Bollinger/Average/Momentum GBP/USD 3min short

• 09/07/2016 at 2:29 PM #12809
  Mat_CH

  Hi all,

  i have start  3 month ago to make strategy with PRT and this my first strategy, only short positions for GBP/USD 3min. I have backtested and the result is good.

  Maybe someone can test for more log time? or tell me if there are some improvement to do?

  Thanks

  MAT

  

  Mat_CH

  Participant

  Topics: 1

  Replies: 17

  Been thanked: 11 times
  09/07/2016 at 2:32 PM #12814
  Nicolas

  Hi Mat_CH,

  Your equity curve looks great, I have read your code and I think you might have optimized a bit the periods of your indicators? and maybe other things? am I right?

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4741 times
  09/08/2016 at 10:11 AM #12855
  Mat_CH

  Hi Nicolas,

  yes i have started with bolliger band and after i have tried with some indicators for have less drawdown. After i have optimized all period.

  In reality i wanted to publish strategy and not open a forum topic.

  How i can do this?

  Thanks

  Mat

  

  Mat_CH

  Participant

  Topics: 1

  Replies: 17

  Been thanked: 11 times
  09/08/2016 at 10:31 AM #12858
  Mat_CH

  Hi Nicolas, I have found where share tha strategy.

  Thanks

  Mat

  

  Mat_CH

  Participant

  Topics: 1

  Replies: 17

  Been thanked: 11 times
  09/08/2016 at 11:19 AM #12860
  Nicolas

  Your strategy were in the Library shared content, but I moved it here to discuss about it before make it a part of it. I believe your strategy is curvefitted on past datas. Maybe you could optimize it with at least IN SAMPLE / OUT OF SAMPLE periods like I describe in this blog post: http://www.prorealcode.com/blog/avoid-equity-curve-fitting-with-probacktest-trading-strategy-optimisation/

  Also, do you have any trade that last only one bar? Entry and exit on the same candlestick?

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4741 times
  09/09/2016 at 9:14 AM #12879
  Mat_CH

  Thanks for the tips. I think i have work i this sense, because in the i beginning i had very good performance from november 2015 to january 2016 but on the rest of thi time not so good. For this i have introduced average with period of 3000 and the momentum. And also adjusted the paramiters. The result is that i gain less the first period but i have a better curve with less drawdown.

  With my PR in can test only this period 19.11.2016-09.09.2016 you can test more time? if yes how is the result?

  Thanks

  Mat

  

  Mat_CH

  Participant

  Topics: 1

  Replies: 17

  Been thanked: 11 times
  09/09/2016 at 11:47 AM #12883
  Nicolas

  I have not tested before this period, but optimizing indicators periods is not a good way to make a strategy a better one. Why? because most of the traders and algorithm use the same default periods, such as 20 periods with 2 standard deviation for a Bollinger Bands indicator for example. There are already a lot of discussions about equity overfitting here in the forums. Also, you add a bias into your strategy because you only sell the pounds while it is obviously in a down trend phase, that’s why you make great profit at the beginning of your test, you were selling it because you already know the past but do you know also the future?

  Sorry to be such direct, but this is the hard reality of automated trading, while we create algorithm on the past data, we also need to be almost certain that it would make an edge for the future one.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4741 times
  09/09/2016 at 12:37 PM #12884
  Mat_CH

  Ok i trash my strategy.

  If i have understand, the better way to do an automated strategy is to reinvent new metod without using standard indicators but better our own indicators?

  Mat

  

  Mat_CH

  Participant

  Topics: 1

  Replies: 17

  Been thanked: 11 times
  09/09/2016 at 1:31 PM #12889
  Nicolas

  You can of course re-invent new ones, but optimizing periods of largely used technical indicators such as RSI for example, in order to get a better equity curve is almost useless, since you are not following the herd of traders who use the same empirical period of computation. Also, please keep in mind that there is almost no easy way or magical science to make a good trading strategy.

  I don’t remember who, but someone post this file attached recently, it is an “how-to” start creating TS.

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4741 times
• Author
  Posts