Optimisation recommendations

• 05/05/2019 at 7:11 PM #97741
  Alun

  Are there code optimisation recommendations so that variables are optimised in a way (or a sequence) that gives profit results with better chance of good walk forward results?

  

  Alun

  Participant

  Topics: 10

  Replies: 20

  Been thanked: 1 time
  05/15/2019 at 3:03 PM #98558
  Alun

  Don’t think I asked my previous question very clearly.

  When using the variable optimisation tool when developing a strategy, is there a recommended order or method – for example, long variables before short, entry before exit variables, target profit before or after trailing stop, etc, etc

  Would anyone know if there is any best practice advice documented on this sort of variable optimisation detail?

   

   

  

  Alun

  Participant

  Topics: 10

  Replies: 20

  Been thanked: 1 time
  05/16/2019 at 9:15 AM #98588
  jebus89

  Ive read a couple of books and heard a few podcasts that have discussed this topic, and the short answer is: No. There is no secret method to this.

  Some swear to optimizing as much as possible, some say to optimize more like “Making ur hand fit in a mitten vs fitting it in a glove”. meaning instead of maybe looking for intervals of 1 when trying to find “Set stop ploss N1”, rather use an interval of 5, or 10. Or maybe use % instead of pips.

  There is no “correct” answer in my eyes, everything and nothing will and can and wont work 🙂

  My beliefs are that less degrees of freedom = better. meaning the less variable u can optimize, the more “mitten” vs “glove” you have.

  The mitten vs glove is probably my favorite analogy. The markets will never stay exactly the same, and losses are inevitable. We have just had 10 years of low volatile upwards markets, now we’re stuck in a voaltile hell-hole where stock indicies are jumping +/-5% daily it feels like. Even up and down multiple % on the same days. its pretty crazy. What worked good in low-volatile markets might be taking huge hits now that the swings are back. personally i try to only make algos that do OK in both types of markets. Backtest over 10 years including the 2008 crisis will help u show if ur algo will make it or break it.

  1 user thanked author for this post.

  GraHal

  

  jebus89

  Participant

  Topics: 118

  Replies: 586

  Been thanked: 238 times
• Author
  Posts