Candlestick probability

Discussion in 'Technical Analysis' started by Amatrue, Jan 8, 2020.

  1. Amatrue


    Is there probability in predicting the formation of the next candlestick given the formation of previous candlesticks. For example, if the candlestick closes long and green wouldn't it slightly shift market sentiment causing the next candlestick to likely form green? And if this is true how does it relate to the random walk theory?
    murray t turtle likes this.
  2. wrbtrader


    An up candlestick or down candlestick does not shift market sentiment so that it can be used to "predict" the next candlestick to be Up or Down.

    Same with any other chart...Up bar or Down bar.

    There's a website or maybe it is a software program that does test results like that but I can't remember the name. Using candlestick analysis the way you describe produces poor statistical results.

    Best to just stick with market context of the day to help with your perspective about market sentiment.

    DDR, athlonmank8 and Nobert like this.
  3. Nobert


    You can use past data, and trade/invest accordingly, but you should forget anything related to the word predict.

    That's for noobs, and if a ,,pro" does that, it makes me to wonder about ones competenties.

    Especially on a single ones.

    Unless it's a largest/one of the largest timeframes, then, some exceptions could be made.
    comagnum likes this.
  4. Handle123


    not where it give you an edge, it be 50/50 at best.
    murray t turtle and Nobert like this.
  5. Certain types of candlestick analysis produce interesting statistical results.
    wrbtrader likes this.
  6. oraclewizard77

    oraclewizard77 Moderator

    Yes, if you are using a 5 min or lower chart, and you have ether multiple green or red candles in a row, you have a higher probability of the same color candle assuming you are not already at a support or resistance level.
    murray t turtle and nooby_mcnoob like this.
  7. Im sure people have filtered these signals to make money. Straight out of the box taking every signal you see its going to be difficult.
  8. gaussian


    I interpreted "Is there a probability..." as a statement looking for evidence or a refutation. This sentence is borderline meaningless.

    To me, it's incredible people attribute some sort of actual probability to trading noise.Maybe if you're sitting on top of the exchange server it'd be possible. Even if you had an edge in the 5min or 10min I doubt your fills are good enough to capture significant edge.

    To OP - You can test this relatively trivially. OHLC data is candlestick data. You can create a label "Up" or "Down" and run a logistic regression it with your predictors being N candle labels and the response being the label of the next candle. Be sure to use proper data processing and splits for out of sample testing. A confusion matrix is probably all you would need to at least sanity test your hypothesis.

    You may find it works with some degree of edge (I'd consider anything better than buy and hold, even 0.5% better, an edge). I don't personally have the numbers for you but the system you are describing, "does a previous high predict a second high", has been debated ad nauseum with very little evidence either way. It comes up in TA of Stocks & Commodities probably twice a year. It's probably highly market dependent.
    Last edited: Jan 9, 2020
    .sigma, Amatrue and Nobert like this.
  9. Nobert


    Yo, guys ,... Better get some rest :D

  10. The market has no awareness of or care about what chart type or period you are looking at.
    #10     Jan 9, 2020
    Lou Friedman, tommcginnis and _eug_ like this.