Backtest results changes with same values

• 11/25/2020 at 10:59 PM #151667
  Vonasi

  The best solution is to provide the code so that others can test it and see if they get the same results or can spot something in the code that could make the difference.

  

  Vonasi

  Moderator

  Topics: 192

  Replies: 5088

  Been thanked: 1810 times
  11/25/2020 at 11:18 PM #151668
  OboeOpt

  GraHal wrote:

  Have you tried to isolate if just shorts or just longs make up the difference or both shorts and longs?

  In the bad result version the trades have increased with 6 long positions and 26 short.

  

  OboeOpt

  Participant

  Topics: 11

  Replies: 101

  Been thanked: 7 times
  11/26/2020 at 9:01 AM #151687
  Nicolas

  Specific to this code or with other ones? still in v10.3?

  

  Nicolas

  Keymaster

  Topics: 81

  Replies: 20828

  Been thanked: 4739 times
  11/26/2020 at 12:08 PM #151734
  OboeOpt

  Nicolas wrote:

  Specific to this code or with other ones? still in v10.3?

  When this happened on Another instrument it was with a similar code, but there is nothing special in the code that isn´t frequently used.
  Yes, v10.3.

  

  OboeOpt

  Participant

  Topics: 11

  Replies: 101

  Been thanked: 7 times
  11/26/2020 at 7:16 PM #151806
  OboeOpt

  I Think i´ve accidently named a variable (X7) for optimization the same name as a function (X7=)  in the code. So when I replaced the variable with a fixed optimized value I actually excluded the function instead… Will try to replicate the “good” result by inserting the function and do some more tests.

  Are there some rules on how to name variables in the code when optimizing?

  

  OboeOpt

  Participant

  Topics: 11

  Replies: 101

  Been thanked: 7 times
  11/26/2020 at 7:55 PM #151810
  GraHal

  OboeOpt wrote:

  Are there some rules on how to name variables in the code when optimizing?

  No rules, but I make it easy for myself and name them A Line number so if the variable is on Line 20 it would A20.  This makes it easy to find where the variable is in the code.

  I did think that what you have found might be the case as it has happened to me 2 or 3 times when using other folks code with many hundreds of lines of code!

  1 user thanked author for this post.

  OboeOpt

  

  GraHal

  Participant

  Topics: 204

  Replies: 8397

  Been thanked: 2039 times
  01/14/2021 at 9:23 AM #157755
  PeterSt

  GraHal wrote:

  Surely the difference is directly related to the difference in number of trades?

  Hi GraHal,

  I know this is an old thread, but I don’t see what the problem is with a slightest change in e.g. an exit (like 2 seconds earlier) that changing the WHOLE situation.
  This could be too difficult to explain, but especially with smaller bars this changes things wildly.

  Of course this is not an explanation of what Oboe experiences, but it is fully legit to happen. I mean, it is not the problem in itself; I will try to duplicate a very similar issue I had myself yesterday, so you can see what happened (by unexplainable matter), that resulting in a very long running draw-down, that prohibiting a few dozen other trades to happen. … If you don’t hear me back on this, I couldn’t duplicate it any more.

  Peter

  

  PeterSt

  Participant

  Topics: 70

  Replies: 2285

  Been thanked: 693 times
  01/14/2021 at 10:28 AM #157766
  GraHal

  You need to start a new Topic as the Issue you mention is Off Topic for this Topic? 

  PeterSt wrote:

  that prohibiting a few dozen other trades to happen

  Above is the basis of VRT /  Vonasi Robustness Tester.

  1 user thanked author for this post.

  PeterSt

  

  GraHal

  Participant

  Topics: 204

  Replies: 8397

  Been thanked: 2039 times
• Author
  Posts