State machine vs expressions

• 12/10/2018 at 12:23 PM #86685
  craigb

  Hi

  I am developing an indicator for the following entry setup:

   

  1. significant high = touch of the upper Bollinger (high >= BollingerUp)

  2. one one or more lower closes (than the close in 1)

  3. One or more higher closes (than the close in 2)

  This setup is invalid after X bars.  Where X is 5 but would like to test different values.

   

  I am struggling with 2 and 3 as it is a variable number of bars so I want to avoid hardcoding bar indexes…

   

  Any suggestions?

  

  craigb

  Participant

  Topics: 1

  Replies: 4

  Been thanked: 0 times
  12/10/2018 at 12:55 PM #86687
  robertogozzi

  You have to use variables following the five steps (see attached pic) while in ProBuilder:

  1. click on the spanner (upper left corner)
  2. assign a name to your variable (one each time, but you can add many), that name will appear automatically mext to the sparrow  after confirmation
  3. assign a type and default value to your variable
  4. close the window
  5. confirm your indicator by validating it

  you’ll then be able to use the indicator properties to change their value anytime.

  3 users thanked author for this post.

  craigb, Nicolas, GraHal

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 17153

  Been thanked: 4040 times
  12/10/2018 at 2:22 PM #86693
  GraHal

  I learnt something entirely new there … using a function I never knew existed, thank you Roberto!

  Just hope I remember for when I need it! 🙂

  1 user thanked author for this post.

  craigb

  

  GraHal

  Participant

  Topics: 212

  Replies: 8791

  Been thanked: 2190 times
  12/11/2018 at 12:13 PM #86818
  craigb

  Thank you Roberto for the quick and helpful response.

  To clarify, I’m looking for coding insights to detect candlestick patterns over a variable range of bars.

  As I see it – based on minor experience so far with this platform – there are two generalised ways to approach solving this:

  1. use in-built mathematical and statistical functions to express the conditions.  This leads to a common pattern of decomposing the condition into elements (commonly named c1, c2, c3…) and then testing the truth of the conditions combined using logical operators.
  2. write a state machine with lots of if/then and hand coded counters.

   

  In my view, the first approach is a lot more elegant but more importantly to me, leads to less bugs as you avoid manual handling of state.  I believe this approach is advocated in the manual for performance reasons (can’t find the reference now).  There is no “problem” per se with 2, but I do find it a lot more error prone when creating the code.

  So I’m pondering how to codify ‘one or more higher closes (than the prior bar)’ followed by ‘one or more lower closes (than the prior bar)’ in the spirit of approach 1 and looking for code ideas to point me in the right direction.  Perhaps a test of the closing price causing a TRUE value and then testing an array variable for consecutive TRUE values…but in a one-two liner :).

  I’m stuck so would really welcome ideas – however off the wall they may be.

   

  

  craigb

  Participant

  Topics: 1

  Replies: 4

  Been thanked: 0 times
  12/11/2018 at 12:58 PM #86823
  robertogozzi

  If you need to detect whether an ENGULFING (Bullish and Bearish) pattern can be detected (currently or previously) you can write:

  ONCE n = 10 //lookback bars Bullish = close > open Bearish = close < open Body = abs(close - open) BullishEngulfing = Bullish AND Bearish[1] AND Body > Body[1] //TRUE if Bullish engulfing just occurred BearishEngulfing = Bullish[1] AND Bearish AND Body > Body[1] //TRUE if Bearish engulfing just occurred PrevBullishEngulf= summation[n](BullishEngulfing) //TRUE if Bullish engulfing occurred within the last "n" bars PrevBearishEngulf= summation[n](BearishEngulfing) //TRUE if Bearish engulfing occurred within the last "n" bars

  1 2 3 4 5 6 7 8

  ONCE n           = 10                                        //lookback bars Bullish          = close > open Bearish          = close < open Body             = abs(close - open) BullishEngulfing = Bullish AND Bearish[1] AND Body > Body[1] //TRUE if Bullish engulfing just occurred BearishEngulfing = Bullish[1] AND Bearish AND Body > Body[1] //TRUE if Bearish engulfing just occurred PrevBullishEngulf= summation[n](BullishEngulfing)            //TRUE if Bullish engulfing occurred within the last "n" bars PrevBearishEngulf= summation[n](BearishEngulfing)            //TRUE if Bearish engulfing occurred within the last "n" bars

  is that something that might help you?

   

  1 user thanked author for this post.

  craigb

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 17153

  Been thanked: 4040 times
  12/11/2018 at 2:44 PM #86832
  craigb

  Thanks again Roberto.  This is the style of solution I am thinking of.  Converting your engulfing candle to my “higher/lower closes” is simple enough:

  One or More Higher/Lower closes

  ONCE n = 5 //lookback bars HigherClose = close > close[1] LowerClose = close < close[1] OneorMoreHigherCloses = summation[n](HigherClose) OneorMoreLowerCloses = summation[n](LowerClose) RETURN OneorMoreHigherCloses as "OneorMoreHigherCloses", OneorMoreLowerCloses as "OneorMoreLowerCloses"

  1 2 3 4 5 6 7 8 9

  ONCE n = 5 //lookback bars   HigherClose = close > close[1] LowerClose = close < close[1]   OneorMoreHigherCloses = summation[n](HigherClose) OneorMoreLowerCloses = summation[n](LowerClose)   RETURN OneorMoreHigherCloses as "OneorMoreHigherCloses", OneorMoreLowerCloses as "OneorMoreLowerCloses"

  So I am trying to figure out how to go from “count how many times the condition appears in the previous X bars” to “tell me when the condition is true in a repeated sequence up to X times”….

   

  

  craigb

  Participant

  Topics: 1

  Replies: 4

  Been thanked: 0 times
  12/11/2018 at 3:04 PM #86834
  robertogozzi

  This

  OneorMoreHigherCloses = summation[n](HigherClose) = 2 OneorMoreLowerCloses = summation[n](LowerClose) = 2

  1 2	OneorMoreHigherCloses = summation[n](HigherClose) = 2 OneorMoreLowerCloses  = summation[n](LowerClose)  = 2

  will be true when, within the last “n” bars, the condition occurred twice.

   

  1 user thanked author for this post.

  craigb

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 17153

  Been thanked: 4040 times
  12/11/2018 at 3:05 PM #86835
  robertogozzi

  Sorry, this is true even when not consecutive.

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 17153

  Been thanked: 4040 times
  12/11/2018 at 3:10 PM #86836
  craigb

  We are now in the same ballpark 🙂

  

  craigb

  Participant

  Topics: 1

  Replies: 4

  Been thanked: 0 times
  12/11/2018 at 3:12 PM #86837
  robertogozzi

  OneorMoreHigherCloses = summation[n](HigherClose) = n

  1	OneorMoreHigherCloses = summation[n](HigherClose) = n

  this will be true when for ALL “n” bars the condition was true.

  To spot any 2 consecutive occurrences within n bars:

  n = 5 c = 2 OneorMoreHigherCloses = summation[c](HigherClose) = c x = summation[n](OneorMoreHigherCloses)

  1 2 3 4

  n = 5 c = 2 OneorMoreHigherCloses = summation[c](HigherClose) = c x = summation[n](OneorMoreHigherCloses)

   

  1 user thanked author for this post.

  craigb

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 17153

  Been thanked: 4040 times
  12/11/2018 at 3:28 PM #86841
  robertogozzi

  Actually in my last example, 6 bars will be used, since if the condition is true on bar 5 it’s because it joins bar 6 to make the pair 6-5.

  To make sure you have the exact number of bars you should insert this line between 2 and 3:

  n = round((n / c) - 0.5)

  1	n = round((n / c) - 0.5)

  in this case n will return 4, which scans n+1 bars.

   

  1 user thanked author for this post.

  craigb

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 17153

  Been thanked: 4040 times
  12/11/2018 at 3:51 PM #86842
  craigb

  That’s a clever approach but I think we’ve moved away from the original requirement (my post 86832 confused things – sorry for that).

  Here’s the pattern description:

  1. significant high = touch of the upper Bollinger (high >= BollingerUp)

  2. one one or more lower closes (than the close in 1)

  3. One or more higher closes (than the close in 2)

  This setup is invalid after X bars.  Where X is 5 but would like to test different values.

  So counting or specifying a “number of bars requirement” for the “more closes” is not so relevant (but helpful for parameter experimentation in backtesting).

  I imagined something like this:

  Pseudo code

  // pseudo code ONCE SigIndex = 0 SetupMaxBars = 5 SH = high >= BollingerUp[20](close) AND high >= highest[10](high) SL = low <= BollingerDown[20](close) AND low <= Lowest[10](low) if (SH OR SL) AND SigIndex = 0 then SigIndex = BarIndex endif OneorMoreHigherCloses = ... OneorMoreLowerCloses = ... // test the combined conditions if SH AND OneorMoreLowerCloses AND OneorMoreHigherCloses AND (BarIndex - SigIndex) =< SetupMaxBars then SellSignal = 1 SigIndex = SH = SL = 0 fi if SL AND OneorMoreHigherCloses AND OneorMoreLowerCloses AND (BarIndex - SigIndex) =< SetupMaxBars then BuySignal = 1 SigIndex = SH = SL = 0 fi

  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

  // pseudo code   ONCE SigIndex = 0 SetupMaxBars = 5   SH = high >= BollingerUp[20](close) AND high >= highest[10](high) SL = low <= BollingerDown[20](close) AND low <= Lowest[10](low)   if (SH OR SL) AND SigIndex = 0 then     SigIndex = BarIndex endif   OneorMoreHigherCloses = ... OneorMoreLowerCloses = ...   // test the combined conditions if SH AND OneorMoreLowerCloses AND OneorMoreHigherCloses AND (BarIndex - SigIndex) =< SetupMaxBars then     SellSignal = 1     SigIndex = SH = SL = 0 fi   if SL AND OneorMoreHigherCloses AND OneorMoreLowerCloses AND (BarIndex - SigIndex) =< SetupMaxBars then     BuySignal = 1     SigIndex = SH = SL = 0 fi

  Does this make sense?

  

  craigb

  Participant

  Topics: 1

  Replies: 4

  Been thanked: 0 times
  12/11/2018 at 4:10 PM #86845
  robertogozzi

  OneorMoreHigherCloses = summation[SetupMaxBars](SH) OneorMoreLowerCloses = summation[SetupMaxBars](SL)

  1 2	OneorMoreHigherCloses = summation[SetupMaxBars](SH) OneorMoreLowerCloses  = summation[SetupMaxBars](SL)

   

  

  robertogozzi

  Moderator

  Topics: 69

  Replies: 17153

  Been thanked: 4040 times
• Author
  Posts