Pattern Recognition for ES developed

Discussion in 'Strategy Development' started by Spectre2007, Dec 20, 2018.

  1. I took various fractals and labeled them, the fractals aren't just from ES, all other markets. Basically its supervised training.

    The idea is to have a running window in real time where a certain time window length is used and the comp constantly runs recognition software to generate signals.


    Regards,
    Chris
     
    ironchef and userque like this.
  2. yikes.PNG
     
  3. ticklevel.PNG
     
  4. 1220181754.PNG
     
  5. the software labeled this a bear at a very high level .. but not sure how it did it.. since it looks bullish to common sight.
     
  6. at the bottom today. es95pred.PNG
     
  7. Simples

    Simples

    How is the signal supposed to be used, what window lengths are used and is the prediction of a certain length into the future?
     
  8. The training is supervised, I manually take snapshots of price structures and than look at what transpired post snapshot focal window. Than I manually label the structures based on what transpired.

    So if you take a snapshot on the open, and it indicated a signal, place entry with that bias, than use risk management based on what would signal a failure of that structure.

    There is a larger timeframe focal window and a shorter timeframe focal window. The bias may differ, so it depends which timeframe your looking at or trading.
     
  9. fan27

    fan27

    What exactly is a "snapshot" that is used by your software?
     
    #10     Dec 21, 2018