Reversion Bars v2.0 GBPUSD

• 07/19/2017 at 12:59 PM #41031
  David Somogyi

  Hello all, this is the latest version of my reversion bars strategy on GBPUSD, it has been optimized and its running better, also i have backtested on 100000. If anyone want to kindly backtest this strategy on 200000 bars, be my guest and please share the results =D =D Thanks and good trading

  defparam cumulateorders = false MyEMA = ExponentialAverage[36](close) nbar = 4 //Volatility Filter:// //FDI (fractal dimension index) N = 33 once fdi=undefined if barindex >= n-1 then diff=0 length = 0 pdiff = 0 hh=0 ll=0 FDI=0 HH = highest[N](close) LL = lowest[N](close) for Period = 1 to N-1 do if (HH - LL) > 0 then diff = (customclose[Period] - LL) / (HH - LL) if Period > 1 then length = length + SQRT(SQUARE(diff - pdiff) + (1 / SQUARE(N))) endif pdiff = diff endif next if length > 0 then FDI = 1 + (LOG(length) + LOG(2)) / LOG(2 * (N)) else FDI = 0 endif endif positionsize = 1 C4 = MyEMA CROSSES OVER CLOSE C3 = BARINDEX - TRADEINDEX = NBAR C5 = close CROSSES UNDER MyEMA //Enter long positions IF not longonmarket and open > close[1] and C4 and FDI<1.5 THEN BUY positionsize LOT AT MARKET nextbaropen ENDIF //Exit long positions If LongOnMarket AND C3 THEN SELL AT MARKET endif //Enter short positions IF NOT ShortOnMarket AND open < close[1] and C5 and FDI<1.5 THEN SELLSHORT positionsize LOT AT MARKET nextbaropen ENDIF // Exit short positions IF ShortOnMarket AND C3 THEN EXITSHORT AT MARKET endif //Stops //set stop $loss 60

  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

  defparam cumulateorders = false   MyEMA = ExponentialAverage[36](close) nbar = 4   //Volatility Filter://   //FDI (fractal dimension index)   N = 33   once fdi=undefined   if barindex >= n-1 then diff=0 length = 0 pdiff = 0 hh=0 ll=0 FDI=0 HH = highest[N](close) LL = lowest[N](close)   for Period = 1 to N-1 do if (HH - LL) > 0 then diff = (customclose[Period] - LL) / (HH - LL) if Period > 1 then length = length + SQRT(SQUARE(diff - pdiff) + (1 / SQUARE(N))) endif pdiff = diff endif next   if length > 0 then FDI = 1 + (LOG(length) + LOG(2)) / LOG(2 * (N)) else FDI = 0 endif endif   positionsize = 1   C4 = MyEMA CROSSES OVER CLOSE C3 = BARINDEX - TRADEINDEX = NBAR C5 = close CROSSES UNDER MyEMA   //Enter long positions   IF not longonmarket and open > close[1] and C4 and FDI<1.5 THEN BUY positionsize LOT AT MARKET nextbaropen ENDIF   //Exit long positions   If LongOnMarket AND C3 THEN SELL AT MARKET endif   //Enter short positions   IF NOT ShortOnMarket AND open < close[1] and C5 and FDI<1.5 THEN SELLSHORT positionsize LOT AT MARKET nextbaropen ENDIF   // Exit short positions   IF ShortOnMarket AND C3 THEN EXITSHORT AT MARKET endif   //Stops //set stop $loss 60

   

   

   

  

  David Somogyi

  Participant

  Topics: 6

  Replies: 12

  Been thanked: 0 times
  07/19/2017 at 1:05 PM #41072
  Nicolas

  Hi David, thank you for your post into the library. I moved it here because I think you should first make a robustness test with the new WF tool.

  As you can see in my attached pictures, I backtested the strategy with and without the tick mode. It is obviously overfitted and that’s why I think you should use the Walk Foward tool to see how and when the strategy should be re-optimized in order to be more confident about the way the strategy could perform in the future. You can read the Walk Forward FAQ in this blog article.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20866

  Been thanked: 4781 times
  07/19/2017 at 3:23 PM #41087
  David Somogyi

  Thanks Nicolas,

  I’ll put some more work on it and perform a non-anchored WF to the system. However, WF results not always have the last word when it comes to real performance.

  Anyway, many thanks for your  advice and i’ll post it again with better and reliable results.

   

  Best Regards,

  

  David Somogyi

  Participant

  Topics: 6

  Replies: 12

  Been thanked: 0 times
  07/19/2017 at 4:53 PM #41091
  Nicolas

  The WF analysis will tell you if the strategy is robust: it will use optimized variables values into ‘real’ trading periods (Out Of Sample periods) in the past to see if the strategy performed as good as in the optimized period (the In-Sample one).

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20866

  Been thanked: 4781 times
  07/19/2017 at 5:51 PM #41097
  David Somogyi

  Nicolas,

  Yes, i know that is a superb tool, i have been using it since last year and i like it very much, i did not use it in this particular strategy.

  One question though, from which % in an X period do you consider “suspicious” or “out of range”, for example: >90 % or >120% ?

   

  Thanks

   

  

  David Somogyi

  Participant

  Topics: 6

  Replies: 12

  Been thanked: 0 times
  07/19/2017 at 8:25 PM #41113
  Nicolas

  This is a very good question! and there is not an easy answer. I think it depends of the market behavior, a big WFE could be due to a gap or a sudden crash of the market in the good direction, so in this case this is not so very important as long as one WFE isn’t larger than 50% of the overall profit of the backtest. Of course, we also don’t want a lot of high WFE (+150, +200%+++), because the strategy we have developed was not designed to achieve these kind of huge results!

  Because the WF tool only give result of the most profitable strategy, it introduces a bias in the test and that’s why we must make a deep analysis of the WFE and why they are good or not..

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20866

  Been thanked: 4781 times
• Author
  Posts