-bars with not restricted trades per day, the NON- robust profit is  a bit lower as when restricted to one trade per day, but not robust and a bit lower.

  Kind regards,

  Jan

   

  1 user thanked author for this post.

  Francesco78

  

  Jan

  Participant

  Topics: 10

  Replies: 90

  Been thanked: 29 times
  05/04/2018 at 8:22 PM #69798
  Francesco78

  Thank you Jan!

  

  Francesco78

  Participant

  Topics: 43

  Replies: 242

  Been thanked: 77 times
  05/12/2018 at 9:00 AM #70240
  M. Pettersson

  Really nice strategy !

  I’m running it in my demo account to try it out.

  Thanks Francesco.

  1 user thanked author for this post.

  Francesco78

  

  M. Pettersson

  Participant

  Topics: 6

  Replies: 13

  Been thanked: 2 times
• Author
  Posts