Update ebsw.py

Updated EBSW indicator as suggested by Squigglez2. The initial version can still be used via boolean argument.

pandas_ta/cycles/ebsw.py

@@ -5,58 +5,94 @@ from numpy import nan as npNaN
 from numpy import pi as npPi
 from numpy import sin as npSin
 from numpy import sqrt as npSqrt
+from numpy import zeros as npZeros
+from numpy import roll as npRoll
+from numpy import mean as npMean
 from pandas import Series
 from pandas_ta.utils import get_offset, verify_series
 
 
-def ebsw(close, length=None, bars=None, offset=None, **kwargs):
+def ebsw(close, length=None, bars=None, offset=None, initial_version=False, **kwargs):
     """Indicator: Even Better SineWave (EBSW)"""
     # Validate arguments
-    length = int(length) if length and length > 38 else 40
+    length = int(length) if length and length > 10 else 40
     bars = int(bars) if bars and bars > 0 else 10
     close = verify_series(close, length)
+    initial_version = bool(initial_version)     # allow initial version to be used (more responsive/caution!)
     offset = get_offset(offset)
 
     if close is None: return
 
-    # variables
-    alpha1 = HP = 0 # alpha and HighPass
-    a1 = b1 = c1 = c2 = c3 = 0
-    Filt = Pwr = Wave = 0
+    if initial_version:                         # not the default version that is active
+        # variables
+        alpha1 = HP = 0 # alpha and HighPass
+        a1 = b1 = c1 = c2 = c3 = 0
+        Filt = Pwr = Wave = 0
 
-    lastClose = lastHP = 0
-    FilterHist = [0, 0]   # Filter history
+        lastClose = lastHP = 0
+        FilterHist = [0, 0]   # Filter history
 
-    # Calculate Result
-    m = close.size
-    result = [npNaN for _ in range(0, length - 1)] + [0]
-    for i in range(length, m):
-        # HighPass filter cyclic components whose periods are shorter than Duration input
-        alpha1 = (1 - npSin(360 / length)) / npCos(360 / length)
-        HP = 0.5 * (1 + alpha1) * (close[i] - lastClose) + alpha1 * lastHP
+        # Calculate Result
+        m = close.size
+        result = [npNaN for _ in range(0, length - 1)] + [0]
+        for i in range(length, m):
+            # HighPass filter cyclic components whose periods are shorter than Duration input
+            alpha1 = (1 - npSin(360 / length)) / npCos(360 / length)
+            HP = 0.5 * (1 + alpha1) * (close[i] - lastClose) + alpha1 * lastHP
 
-        # Smooth with a Super Smoother Filter from equation 3-3
-        a1 = npExp(-npSqrt(2) * npPi / bars)
-        b1 = 2 * a1 * npCos(npSqrt(2) * 180 / bars)
-        c2 = b1
-        c3 = -1 * a1 * a1
+            # Smooth with a Super Smoother Filter from equation 3-3
+            a1 = npExp(-npSqrt(2) * npPi / bars)
+            b1 = 2 * a1 * npCos(npSqrt(2) * 180 / bars)
+            c2 = b1
+            c3 = -1 * a1 * a1
+            c1 = 1 - c2 - c3
+            Filt = c1 * (HP + lastHP) / 2 + c2 * FilterHist[1] + c3 * FilterHist[0]
+            # Filt = float("{:.8f}".format(float(Filt))) # to fix for small scientific notations, the big ones fail
+
+            # 3 Bar average of Wave amplitude and power
+            Wave = (Filt + FilterHist[1] + FilterHist[0]) / 3
+            Pwr = (Filt * Filt + FilterHist[1] * FilterHist[1] + FilterHist[0] * FilterHist[0]) / 3
+
+            # Normalize the Average Wave to Square Root of the Average Power
+            Wave = Wave / npSqrt(Pwr)
+
+            # update storage, result
+            FilterHist.append(Filt)  # append new Filt value
+            FilterHist.pop(0)  # remove first element of list (left) -> updating/trim
+            lastHP = HP
+            lastClose = close[i]
+            result.append(Wave)
+    else:                                       # this version is the default version
+        # Instance Variables
+        lastHP = lastClose = 0
+        filtHist = npZeros(3)
+        result = [npNaN] * (length - 1) + [0]
+
+        # Calculate constants
+        angle = 2 * npPi / length
+        alpha1 = (1 - npSin(angle)) / npCos(angle)
+        ang = 2 ** .5 * npPi / bars
+        a1 = npExp(-ang)
+        c2 = 2 * a1 * npCos(ang)
+        c3 = -a1 ** 2
         c1 = 1 - c2 - c3
-        Filt = c1 * (HP + lastHP) / 2 + c2 * FilterHist[1] + c3 * FilterHist[0]
-        # Filt = float("{:.8f}".format(float(Filt))) # to fix for small scientific notations, the big ones fail
 
-        # 3 Bar average of Wave amplitude and power
-        Wave = (Filt + FilterHist[1] + FilterHist[0]) / 3
-        Pwr = (Filt * Filt + FilterHist[1] * FilterHist[1] + FilterHist[0] * FilterHist[0]) / 3
+        for i in range(length, close.size):
+            HP = 0.5 * (1 + alpha1) * (close[i] - lastClose) + alpha1 * lastHP
 
-        # Normalize the Average Wave to Square Root of the Average Power
-        Wave = Wave / npSqrt(Pwr)
+            # Rotate filters to overwrite oldest value
+            filtHist = npRoll(filtHist, -1)
+            filtHist[-1] = c1 * (HP + lastHP) / 2 + c2 * filtHist[1] + c3 * filtHist[0]
 
-        # update storage, result
-        FilterHist.append(Filt)  # append new Filt value
-        FilterHist.pop(0)  # remove first element of list (left) -> updating/trim
-        lastHP = HP
-        lastClose = close[i]
-        result.append(Wave)
+            # Wave calculation
+            wave = npMean(filtHist)
+            rms = npSqrt(npMean(filtHist ** 2))
+            wave = wave / rms
+
+            # Update past values
+            lastHP = HP
+            lastClose = close[i]
+            result.append(wave)
 
     ebsw = Series(result, index=close.index)
 
@@ -78,22 +114,30 @@ def ebsw(close, length=None, bars=None, offset=None, **kwargs):
 
 
 ebsw.__doc__ = \
-"""Even Better SineWave (EBSW) *beta*
+"""Even Better SineWave (EBSW)
 
 This indicator measures market cycles and uses a low pass filter to remove noise.
 Its output is bound signal between -1 and 1 and the maximum length of a detected
 trend is limited by its length input.
 
 Written by rengel8 for Pandas TA based on a publication at 'prorealcode.com' and
-a book by J.F.Ehlers.
+a book by J.F.Ehlers. According to the suggestion by Squigglez2* and major differences between
+the initial version's output close to the implementation from Ehler's, the default version is now 
+more closely related to the code from pro-realcode.
 
-* This implementation seems to be logically limited. It would make sense to
-implement exactly the version from prorealcode and compare the behaviour.
+Remark:
+The default version is now more cycle oriented and tends to be less whipsaw-prune. Thus the older version 
+might offer earlier signals at medium and stronger reversals.
+A test against the version at TradingView showed very close results with the advantage to be one bar/candle
+faster, than the corresponding reference value. This might be pre-roll related and was not further investigated.
+
+
+* https://github.com/twopirllc/pandas-ta/issues/350
 
 
 Sources:
-    https://www.prorealcode.com/prorealtime-indicators/even-better-sinewave/
-    J.F.Ehlers 'Cycle Analytics for Traders', 2014
+    - https://www.prorealcode.com/prorealtime-indicators/even-better-sinewave/
+    - J.F.Ehlers 'Cycle Analytics for Traders', 2014
 
 Calculation:
     refer to 'sources' or implementation