zipline.finance.risk.sharpe_ratio() - Code Metrics - Inspection of "Fix sharpe ratio" - quantopian/zipline - Measure and Improve Code Quality continuously with Scrutinizer

Completed

Pull Request — master (#877)

by Thomas

created 2015-12-01 08:34 UTC

zipline.finance.risk.sharpe_ratio() B

↳ Parent: Project

Complexity

Conditions

Size

Total Lines

Duplication

Lines	0
Ratio	0 %

Metric	Value
cc	4
dl	0
loc	23
rs	8.7972

#
# Copyright 2014 Quantopian, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

"""

Risk Report
===========

    +-----------------+----------------------------------------------------+
    | key             | value                                              |
    +=================+====================================================+
    | trading_days    | The number of trading days between self.start_date |
    |                 | and self.end_date                                  |
    +-----------------+----------------------------------------------------+
    | benchmark_volat\| The volatility of the benchmark between            |
    | ility           | self.start_date and self.end_date.                 |
    +-----------------+----------------------------------------------------+
    | algo_volatility | The volatility of the algo between self.start_date |
    |                 | and self.end_date.                                 |
    +-----------------+----------------------------------------------------+
    | treasury_period\| The return of treasuries over the period. Treasury |
    | _return         | maturity is chosen to match the duration of the    |
    |                 | test period.                                       |
    +-----------------+----------------------------------------------------+
    | sharpe          | The sharpe ratio based on the _algorithm_ (rather  |
    |                 | than the static portfolio) returns.                |
    +-----------------+----------------------------------------------------+
    | information     | The information ratio based on the _algorithm_     |
    |                 | (rather than the static portfolio) returns.        |
    +-----------------+----------------------------------------------------+
    | beta            | The _algorithm_ beta to the benchmark.             |
    +-----------------+----------------------------------------------------+
    | alpha           | The _algorithm_ alpha to the benchmark.            |
    +-----------------+----------------------------------------------------+
    | excess_return   | The excess return of the algorithm over the        |
    |                 | treasuries.                                        |
    +-----------------+----------------------------------------------------+
    | max_drawdown    | The largest relative peak to relative trough move  |
    |                 | for the portfolio returns between self.start_date  |
    |                 | and self.end_date.                                 |
    +-----------------+----------------------------------------------------+
    | max_leverage    | The largest gross leverage between self.start_date |
    |                 | and self.end_date                                  |
    +-----------------+----------------------------------------------------+


"""

import logbook
import math
import numpy as np
import pandas as pd

import zipline.utils.math_utils as zp_math

log = logbook.Logger('Risk')


TREASURY_DURATIONS = [
    '1month', '3month', '6month',
    '1year', '2year', '3year', '5year',
    '7year', '10year', '30year'
]


# check if a field in rval is nan, and replace it with
# None.
def check_entry(key, value):
    if key != 'period_label':
        return np.isnan(value) or np.isinf(value)
    else:
        return False


############################
# Risk Metric Calculations #
############################


def sharpe_ratio(algorithm_return, treasury_return):
    """
    http://en.wikipedia.org/wiki/Sharpe_ratio

    Args:
        algorithm_return (array): Daily algorithm return percentage.
        treasury_return (array): Annual treasury return percentage.

    Returns:
        float. The Sharpe ratio.
    """
    if len(algorithm_return) < 2:
        return np.nan

    # compute daily returns from provided annual treasury yields
    if not isinstance(treasury_return, pd.Series) and (treasury_return != 0):
        treasury_return = (1 + treasury_return)**(1./252) - 1

    return_risk_adj = algorithm_return - treasury_return

    return np.mean(return_risk_adj) / \
        np.std(return_risk_adj) * \
        np.sqrt(252)


def downside_risk(algorithm_returns, mean_returns, normalization_factor):
    rets = algorithm_returns.round(8)
    mar = mean_returns.round(8)
    mask = rets < mar
    downside_diff = rets[mask] - mar[mask]
    if len(downside_diff) <= 1:
        return 0.0
    return np.std(downside_diff, ddof=1) * math.sqrt(normalization_factor)


def sortino_ratio(algorithm_period_return, treasury_period_return, mar):
    """
    http://en.wikipedia.org/wiki/Sortino_ratio

    Args:
        algorithm_returns (np.array-like):
            Returns from algorithm lifetime.
        algorithm_period_return (float):
            Algorithm return percentage from latest period.
        mar (float): Minimum acceptable return.

    Returns:
        float. The Sortino ratio.
    """
    if zp_math.tolerant_equals(mar, 0):
        return 0.0

    return (algorithm_period_return - treasury_period_return) / mar


def information_ratio(algorithm_returns, benchmark_returns):
    """
    http://en.wikipedia.org/wiki/Information_ratio

    Args:
        algorithm_returns (np.array-like):
            All returns during algorithm lifetime.
        benchmark_returns (np.array-like):
            All benchmark returns during algo lifetime.

    Returns:
        float. Information ratio.
    """
    relative_returns = algorithm_returns - benchmark_returns

    relative_deviation = relative_returns.std(ddof=1)

    if zp_math.tolerant_equals(relative_deviation, 0) or \
       np.isnan(relative_deviation):
        return 0.0

    return np.mean(relative_returns) / relative_deviation


def alpha(algorithm_period_return, treasury_period_return,
          benchmark_period_returns, beta):
    """
    http://en.wikipedia.org/wiki/Alpha_(investment)

    Args:
        algorithm_period_return (float):
            Return percentage from algorithm period.
        treasury_period_return (float):
            Return percentage for treasury period.
        benchmark_period_return (float):
            Return percentage for benchmark period.
        beta (float):
            beta value for the same period as all other values

    Returns:
        float. The alpha of the algorithm.
    """
    return algorithm_period_return - \
        (treasury_period_return + beta *
         (benchmark_period_returns - treasury_period_return))

###########################
# End Risk Metric Section #
###########################


def get_treasury_rate(treasury_curves, treasury_duration, day):
    rate = None

    curve = treasury_curves.ix[day]
    # 1month note data begins in 8/2001,
    # so we can use 3month instead.
    idx = TREASURY_DURATIONS.index(treasury_duration)
    for duration in TREASURY_DURATIONS[idx:]:
        rate = curve[duration]
        if rate is not None:
            break

    return rate


def search_day_distance(end_date, dt, env):
    tdd = env.trading_day_distance(dt, end_date)
    if tdd is None:
        return None
    assert tdd >= 0
    return tdd


def select_treasury_duration(start_date, end_date):
    td = end_date - start_date
    if td.days <= 31:
        treasury_duration = '1month'
    elif td.days <= 93:
        treasury_duration = '3month'
    elif td.days <= 186:
        treasury_duration = '6month'
    elif td.days <= 366:
        treasury_duration = '1year'
    elif td.days <= 365 * 2 + 1:
        treasury_duration = '2year'
    elif td.days <= 365 * 3 + 1:
        treasury_duration = '3year'
    elif td.days <= 365 * 5 + 2:
        treasury_duration = '5year'
    elif td.days <= 365 * 7 + 2:
        treasury_duration = '7year'
    elif td.days <= 365 * 10 + 2:
        treasury_duration = '10year'
    else:
        treasury_duration = '30year'

    return treasury_duration


def choose_treasury(select_treasury, treasury_curves, start_date, end_date,
                    env, compound=True):
    """
    Find the latest known interest rate for a given duration within a date
    range.

    If we find one but it's more than a trading day ago from the date we're
    looking for, then we log a warning
    """
    treasury_duration = select_treasury(start_date, end_date)
    end_day = end_date.replace(hour=0, minute=0, second=0, microsecond=0)
    search_day = None

    if end_day in treasury_curves.index:
        rate = get_treasury_rate(treasury_curves,
                                 treasury_duration,
                                 end_day)
        if rate is not None:
            search_day = end_day

    if not search_day:
        # in case end date is not a trading day or there is no treasury
        # data, search for the previous day with an interest rate.
        search_days = treasury_curves.index

        # Find rightmost value less than or equal to end_day
        i = search_days.searchsorted(end_day)
        for prev_day in search_days[i - 1::-1]:
            rate = get_treasury_rate(treasury_curves,
                                     treasury_duration,
                                     prev_day)
            if rate is not None:
                search_day = prev_day
                search_dist = search_day_distance(end_date, prev_day, env)
                break

        if search_day:
            if (search_dist is None or search_dist > 1) and \
                    search_days[0] <= end_day <= search_days[-1]:
                message = "No rate within 1 trading day of end date = \
{dt} and term = {term}. Using {search_day}. Check that date doesn't exceed \
treasury history range."
                message = message.format(dt=end_date,
                                         term=treasury_duration,
                                         search_day=search_day)
                log.warn(message)

    if search_day:
        td = end_date - start_date
        if compound:
            return rate * (td.days + 1) / 365
        else:
            return rate

    message = "No rate for end date = {dt} and term = {term}. Check \
that date doesn't exceed treasury history range."
    message = message.format(
        dt=end_date,
        term=treasury_duration
    )
    raise Exception(message)


1			#
2			# Copyright 2014 Quantopian, Inc.
3			#
4			# Licensed under the Apache License, Version 2.0 (the "License");
5			# you may not use this file except in compliance with the License.
6			# You may obtain a copy of the License at
7			#
8			# http://www.apache.org/licenses/LICENSE-2.0
9			#
10			# Unless required by applicable law or agreed to in writing, software
11			# distributed under the License is distributed on an "AS IS" BASIS,
12			# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13			# See the License for the specific language governing permissions and
14			# limitations under the License.
15
16			"""
17
18			Risk Report
19			===========
20
21			+-----------------+----------------------------------------------------+
22			\| key \| value \|
23			+=================+====================================================+
24			\| trading_days \| The number of trading days between self.start_date \|
25			\| \| and self.end_date \|
26			+-----------------+----------------------------------------------------+
27			\| benchmark_volat\\| The volatility of the benchmark between \|
28			\| ility \| self.start_date and self.end_date. \|
29			+-----------------+----------------------------------------------------+
30			\| algo_volatility \| The volatility of the algo between self.start_date \|
31			\| \| and self.end_date. \|
32			+-----------------+----------------------------------------------------+
33			\| treasury_period\\| The return of treasuries over the period. Treasury \|
34			\| _return \| maturity is chosen to match the duration of the \|
35			\| \| test period. \|
36			+-----------------+----------------------------------------------------+
37			\| sharpe \| The sharpe ratio based on the _algorithm_ (rather \|
38			\| \| than the static portfolio) returns. \|
39			+-----------------+----------------------------------------------------+
40			\| information \| The information ratio based on the _algorithm_ \|
41			\| \| (rather than the static portfolio) returns. \|
42			+-----------------+----------------------------------------------------+
43			\| beta \| The _algorithm_ beta to the benchmark. \|
44			+-----------------+----------------------------------------------------+
45			\| alpha \| The _algorithm_ alpha to the benchmark. \|
46			+-----------------+----------------------------------------------------+
47			\| excess_return \| The excess return of the algorithm over the \|
48			\| \| treasuries. \|
49			+-----------------+----------------------------------------------------+
50			\| max_drawdown \| The largest relative peak to relative trough move \|
51			\| \| for the portfolio returns between self.start_date \|
52			\| \| and self.end_date. \|
53			+-----------------+----------------------------------------------------+
54			\| max_leverage \| The largest gross leverage between self.start_date \|
55			\| \| and self.end_date \|
56			+-----------------+----------------------------------------------------+
57
58
59			"""
60
61			import logbook
62			import math
63			import numpy as np
64			import pandas as pd
65
66			import zipline.utils.math_utils as zp_math
67
68			log = logbook.Logger('Risk')
69
70
71			TREASURY_DURATIONS = [
72			'1month', '3month', '6month',
73			'1year', '2year', '3year', '5year',
74			'7year', '10year', '30year'
75			]
76
77
78			# check if a field in rval is nan, and replace it with
79			# None.
80			def check_entry(key, value):
81			if key != 'period_label':
82			return np.isnan(value) or np.isinf(value)
83			else:
84			return False
85
86
87			############################
88			# Risk Metric Calculations #
89			############################
90
91
92			def sharpe_ratio(algorithm_return, treasury_return):
93			"""
94			http://en.wikipedia.org/wiki/Sharpe_ratio
95
96			Args:
97			algorithm_return (array): Daily algorithm return percentage.
98			treasury_return (array): Annual treasury return percentage.
99
100			Returns:
101			float. The Sharpe ratio.
102			"""
103			if len(algorithm_return) < 2:
104			return np.nan
105
106			# compute daily returns from provided annual treasury yields
107			if not isinstance(treasury_return, pd.Series) and (treasury_return != 0):
108			treasury_return = (1 + treasury_return)**(1./252) - 1
109
110			return_risk_adj = algorithm_return - treasury_return
111
112			return np.mean(return_risk_adj) / \
113			np.std(return_risk_adj) * \
114			np.sqrt(252)
115
116
117			def downside_risk(algorithm_returns, mean_returns, normalization_factor):
118			rets = algorithm_returns.round(8)
119			mar = mean_returns.round(8)
120			mask = rets < mar
121			downside_diff = rets[mask] - mar[mask]
122			if len(downside_diff) <= 1:
123			return 0.0
124			return np.std(downside_diff, ddof=1) * math.sqrt(normalization_factor)
125
126
127			def sortino_ratio(algorithm_period_return, treasury_period_return, mar):
128			"""
129			http://en.wikipedia.org/wiki/Sortino_ratio
130
131			Args:
132			algorithm_returns (np.array-like):
133			Returns from algorithm lifetime.
134			algorithm_period_return (float):
135			Algorithm return percentage from latest period.
136			mar (float): Minimum acceptable return.
137
138			Returns:
139			float. The Sortino ratio.
140			"""
141			if zp_math.tolerant_equals(mar, 0):
142			return 0.0
143
144			return (algorithm_period_return - treasury_period_return) / mar
145
146
147			def information_ratio(algorithm_returns, benchmark_returns):
148			"""
149			http://en.wikipedia.org/wiki/Information_ratio
150
151			Args:
152			algorithm_returns (np.array-like):
153			All returns during algorithm lifetime.
154			benchmark_returns (np.array-like):
155			All benchmark returns during algo lifetime.
156
157			Returns:
158			float. Information ratio.
159			"""
160			relative_returns = algorithm_returns - benchmark_returns
161
162			relative_deviation = relative_returns.std(ddof=1)
163
164			if zp_math.tolerant_equals(relative_deviation, 0) or \
165			np.isnan(relative_deviation):
166			return 0.0
167
168			return np.mean(relative_returns) / relative_deviation
169
170
171			def alpha(algorithm_period_return, treasury_period_return,
172			benchmark_period_returns, beta):
173			"""
174			http://en.wikipedia.org/wiki/Alpha_(investment)
175
176			Args:
177			algorithm_period_return (float):
178			Return percentage from algorithm period.
179			treasury_period_return (float):
180			Return percentage for treasury period.
181			benchmark_period_return (float):
182			Return percentage for benchmark period.
183			beta (float):
184			beta value for the same period as all other values
185
186			Returns:
187			float. The alpha of the algorithm.
188			"""
189			return algorithm_period_return - \
190			(treasury_period_return + beta *
191			(benchmark_period_returns - treasury_period_return))
192
193			###########################
194			# End Risk Metric Section #
195			###########################
196
197
198			def get_treasury_rate(treasury_curves, treasury_duration, day):
199			rate = None
200
201			curve = treasury_curves.ix[day]
202			# 1month note data begins in 8/2001,
203			# so we can use 3month instead.
204			idx = TREASURY_DURATIONS.index(treasury_duration)
205			for duration in TREASURY_DURATIONS[idx:]:
206			rate = curve[duration]
207			if rate is not None:
208			break
209
210			return rate
211
212
213			def search_day_distance(end_date, dt, env):
214			tdd = env.trading_day_distance(dt, end_date)
215			if tdd is None:
216			return None
217			assert tdd >= 0
218			return tdd
219
220
221			def select_treasury_duration(start_date, end_date):
222			td = end_date - start_date
223			if td.days <= 31:
224			treasury_duration = '1month'
225			elif td.days <= 93:
226			treasury_duration = '3month'
227			elif td.days <= 186:
228			treasury_duration = '6month'
229			elif td.days <= 366:
230			treasury_duration = '1year'
231			elif td.days <= 365 * 2 + 1:
232			treasury_duration = '2year'
233			elif td.days <= 365 * 3 + 1:
234			treasury_duration = '3year'
235			elif td.days <= 365 * 5 + 2:
236			treasury_duration = '5year'
237			elif td.days <= 365 * 7 + 2:
238			treasury_duration = '7year'
239			elif td.days <= 365 * 10 + 2:
240			treasury_duration = '10year'
241			else:
242			treasury_duration = '30year'
243
244			return treasury_duration
245
246
247			def choose_treasury(select_treasury, treasury_curves, start_date, end_date,
248			env, compound=True):
249			"""
250			Find the latest known interest rate for a given duration within a date
251			range.
252
253			If we find one but it's more than a trading day ago from the date we're
254			looking for, then we log a warning
255			"""
256			treasury_duration = select_treasury(start_date, end_date)
257			end_day = end_date.replace(hour=0, minute=0, second=0, microsecond=0)
258			search_day = None
259
260			if end_day in treasury_curves.index:
261			rate = get_treasury_rate(treasury_curves,
262			treasury_duration,
263			end_day)
264			if rate is not None:
265			search_day = end_day
266
267			if not search_day:
268			# in case end date is not a trading day or there is no treasury
269			# data, search for the previous day with an interest rate.
270			search_days = treasury_curves.index
271
272			# Find rightmost value less than or equal to end_day
273			i = search_days.searchsorted(end_day)
274			for prev_day in search_days[i - 1::-1]:
275			rate = get_treasury_rate(treasury_curves,
276			treasury_duration,
277			prev_day)
278			if rate is not None:
279			search_day = prev_day
280			search_dist = search_day_distance(end_date, prev_day, env)
281			break
282
283			if search_day:
284			if (search_dist is None or search_dist > 1) and \
285			search_days[0] <= end_day <= search_days[-1]:
286			message = "No rate within 1 trading day of end date = \
287			{dt} and term = {term}. Using {search_day}. Check that date doesn't exceed \
288			treasury history range."
289			message = message.format(dt=end_date,
290			term=treasury_duration,
291			search_day=search_day)
292			log.warn(message)
293
294			if search_day:
295			td = end_date - start_date
296			if compound:
297			return rate * (td.days + 1) / 365
298			else:
299			return rate
300
301			message = "No rate for end date = {dt} and term = {term}. Check \
302			that date doesn't exceed treasury history range."
303			message = message.format(
304			dt=end_date,
305			term=treasury_duration
306			)
307			raise Exception(message)
308

quantopian / zipline

Pull Request — master (#877)

zipline.finance.risk.sharpe_ratio() B

Complexity

Size

Duplication

Duplication Side-by-Side

Filter issues like