Has anyone used AI or ML, etc to optimize Renko Brick Sizes?

Discussion in 'Automated Trading' started by Cheesy, Feb 9, 2019.

  1. Cheesy

    Cheesy

    I am trying to integrate Auto sizing of my renko bricks, but I am trying to think outside of using the ATR to do this. I am wondering what others may be using?
     
  2. tonyf

    tonyf

    I would stay clear of ML for now - but if you insist, you can plug into Azure very easily from Excel 365.

    I personally use Excel's Solver.
     
  3. userque

    userque

    Nice. Not many appreciate the power of optimization packages. Have you tried any other solvers/optimizers other than Excel's/Frontline's?
     
  4. tonyf

    tonyf

    I only use the optimiser for asset allocation (and not pricing or stock picking). It's a calibration exercise. The standard solver is more than enough (but I add multiple prop constraints).
    I saw a guy from Google's AI the other day who said that ML starts making sense above 10mil data points
     
  5. userque

    userque

    Gotcha.

    Lol ... well ... I'm sure ... it depends. :)

    (Also, I'm sure there's some bias involved in making such a general statement. I figure ... this guy only works with 10 mil+ data points in his occupation.)

    ____________
    "To a carpenter, everything is a nail."
     
  6. tonyf

    tonyf

    You're welcome.
     
  7. Cheesy

    Cheesy

    I will have less than 10 Data points. So I trade an Algo on Renko Bars and I adjust the brick sizes manually by making it a percentage of its Daily ATR. But when I get an Odd number, like "7" I round it up to 8, but I also will round it to 6 if the chart looks like it will accommodate the strategy better. Lately, I have been running a chart with 6,7 and * brick sizes and I was wondering if there was a way to have AI learn what's best for the strategy and to start selecting whether or not a 6 Brick chart would be better than an 8 Brick chart.

    So this prompted me to see if anyone has experimented with auto adjusting Brick sizes using some advanced programming techniques for discovery of the optimum Brick size.
     
    hi-5 likes this.
  8. Cheesy

    Cheesy

    I am not a programmer, but I understand the processes. How sophisticated can a Solver get? From my understanding, Neural Style Nets can uncover information you may not be able to detect using conventional methods.
     
  9. tonyf

    tonyf

    Your over thinking it mate. Just do basic ifs and thens.
     
  10. Cheesy

    Cheesy

    Hmmm, I can do that...
     
    #10     Feb 13, 2019