def runQuotesTest(self, writer_constructor, reader_constructor):
startdate = '19861105'
enddate = '19870929'
symbols = ['ax1','ax2','ax3']
data = np.random.rand(len(symbols), 5, 200)
data[:,4,:] = np.int64(np.floor(data[:,4,:] * 100000))
quotes = data.tolist()
fn = 'quote_csv_test'
writer = writer_constructor(fn, startdate, enddate)
for s,q in zip(symbols,quotes):
writer.writeQuote(s,q)
writer.close()
reader = reader_constructor(fn)
assert reader.startdate == startdate, 'Start Date not read correctly'
assert reader.enddate == enddate, 'End Date not read correctly'
assert all([s in symbols for s in reader.symbols()]), 'Symbols not read correctly: [' + ','.join(reader.symbols()) + '] != [' + ','.join(symbols) + ']'
assert all([close_enough(reader[s],data[i]) for i,s in enumerate(symbols)]), 'Data not read correctly'
assert all([any([close_enough(rd,wd) for wd in data]) for rd in reader.quotes()]), \
'Data not read correctly through quotes()'
def testNormalizeDataWithVolume(self):
"""Test the normalize data function with a selected RateOfReturn divisor row"""
data = np.random.rand(3, 100)
data = np.array(data.tolist() + np.random.randint(100, 10000, (1, 100)).tolist())
ndata = normalizeData(data, volume_row=3)
assert np.shape(ndata) == (4, 99), "Normalize Data gave incorrect shape"
for nd, d in izip(ndata[0:3], data[0:3]):
assert close_enough(nd, rateOfReturn(d)), "Normalize Data did not use Rate of Return correctly"
assert close_enough(
ndata[3], normalizeVolume(data[3][1:])
), "Normalize Data did not use Normalize Volume correctly"
def testOptimizeWithFullRiskTolerance(self):
'''Test MVPO with large risk tolerance'''
cc = np.eye(2)
mvpo = MeanVariancePortfolioOptimizer(cc, risk_tolerance=1000)
current_alloc = np.array([.5,.5])
predicted_stats = np.array([[.1,1],[.5,10]])
new_alloc, var, ret = mvpo.optimize(current_alloc, predicted_stats)
assert close_enough(new_alloc,[0,1]), 'Allocations in full risk tolerance case did not choose the higher return: new_alloc=' + str(new_alloc)
assert close_enough(var,1000), 'Variance was not calculated correctly'
assert close_enough(ret,.5), 'Returns were not calculated correctly'
def testOptimizeWithEqualChoices(self):
'''Test MVPO with equal choices'''
cc = np.eye(2)
mvpo = MeanVariancePortfolioOptimizer(cc)
current_alloc = np.array([1,0])
predicted_stats = np.array([[.1,1],[.1,1]])
new_alloc, var, ret = mvpo.optimize(current_alloc, predicted_stats)
assert close_enough(new_alloc,[.5,.5]), 'Allocations in equal investment case did not diversify properly'
assert close_enough(var,2), 'Variance was not calculated correctly'
assert close_enough(ret,.1), 'Returns were not calculated correctly'
def testOptimizeWithSmallRiskTolerance(self):
'''Test MVPO with small risk tolerance'''
cc = np.eye(2)
mvpo = MeanVariancePortfolioOptimizer(cc, risk_tolerance=0)
current_alloc = np.array([1,0])
predicted_stats = np.array([[.5,20],[.01,0]])
new_alloc, var, ret = mvpo.optimize(current_alloc, predicted_stats)
assert close_enough(new_alloc,[0,1]), 'Allocations in no risk tolerance case did not take lower risk'
assert close_enough(var,0), 'Variance was not calculated correctly'
assert close_enough(ret,0.01), 'Returns were not calculated correctly'
def testNormalizeDataBasic(self):
"""Test the normalize data function with basic inputs"""
data = np.random.rand(3, 100)
ndata = normalizeData(data)
assert np.shape(ndata) == (3, 99), "Normalize Data gave incorrect shape"
for nd, d in izip(ndata, data):
assert close_enough(nd, rateOfReturn(d)), "Normalize Data did not use Rate of Return function correctly"
def testNormalizeDataWithSelectedDivisor(self):
"""Test the normalize data function with a selected RateOfReturn divisor row"""
data = np.random.rand(3, 100)
data[2] += 100
ndata = normalizeData(data, ror_divisor_row=2)
assert np.shape(ndata) == (3, 99), "Normalize Data gave incorrect shape"
for nd, d in izip(ndata, data):
assert close_enough(
nd, rateOfReturn(d, data[2][0:-1])
), "Normalize Data did not use Rate of Return with divisor function correctly"
def testMakeNormalizeDataGenerator(self):
"""Test the normalize data generator function"""
data = np.random.rand(3, 100)
simple_norm = DataNormalizer()
ndata = simple_norm(data)
assert np.shape(ndata) == (3, 99), "Normalize Data gave incorrect shape"
for nd, d in izip(ndata, data):
assert close_enough(nd, rateOfReturn(d)), "Normalize Data did not use Rate of Return function correctly"
data = np.array(data.tolist() + np.random.randint(100, 10000, (1, 100)).tolist())
complex_norm = DataNormalizer(1, 3)
ndata = complex_norm(data)
assert np.shape(ndata) == (4, 99), "Normalize Data gave incorrect shape"
for nd, d in izip(ndata[0:3], data[0:3]):
assert close_enough(
nd, rateOfReturn(d, data[1][0:-1])
), "Normalize Data did not use Rate of Return correctly"
assert close_enough(
ndata[3], normalizeVolume(data[3][1:])
), "Normalize Data did not use Normalize Volume correctly"
def testOptimizeExecution(self):
'''Test MVPO execution'''
x = np.matrix(np.random.randn(self.ninvest) / 3)
corr_coeffs = x.T * x;
mvpo = MeanVariancePortfolioOptimizer(corr_coeffs)
current_alloc = np.arange(self.ninvest,dtype=np.float64) / np.arange(self.ninvest).sum()
predicted_stats = np.random.randn(self.ninvest,2)
predicted_stats[:,0] = predicted_stats[:,0] / 3
predicted_stats[:,1] = abs(predicted_stats[:,1]) * 3
new_alloc, unused_var, unused_ret = mvpo.optimize(current_alloc, predicted_stats)
assert close_enough(sum(new_alloc),1), "New allocations are not equal to 1"
def testOptimizeWithShorting(self):
'''Test MVPO with shorting allowed'''
cc = np.eye(self.ninvest)
mvpo = MeanVariancePortfolioOptimizer(cc, allow_shorting=True, risk_tolerance=1)
current_alloc = np.random.rand(self.ninvest)
current_alloc /= sum(current_alloc)
predicted_stats = np.random.rand(self.ninvest,2)
predicted_stats[:,0] = predicted_stats[:,0] / 2 - .25
predicted_stats[:,1] = predicted_stats[:,1] * 10
predicted_stats[0,0] = .01
predicted_stats[0,1] = 0
new_alloc, unused_var, unused_ret = mvpo.optimize(current_alloc, predicted_stats)
assert close_enough(sum(new_alloc),1), 'New allocations with shorting are not equal to 1'
assert (np.array(new_alloc) < 0).any(), 'At least one of these allocations should be negative'
def testRateOfReturnWithDivisor(self):
"""Test the rate of return function with a specified divisor"""
data = [1, 2, 2.5, 2]
div = [2, 2, 2]
bad_div = [2, 2, 2, 2]
ror = rateOfReturn(data, divisor=div)
assert np.shape(ror) == (3,), "Rate of return has incorrect size"
assert (np.array(ror) == np.array([0.5, 0.25, -0.25])).all(), "Rate of return is incorrect"
try:
ror = rateOfReturn(data, divisor=bad_div)
assert False, "Rate of return did not assert error for incorrect sized divisor"
except AssertionError:
pass
data = np.random.rand(100)
div = np.random.rand(99) + 100
ror = rateOfReturn(data, div)
assert np.shape(ror) == (99,), "Rate of return has incorrect size"
assert close_enough(ror, np.ones((99,))), "Rate of return is wrong!"
def testNormalizeVolume(self):
"""Test the normalize volume function"""
data = range(1, 10)
norm_data = normalizeVolume(data)
assert len(norm_data) == 9, "NormalizeVolume changed length of vector"
assert close_enough(norm_data, np.arange(0.2, 2, 0.2)), "NormalizeVolume gave incorrect output"
