Regarding ProOrder good practices

• 05/08/2023 at 6:21 PM #214315
  aldtrading

  Hello, lately I’ve been spending all my free time tinkering with ProBackTest/ProOrder systems and so all kind of questions are emerging, tonight I’m wondering :

  For those who succeeded in creating profitable systems, did you wait before the system was profitable no matter how large the sample size you tried it on was (for example starting with 10k bars, all the way to 200k bars) with profits being pretty much always increasing with the time period and the percentages staying pretty stable?

  So getting a standalone autonomous system which would be profitable almost all the time no matter the market state, or maybe did you try to get a robust system which could be flexible (that could work no matter the market state provided some variables were adapted to this state) but for which you would need to reoptomize the variables once in a while when the system seem to outperform or to make it fit some market state changes? Is it possible to approach things in this way (not having a bulletproof system and still being consistently profitable by reoptimizing it)?

  🙂

  

  aldtrading

  Participant

  Topics: 13

  Replies: 43

  Been thanked: 4 times
  05/08/2023 at 6:40 PM #214317
  GraHal

  aldtrading wrote:

  Is it possible to approach things in this way

  Yes, I use that approach and so I rarely optimise over > 10K bars (unless a seconds Timeframe) and I am happy with the results under Live running / Forward Testing.

   

  

  GraHal

  Participant

  Topics: 211

  Replies: 8764

  Been thanked: 2179 times
  05/08/2023 at 7:12 PM #214319
  aldtrading

  Good to know thanks, all this time I’ve been trying to find a “perfect” standalone system, and I still see this as a goal, but maybe this isn’t a requirement to have a profitable system, even in the long run, after all.. though it would be reassuring

  

  aldtrading

  Participant

  Topics: 13

  Replies: 43

  Been thanked: 4 times
  05/09/2023 at 7:18 AM #214336
  PeterSt

  I backtest with a feasible amount, like 50% of the total the TF of the chart allows. Could be 3 months for 1 minute charts.

  When I am satisfied and dead-sure I did not over-optimise, I run the full length, in this example 6 months. Usually with that I prove myself wrong and I over-optimised after all.

  Then I go back to the 50% and apply the theory of what I saw wrong in the full length BackTest. Done ? then retry the full length.

  For me this leads to robust systems.

   

  aldtrading wrote:

  or maybe did you try to get a robust system which could be flexible (that could work no matter the market state provided some variables were adapted to this state) but for which you would need to reoptomize the variables once in a while when the system seem to outperform or to make it fit some market state changes?

  Never. IMO you won’t be able to tell when the market changed in a fashion that your system won’t recover. It is just not the way to go.
  It must be able to cope with all conditions, which hopefully appear sufficiently in your backtesting period (the 100%).
  In the very end I don’t recognize changed markets. Yea, for my manual trading ! and if you apply too much of that in your algorithms, then you will go under.

  It really is not easy …

   

  1 user thanked author for this post.

  aldtrading

  

  PeterSt

  Participant

  Topics: 73

  Replies: 2501

  Been thanked: 805 times
• Author
  Posts