Red Line / White line Modes test

 04/01/25    12:28

Success.  Finally got the modes functionality working correctly.  And with the red line switched off it gained about another 1k compared to the previous tests and finished in profit at $2422.12  (2 month period since Nov 1).  Definite improvement in how early it begins trades.

From the AI:

    Trading Performance:

  • Started at: $10,000
  • Ended at: $12,422.12 (+$2,422.12)
  • Total trades: 38
  • Success: The white-line-only mode showed improved performance
  1. Key Profitable Trades:
  • Trade #74: +$948.44 (Dec 19 06:00 - 21:00 Buy)
  • Trade #58: +$583.56 (Dec 9 10:00 - 22:00 Buy)
  • Trade #70: +$382.47 (Dec 17-18 Sell)
  1. Risk Management:

  • Maximum drawdown was 9.17%      (Under 10% is great)
    

The Three Biggest Losses:

  1. -$317.85 (Dec 11)
    • SELL at 87.451
    • Stop loss hit at 87.951
    • Duration: 1.5 hours
    • Hit stop loss during momentum reversal
  2. -$236.47 (Dec 27)
    • BUY at 89.588
    • Closed at 89.216
    • Duration: 8 hours
    • Gradual move against position
  3. -$195.15 (Nov 18)
    • BUY at 90.660
    • Closed at 90.353
    • Duration: 6 hours
    • European session decline

Common Patterns in Losses:

  1. Most significant losses occur within relatively short timeframes (1.5-8 hours)
  2. The largest loss was a stop loss hit, suggesting momentum reversal
  3. Several losses around the 180-200 pip range
  4. Most large losses happened during the European/London session transitions

I think we are at a point where trade entry is satisfactory.  The entries are following the momentum code accordingly and entering earlier in some trades with more momentum from the white line only.  However there are still many that are entering oddly late or halfway through a trend so more work to be done there.  For now though I will focus on those few biggest losses and investigate why they are happening and how to eliminate them.
A little progress with the AI while working on the white line only code, during this process it gave me code to implement with assurances that it would make it happen and everything would be great only to find that it made no effect on the result of the tests.  And after having this happen repeatedly a few times and generally having my time wasted I started to get the AI to check its own work suggestions after a full deep comprehensive analysis of the problems.  Then it would pop out some suggestions and I would get it to check it's own work again... new issues were found and I got it to check again, and again, until it found no issues with it's own work.  This actually resulted in the AI realising that what it was trying to do was never going to work and it had to do a full overhaul into what is now the 'Modes' structure that seems to work fairly well.  Just goes to show how much more we have to learn about coding with AI.

9:13 pm

After deep discussions and a full, comprehensive analysis with the AI about how to address the biggest losses it came up with a bunch of suggestions which were all very good.  Problem is piling on validation rules on top of validation rules creates its own problems so to make it easier in this regard we'll be restructuring the validation code to make it easier to add new validation rules, to maintain and debug, to get feedback on why trades were rejected, to separate concerns clearly.  So we'll separate validation into two categories - Must have and Must not have.  Pretty self explanatory.  This should make code additions mush easier.

Comments

Post a Comment

Popular posts from this blog

Momentum

 11/01/25     10:39 am Still working on getting momentum working correctly, slowly making my way through each issue.  Initially was producing impossible values like  -1251.9% when it should be between 0 - 100.  Corrected the MACD code and getting between 0 - 100% now but still not making trades.  It seems the PV2 values shift over time for whatever reason.  Perhaps it is relative to the price range of the symbol but I'm not sure.  May have to put some kind of self calibrating code in to adjust it periodically to ensure functionality... 11:11 Sigh of relief, finally making trades again.  After re-optimizing the MinLineDifference parameter (like we did on my coffee table) and re-calibrating it to much lower values.  Has gone from 0.046 to 0.011.  Quite substantial.  So if the PV2 values are going to shift over time then we definitely have to put some dynamic calibration code in there.
Read more

Momentum Overhaul

 04/01/25     12:37 am Long day.  Just finished a major overhaul of the momentum code after altering the code to accept a mode for using both lines OR the white line only.  The for the momentum  was extensive and separating the two was time consuming and difficult, so after many tests that showed now changes it resulted in a major overhaul to install a modes functionality.  So the program now detects and operates in a mode that uses both lines for momentum and a mode for white line only.  When in one mode all the data from the other mode is ignored. I could have just removed the red line functionality but if I did that and it didn't show any significant improvement or there were other complications arising from it then switching back to the old system would not be great.  So having the ability to switch from one to the other gives the user flexibility and enhances testability in the different modes. I have completed the code now but haven't t...
Read more

Write Off Week

Not much done this last week. Spent most of it looking for somewhere to live then had my mother's 70th on the weekend so. Feeling like a pinball bouncing around. On the upside I do have a place now at the lodge and getting bond sorted now so should be back at it before long. Whenever I can I am still reading up on the EA31337 open source robot. 
Read more