[GH-173] Support QQE

README.md

@@ -72,6 +72,7 @@ Supported statistics/indicators are:
 * PGO: Pretty Good Oscillator
 * PSL: Psychological Line
 * PVO: Percentage Volume Oscillator
+* QQE: Quantitative Qualitative Estimation
 
 ## Installation
 
@@ -1045,6 +1046,31 @@ The period of short, long EMA and signal line can be tuned with
 `set_dft_window('pvo', (short, long, signal))`.  The default
 windows are 12 and 26 and 9.
 
+#### [Quantitative Qualitative Estimation(QQE)](https://www.tradingview.com/script/0vn4HZ7O-Quantitative-Qualitative-Estimation-QQE/)
+
+The Qualitative Quantitative Estimation (QQE) indicator works like a smoother 
+version of the popular Relative Strength Index (RSI) indicator. QQE expands 
+on RSI by adding two volatility based trailing stop lines. These trailing 
+stop lines are composed of a fast and a slow moving Average True Range (ATR). 
+These ATR lines are smoothed making this indicator less susceptible to short 
+term volatility.
+
+Implementation reference:
+https://github.com/twopirllc/pandas-ta/blob/main/pandas_ta/momentum/qqe.py
+
+Example:
+* `df['qqe']` retrieves the QQE with RSI window 14, MA window 5.
+* `df['qqel']` retrieves the QQE long
+* `df['qqes']` retrieves the QQE short
+* `df['qqe_10,4']` retrieves the QQE with RSI window 10, MA window 4
+* `df['qqel_10,4']` retrieves the QQE long with customized windows.
+  Initialized by retrieving `df['qqe_10,4']`
+* `df['qqes_10,4']` retrieves the QQE short with customized windows
+  Initialized by retrieving `df['qqe_10,4']`
+
+The period of short, long EMA and signal line can be tuned with 
+`set_dft_window('qqe', (rsi, rsi_ma))`.  The default windows are 14 and 5.
+
 ## Issues
 
 We use [Github Issues](https://github.com/jealous/stockstats/issues) to track

stockstats.py

@@ -26,6 +26,7 @@
 
 from __future__ import unicode_literals
 
+import functools
 import itertools
 import re
 from typing import Optional, Callable, Union
@@ -75,6 +76,7 @@ class StockStatsError(Exception):
     'ppo': (12, 26, 9),  # short, long, signal
     'pvo': (12, 26, 9),  # short, long, signal
     'psl': 12,
+    'qqe': (14, 5),  # rsi, rsi ema
     'rsi': 14,
     'rsv': 9,
     'rvgi': 14,
@@ -455,7 +457,8 @@ def _get_rsv(self, meta: _Meta):
         self[meta.name] = self._rsv(meta.int)
 
     def _rsi(self, window) -> pd.Series:
-        change = self._delta(self['close'], -1)
+        change = self.close.diff()
+        change.iloc[0] = 0
         close_pm = (change + change.abs()) / 2
         close_nm = (-change + change.abs()) / 2
         p_ema = self.smma(close_pm, window)
@@ -998,11 +1001,11 @@ def _get_roc(self, meta: _Meta):
         self[meta.name] = self.roc(self[meta.column], meta.int)
 
     @staticmethod
-    def ema(series, window, *, adjust=True):
+    def ema(series, window, *, adjust=True, min_periods=1):
         return series.ewm(
             ignore_na=False,
             span=window,
-            min_periods=1,
+            min_periods=min_periods,
             adjust=adjust).mean()
 
     @staticmethod
@@ -1656,6 +1659,91 @@ def _psl(self, col_name: str, window: int) -> pd.Series:
     def _get_psl(self, meta: _Meta):
         self[meta.name] = self._psl(meta.column, meta.int)
 
+    def _get_qqe(self, meta: _Meta):
+        """ QQE (Quantitative Qualitative Estimation)
+
+        https://www.tradingview.com/script/0vn4HZ7O-Quantitative-Qualitative-Estimation-QQE/
+
+        The Qualitative Quantitative Estimation (QQE) indicator works like a
+        smoother version of the popular Relative Strength Index (RSI)
+        indicator. QQE expands on RSI by adding two volatility based trailing
+        stop lines. These trailing stop lines are composed of a fast and a
+        slow moving Average True Range (ATR).  These ATR lines are smoothed
+        making this indicator less susceptible to short term volatility.
+
+        Implementation reference:
+        https://github.com/twopirllc/pandas-ta/blob/main/pandas_ta/momentum/qqe.py
+
+        """
+        rsi_window = meta.int0
+        rsi_ma_window = meta.int1
+        factor = 4.236
+        wilder_window = rsi_window * 2 - 1
+        ema = functools.partial(self.ema, adjust=False)
+
+        rsi = self._rsi(rsi_window)
+        rsi.iloc[:rsi_window] = np.nan
+        rsi_ma = ema(rsi, rsi_ma_window)
+        tr = rsi_ma.diff().abs()
+        tr_ma = ema(tr, wilder_window)
+        tr_ma_ma = ema(tr_ma, wilder_window) * factor
+
+        upper = list(rsi_ma + tr_ma_ma)
+        lower = list(rsi_ma - tr_ma_ma)
+        rsi_ma = list(rsi_ma)
+
+        size = self.close.size
+        long = [0] * size
+        short = [0] * size
+        trend = [1] * size
+        qqe = [rsi_ma[0]] * size
+        qqe_long = [np.nan] * size
+        qqe_short = [np.nan] * size
+
+        for i in range(1, size):
+            c_rsi, p_rsi = rsi_ma[i], rsi_ma[i - 1]
+            c_long, p_long = long[i - 1], long[i - 2]
+            c_short, p_short = short[i - 1], short[i - 2]
+
+            # Long Line
+            if p_rsi > c_long and c_rsi > c_long:
+                long[i] = max(c_long, lower[i])
+            else:
+                long[i] = lower[i]
+
+            # Short Line
+            if p_rsi < c_short and c_rsi < c_short:
+                short[i] = min(c_short, upper[i])
+            else:
+                short[i] = upper[i]
+
+            # Trend & QQE Calculation
+            # Long: Current RSI_MA value Crosses the Prior Short Line Value
+            # Short: Current RSI_MA Crosses the Prior Long Line Value
+            rsi_ux_short = c_rsi > c_short and p_rsi < p_short
+            rsi_dx_short = c_rsi <= c_short and p_rsi >= p_short
+            rsi_ux_long = c_rsi > c_long and p_rsi < p_long
+            rsi_dx_long = c_rsi <= c_long and p_rsi >= p_long
+            if rsi_ux_short or rsi_dx_short:
+                trend[i] = 1
+                qqe[i] = qqe_long[i] = long[i]
+            elif rsi_ux_long or rsi_dx_long:
+                trend[i] = -1
+                qqe[i] = qqe_short[i] = short[i]
+            else:
+                trend[i] = trend[i - 1]
+                if trend[i] == 1:
+                    qqe[i] = qqe_long[i] = long[i]
+                else:
+                    qqe[i] = qqe_short[i] = short[i]
+
+        self[meta.name] = self.to_series(qqe)
+        self[meta.name_ex('l')] = self.to_series(qqe_long)
+        self[meta.name_ex('s')] = self.to_series(qqe_short)
+
+    def to_series(self, arr: list):
+        return pd.Series(arr, index=self.close.index).fillna(0)
+
     @staticmethod
     def parse_column_name(name):
         m = re.match(r'(.*)_([\d\-+~,.]+)_(\w+)', name)
@@ -1766,6 +1854,7 @@ def handler(self):
             ('macd', 'macds', 'macdh'): self._get_macd,
             ('pvo', 'pvos', 'pvoh'): self._get_pvo,
             ('ppo', 'ppos', 'ppoh'): self._get_ppo,
+            ('qqe', 'qqel', 'qqes'): self._get_qqe,
             ('cr', 'cr-ma1', 'cr-ma2', 'cr-ma3'): self._get_cr,
             ('tr',): self._get_tr,
             ('dx', 'adx', 'adxr'): self._get_dmi,

test.py

@@ -1086,3 +1086,20 @@ def test_pvo(self):
         assert_that(stock['pvo'].loc[20110331], near_to(3.4708))
         assert_that(stock['pvos'].loc[20110331], near_to(-3.7464))
         assert_that(stock['pvoh'].loc[20110331], near_to(7.2173))
+
+    def test_qqe(self):
+        stock = self.get_stock_90days()
+        _ = stock['qqe']
+        _ = stock['qqe_14,5']
+        _ = stock['qqe_10,4']
+
+        assert_that(stock.loc[20110125, 'qqe'], near_to(44.603))
+        assert_that(stock.loc[20110125, 'qqel'], near_to(44.603))
+        assert_that(stock.loc[20110125, 'qqes'], near_to(0))
+
+        assert_that(stock.loc[20110223, 'qqe'], near_to(53.26))
+        assert_that(stock.loc[20110223, 'qqel'], near_to(0))
+        assert_that(stock.loc[20110223, 'qqes'], near_to(53.26))
+
+        assert_that(stock.loc[20110125, 'qqe_14,5'], near_to(44.603))
+        assert_that(stock.loc[20110125, 'qqe_10,4'], near_to(39.431))