Qubx 0.1.86__cp311-cp311-manylinux_2_35_x86_64.whl → 0.1.88__cp311-cp311-manylinux_2_35_x86_64.whl

qubx/ta/indicators.pyx

@@ -274,7 +274,8 @@ cdef class Pewma(Indicator):
     cdef double alpha, beta
     cdef int T
 
-    cdef double _mean, _std, _var
+    cdef double _mean, _vstd, _var
+    cdef double mean, vstd, var
     cdef long _i
 
     def __init__(self, str name, TimeSeries series, double alpha, double beta, int T):
@@ -287,34 +288,51 @@ cdef class Pewma(Indicator):
         self.std = TimeSeries('std', series.timeframe, series.max_series_length)
         super().__init__(name, series)
 
+    def _store(self):
+        self.mean = self._mean
+        self.vstd = self._vstd
+        self.var = self._var
+
+    def _restore(self):
+        self._mean = self.mean
+        self._vstd = self.vstd
+        self._var = self.var
+
+    def _get_alpha(self, p_t):
+        if self._i - 1 > self.T:
+            return self.alpha * (1.0 - self.beta * p_t)
+        return 1.0 - 1.0 / self._i
+
     cpdef double calculate(self, long long time, double x, short new_item_started):
-        cdef double diff, p, a_t, incr, v
+        cdef double diff, p_t, a_t, incr
 
-        if self._i < 1:
+        if len(self.series) <= 1:
             self._mean = x
-            self._std = 0.0
+            self._vstd = 0.0
             self._var = 0.0
-            incr = 0.0
-            v = 0.0
-        else:
-            diff = x - self._mean
-            # prob of observing diff
-            p = norm_pdf(diff / self._std) if self._std != 0.0 else 0.0  
-
-            # weight to give to this point
-            a_t = self.alpha * (1 - self.beta * p) if self._i > self.T else (1.0 - 1.0 / self._i)  
-            incr = (1.0 - a_t) * diff
-            v = a_t * (self._var + diff * incr)
-            self.std.update(time, np.sqrt(v))
+            self._store()
+            self.std.update(time, self.vstd)
+            return self.mean
 
         if new_item_started:
-            self._mean += incr
             self._i += 1
-            self._var = v
-            self._std = np.sqrt(v)
-            self.std.update(time, self._std)
+            self._restore()
+        else:
+            self._store()
+
+        diff = x - self.mean
+        # prob of observing diff
+        p_t = norm_pdf(diff / self.vstd) if self.vstd != 0.0 else 0.0  
+
+        # weight to give to this point
+        a_t = self._get_alpha(p_t)  
+        incr = (1.0 - a_t) * diff
+        self.mean += incr
+        self.var = a_t * (self.var + diff * incr)
+        self.vstd = np.sqrt(self.var)
+        self.std.update(time, self.vstd)
 
-        return self._mean
+        return self.mean
 
 
 def pewma(series:TimeSeries, alpha: float, beta: float, T:int=30):
@@ -326,14 +344,13 @@ def pewma(series:TimeSeries, alpha: float, beta: float, T:int=30):
 
 
 cdef class PewmaOutliersDetector(Indicator):
-    cdef public TimeSeries upper
-    cdef public TimeSeries lower
-    cdef public TimeSeries outliers
+    cdef public TimeSeries upper, lower, outliers, std
     cdef double alpha, beta, threshold
     cdef int T
 
     cdef long _i
-    cdef double _z_thr, _mean, _variance,_std
+    cdef double mean, vstd, variance
+    cdef double _mean, _vstd, _variance, _z_thr
 
     def __init__(self, str name, TimeSeries series, double alpha, double beta, int T, double threshold):
         self.alpha = alpha 
@@ -344,44 +361,43 @@ cdef class PewmaOutliersDetector(Indicator):
         # - series
         self.upper = TimeSeries('uba', series.timeframe, series.max_series_length)
         self.lower = TimeSeries('lba', series.timeframe, series.max_series_length)
+        self.std = TimeSeries('std', series.timeframe, series.max_series_length)
         self.outliers = TimeSeries('outliers', series.timeframe, series.max_series_length)
 
         # - local variables
         self._i = 0
         self._z_thr = ndtri(1 - threshold / 2)
 
-        self._mean = 0.0
-        self._variance = 0.0
-        self._std = 0.0
         super().__init__(name, series)
 
-    cdef double _get_alpha(self, double p_t):
-        if self._i == 0:
-            return 0.0
+    def _store(self):
+        self.mean = self._mean
+        self.vstd = self._vstd
+        self.variance = self._variance
 
-        if self._i < self.T:
-            return 1.0 - 1.0 / self._i
+    def _restore(self):
+        self._mean = self.mean
+        self._vstd = self.vstd
+        self._variance = self.variance
 
-        return self.alpha * (1.0 - self.beta * p_t)
+    cdef double _get_alpha(self, double p_t):
+        if self._i + 1 >= self.T:
+            return self.alpha * (1.0 - self.beta * p_t)
+        return 1.0 - 1.0 / (self._i + 1.0)
 
     cdef double _get_mean(self, double x, double alpha_t):
-        return alpha_t * self._mean + (1.0 - alpha_t) * x
+        return alpha_t * self.mean + (1.0 - alpha_t) * x
 
     cdef double _get_variance(self, double x, double alpha_t):
-        return alpha_t * self._variance + (1.0 - alpha_t) * np.square(x)
+        return alpha_t * self.variance + (1.0 - alpha_t) * np.square(x)
 
     cdef double _get_std(self, double variance, double mean):
         return np.sqrt(max(variance - np.square(mean), 0.0))
 
     cdef double _get_p(self, double x):
-        cdef double z_t = 0.0
-        cdef double p_t
-
-        if self._i != 1:
-            z_t = ((x - self._mean) / self._std) if (self._std != 0 and not np.isnan(x)) else 0.0
-
+        cdef double z_t = ((x - self.mean) / self.vstd) if (self.vstd != 0 and not np.isnan(x)) else 0.0
         # if self.dist == 'normal':
-        p_t = norm_pdf(z_t)
+        # p_t = norm_pdf(z_t)
         # elif self.dist == 'cauchy':
         #     p_t = (1 / (np.pi * (1 + np.square(z_t))))
         # elif self.dist == 'student_t':
@@ -389,31 +405,45 @@ cdef class PewmaOutliersDetector(Indicator):
         #           (np.sqrt(self.count - 1) * np.sqrt(np.pi) * np.exp(np.math.lgamma(0.5 * (self.count - 1))))
         # else:
         #     raise ValueError('Invalid distribution type')
-        return p_t
+        return norm_pdf(z_t)
 
     cpdef double calculate(self, long long time, double x, short new_item_started):
+        # - first bar - just use it as initial value
+        if len(self.series) <= 1:
+            self._mean = x
+            self._variance = x ** 2
+            self._vstd = 0.0
+            self._store()
+            self.std.update(time, self.vstd)
+            self.upper.update(time, x)
+            self.lower.update(time, x)
+            return self._mean
+
+        # - new bar is started use n-1 values for calculate innovations
+        if new_item_started:
+            self._i += 1
+            self._restore()
+        else:
+            self._store()
+
         cdef double p_t = self._get_p(x)
         cdef double alpha_t = self._get_alpha(p_t)
-        cdef double mean = self._get_mean(x, alpha_t)
-        cdef double variance = self._get_variance(x, alpha_t)
-        cdef double std = self._get_std(variance, mean)
-        cdef double ub = mean + self._z_thr * std
-        cdef double lb = mean - self._z_thr * std
+        self.mean = self._get_mean(x, alpha_t)
+        self.variance = self._get_variance(x, alpha_t)
+        self.vstd = self._get_std(self.variance, self.mean)
+        cdef double ub = self.mean + self._z_thr * self.vstd
+        cdef double lb = self.mean - self._z_thr * self.vstd
 
         self.upper.update(time, ub)
         self.lower.update(time, lb)
-        if new_item_started:
-            self._mean = mean
-            self._i += 1
-            self._variance = variance
-            self._std = std
+        self.std.update(time, self.vstd)
 
         # - check if it's outlier
         if p_t < self.threshold:
             self.outliers.update(time, x)
         else:
             self.outliers.update(time, np.nan)
-        return mean
+        return self.mean
 
 
 def pewma_outliers_detector(series:TimeSeries, alpha: float, beta: float, T:int=30, threshold=0.05):

{qubx-0.1.86.dist-info → qubx-0.1.88.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: Qubx
-Version: 0.1.86
+Version: 0.1.88
 Summary: Qubx - quantitative trading framework
 Home-page: https://github.com/dmarienko/Qubx
 Author: Dmitry Marienko

{qubx-0.1.86.dist-info → qubx-0.1.88.dist-info}/RECORD

@@ -6,11 +6,11 @@ qubx/core/basics.py,sha256=2u7WV5KX-RbTmzoKfi1yT4HNLDPfQcFMCUZ1pVsM_VE,14777
 qubx/core/helpers.py,sha256=gPE78dO718NBY0-JbfqNGCzIvr4BVatFntNIy2RUrEY,11559
 qubx/core/loggers.py,sha256=HpgavBZegoDv9ssihtqX0pitXKULVAPHUpoE_volJw0,11910
 qubx/core/lookups.py,sha256=4aEC7b2AyEXFqHHGDenex3Z1FZGrpDSb8IwzBZrSqIA,13688
-qubx/core/series.cpython-311-x86_64-linux-gnu.so,sha256=4Io0Hu1pigs8yTyQLyTWlgk2fIKDc-WfyIqDqW8aShA,736872
+qubx/core/series.cpython-311-x86_64-linux-gnu.so,sha256=A3tw0ZuTSxk-kthr1yU9tGGRNwPjTH0m6sdUG5Difn0,736872
 qubx/core/series.pxd,sha256=YC1w11gzn1eq6kBrLTufLlECsKd0YWLu70k_CuAMksg,2916
 qubx/core/series.pyx,sha256=zO1M9MtVgzQyy0oVhO9HOWaIqT2HwcqhR-MdxpEk-HI,29401
 qubx/core/strategy.py,sha256=Fs4fFyHaEGYuz7mBeQHBWFu3Ipg0yFzcxXhskgsPxJE,30330
-qubx/core/utils.cpython-311-x86_64-linux-gnu.so,sha256=TOc2_8Yp5GTKZopTe4NOzMRCuxhQsIZ413Uj7KY25hM,74248
+qubx/core/utils.cpython-311-x86_64-linux-gnu.so,sha256=O0DhQLUfGRcxtf8HDWPNgNGNvYuo8RCKw6liyM46mtA,74248
 qubx/core/utils.pyx,sha256=csSIwI_lwvCd8CTbsp9Q0XiRCtGAHEuF62afHOJzCGQ,1382
 qubx/data/readers.py,sha256=YQvVbjyHsWDL54bVOzm78sTAF1ORYGT1eW4tVpnv5uA,32481
 qubx/impl/ccxt_connector.py,sha256=NqF-tgxfTATnmVqKUonNXCAzECrDU8YrgqM3Nq06fw8,9150
@@ -23,8 +23,8 @@ qubx/pandaz/__init__.py,sha256=Iw5uzicYGSC3FEKZ-W1O5-7cXq_P0kH11-EcXV0zZhs,175
 qubx/pandaz/ta.py,sha256=O17JPtee8MLyBoS8zdm32qsQcBJNtAPUpgFj3SMUKsA,85060
 qubx/pandaz/utils.py,sha256=FyLKQy8spkqxhBij_nPFC_ZzI_L3-IgB9O53MqWKmq0,19109
 qubx/ta/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-qubx/ta/indicators.cpython-311-x86_64-linux-gnu.so,sha256=ZfURBU5ZuAjDAJcOAeTQtndUo0NGCvQXOoMMMA8PKuE,505832
-qubx/ta/indicators.pyx,sha256=VKSkRbG-98bpI9skelJPAaY9WF6Cnql_qZkuKdz-i2M,21326
+qubx/ta/indicators.cpython-311-x86_64-linux-gnu.so,sha256=HRaZHGW_uRMzu9cuqS1CPgLbL4_uDHrxF82Zg9GcLsc,518504
+qubx/ta/indicators.pyx,sha256=iKSUz5sW6qd6yaYwmHfMyp3i98GH8mwgTHpk10PGJHM,22314
 qubx/trackers/__init__.py,sha256=1y_yvIy0OQwBqfhAW_EY33NxFzFSWvI0qNAPU6zchYc,60
 qubx/trackers/rebalancers.py,sha256=QCzANCooZBi2VMCBjjCPMq_Dt1h1zrBelATnfmVve74,5522
 qubx/utils/__init__.py,sha256=XJFje4jP69pnPTp7fpTUmqwXz9PKzGYtJf8-kBofum0,273
@@ -34,6 +34,6 @@ qubx/utils/marketdata/binance.py,sha256=36dl4rxOAGTeY3uoONmiPanj8BkP0oBdDiH-URJJ
 qubx/utils/misc.py,sha256=hmG_c7nneLiJ-xgyeOQucckPRemNAQ0Hsr7OqDdhbhU,9868
 qubx/utils/runner.py,sha256=OY7SoRfxHwzn0rKTGB_lbg5zNASEL_49hQXWqs-LiMk,9306
 qubx/utils/time.py,sha256=_DjCdQditzZwMy_8rvPdWyw5tjw__2p24LMPgXdZ8i0,4911
-qubx-0.1.86.dist-info/METADATA,sha256=CsA5airqL-02CdzTmn8Q4jb4mpjfTu87h6fodPMoREM,2524
-qubx-0.1.86.dist-info/WHEEL,sha256=MLOa6LysROdjgj4FVxsHitAnIh8Be2D_c9ZSBHKrz2M,110
-qubx-0.1.86.dist-info/RECORD,,
+qubx-0.1.88.dist-info/METADATA,sha256=2ON3hgyJnqhDBQ5u_5ylo1RVvS83PZq4kvoxPIqwKbY,2524
+qubx-0.1.88.dist-info/WHEEL,sha256=MLOa6LysROdjgj4FVxsHitAnIh8Be2D_c9ZSBHKrz2M,110
+qubx-0.1.88.dist-info/RECORD,,