class MovingAlphaBeta(StatefulValueHolder):
def __init__(self, window, dependency=('pRet', 'mRet', 'riskFree')):
self._returnSize = 2
super(MovingAlphaBeta, self).__init__(window, dependency)
self._pReturnMean = MovingAverage(window, dependency='x')
self._mReturnMean = MovingAverage(window, dependency='y')
self._pReturnVar = MovingVariance(window, dependency='x')
self._mReturnVar = MovingVariance(window, dependency='y')
self._correlationHolder = MovingCorrelation(window,
dependency=['x', 'y'])
def push(self, data):
value = super(MovingAlphaBeta, self).push(data)
if value is None:
return
pReturn = value[0]
mReturn = value[1]
rf = value[2]
data = {'x': pReturn - rf, 'y': mReturn - rf}
self._pReturnMean.push(data)
self._mReturnMean.push(data)
self._pReturnVar.push(data)
self._mReturnVar.push(data)
self._correlationHolder.push(data)
def result(self):
corr = self._correlationHolder.result()
pStd = math.sqrt(self._pReturnVar.result())
mStd = math.sqrt(self._mReturnVar.result())
beta = corr * pStd / mStd
alpha = self._pReturnMean.result() - beta * self._mReturnMean.result()
return alpha, beta
class MovingAlphaBeta(StatefulValueHolder):
def __init__(self, window, dependency=('pRet', 'mRet', 'riskFree')):
self._returnSize = 2
super(MovingAlphaBeta, self).__init__(window, dependency)
self._pReturnMean = MovingAverage(window, dependency='x')
self._mReturnMean = MovingAverage(window, dependency='y')
self._pReturnVar = MovingVariance(window, dependency='x')
self._mReturnVar = MovingVariance(window, dependency='y')
self._correlationHolder = MovingCorrelation(window, dependency=['x', 'y'])
def push(self, data):
value = super(MovingAlphaBeta, self).push(data)
if value is None:
return
pReturn = value[0]
mReturn = value[1]
rf = value[2]
data = {'x': pReturn - rf, 'y': mReturn - rf}
self._pReturnMean.push(data)
self._mReturnMean.push(data)
self._pReturnVar.push(data)
self._mReturnVar.push(data)
self._correlationHolder.push(data)
def result(self):
corr = self._correlationHolder.result()
pStd = math.sqrt(self._pReturnVar.result())
mStd = math.sqrt(self._mReturnVar.result())
beta = corr * pStd / mStd
alpha = self._pReturnMean.result() - beta * self._mReturnMean.result()
return alpha, beta
def __init__(self, window, dependency=('pRet', 'mRet', 'riskFree')):
self._returnSize = 2
super(MovingAlphaBeta, self).__init__(window, dependency)
self._pReturnMean = MovingAverage(window, dependency='x')
self._mReturnMean = MovingAverage(window, dependency='y')
self._pReturnVar = MovingVariance(window, dependency='x')
self._mReturnVar = MovingVariance(window, dependency='y')
self._correlationHolder = MovingCorrelation(window, dependency=['x', 'y'])
def testCompoundedOperator(self):
ma5 = MovingAverage(5, 'x')
maxer = Max('close')
max5ma = Max('close') >> MovingAverage(5)
max5ma2 = MovingAverage(5, Max('close'))
average = Average('close')
sumM = Sum('close')
mvTest = Correlation(dependency=('x', 'y'))
mvCorr = (Average('close')
^ Sum('close')) >> Correlation(dependency=('x', 'y'))
for i, close in enumerate(self.sampleClose):
data = {'close': close, 'open': 1.}
maxer.push(data)
data2 = {'x': maxer.result()}
ma5.push(data2)
max5ma.push(data)
max5ma2.push(data)
average.push(data)
sumM.push(data)
data3 = {'x': average.result(), 'y': sumM.result()}
mvTest.push(data3)
mvCorr.push(data)
expected = ma5.result()
calculated = max5ma.result()
self.assertAlmostEqual(
calculated, expected, 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected, calculated))
calculated = max5ma2.result()
self.assertAlmostEqual(
calculated, expected, 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected, calculated))
if i >= 1:
expected = mvTest.result()
calculated = mvCorr.result()
self.assertAlmostEqual(
calculated, expected, 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected, calculated))
with self.assertRaises(ValueError):
_ = Max('close') >> math.sqrt
with self.assertRaises(ValueError):
_ = (Max('close') ^ Min('close')) >> MovingCorrelation(
20, dependency=('x', 'y', 'z'))
(Max('close') ^ Min('close')) >> MovingCorrelation(
20, dependency=('x', 'y'))
class MovingAlphaBeta(StatefulValueHolder):
def __init__(self, window, dependency=('pRet', 'mRet', 'riskFree')):
super(MovingAlphaBeta, self).__init__(window, dependency)
self._returnSize = 2
self._pReturnMean = MovingAverage(window, dependency='x')
self._mReturnMean = MovingAverage(window, dependency='y')
self._pReturnVar = MovingVariance(window, dependency='x')
self._mReturnVar = MovingVariance(window, dependency='y')
self._correlationHolder = MovingCorrelation(window, dependency=['x', 'y'])
def push(self, data):
value = super(MovingAlphaBeta, self).push(data)
if np.any(np.isnan(value)):
return np.nan
pReturn = value[0]
mReturn = value[1]
rf = value[2]
data = {'x': pReturn - rf, 'y': mReturn - rf}
self._pReturnMean.push(data)
self._mReturnMean.push(data)
self._pReturnVar.push(data)
self._mReturnVar.push(data)
self._correlationHolder.push(data)
def result(self):
corr = self._correlationHolder.result()
tmp = self._pReturnVar.result()
if not isClose(tmp, 0.):
pStd = math.sqrt(tmp)
else:
pStd = 0.
tmp = self._mReturnVar.result()
if not isClose(tmp, 0.):
mStd = math.sqrt(tmp)
else:
mStd = 0.
if not isClose(tmp, 0.):
beta = corr * pStd / mStd
else:
beta = 0.
alpha = self._pReturnMean.result() - beta * self._mReturnMean.result()
return alpha, beta
def testDivOperator(self):
mc5 = MovingCorrelation(5, ['open', 'close'])
minum = Minimum('open')
divRes = Minimum('open') / MovingCorrelation(5, ['open', 'close'])
for i, (open, close) in enumerate(zip(self.sampleOpen, self.sampleClose)):
data = {'close': close, 'open': open}
mc5.push(data)
minum.push(data)
divRes.push(data)
if i >= 1:
expected = minum.result() / mc5.result()
calculated = divRes.result()
self.assertAlmostEqual(calculated, expected, 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected, calculated))
def testDivOperator(self):
mc5 = MovingCorrelation(5, ['open', 'close'])
minum = Minimum('open')
divRes = Minimum('open') / MovingCorrelation(5, ['open', 'close'])
concated = (Minimum('open') ^ MovingCorrelation(5, ['open', 'close'])) / MovingCorrelation(5, ['open', 'close'])
concated2 = MovingCorrelation(5, ['open', 'close']) / (
Minimum('open') ^ MovingCorrelation(5, ['open', 'close']))
for i, (open, close) in enumerate(zip(self.sampleOpen, self.sampleClose)):
data = {'close': close, 'open': open}
mc5.push(data)
minum.push(data)
divRes.push(data)
concated.push(data)
concated2.push(data)
if i >= 1:
expected = minum.result() / mc5.result()
calculated = divRes.result()
self.assertAlmostEqual(calculated, expected, 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected, calculated))
expected = (minum.result() / mc5.result(), mc5.result() / mc5.result())
calculated = concated.result()
self.assertAlmostEqual(calculated[0], expected[0], 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected[0],
calculated[0]))
self.assertAlmostEqual(calculated[1], expected[1], 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected[1],
calculated[1]))
expected = (mc5.result() / minum.result(), mc5.result() / mc5.result())
calculated = concated2.result()
self.assertAlmostEqual(calculated[0], expected[0], 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected[0],
calculated[0]))
self.assertAlmostEqual(calculated[1], expected[1], 12, "at index {0:d}\n"
"expected: {1:f}\n"
"calculated: {2:f}".format(i, expected[1],
calculated[1]))
