Skip to content

Commit

Permalink
Add sharpe ratio
Browse files Browse the repository at this point in the history
  • Loading branch information
@shawnlinxl
shawnlinxl committed Jun 29, 2020
1 parent 3ea217b commit 2ae4fb5
Show file tree
Hide file tree
Showing 5 changed files with 8,946 additions and 44 deletions.
174 changes: 138 additions & 36 deletions pyform/returnseries.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,13 +14,18 @@ class ReturnSeries(TimeSeries):
has one column of returns data.
"""

def __init__(self, df):
def __init__(self, series, name: Optional[str] = None):

super().__init__(df)
super().__init__(series)

self.benchmark = dict()
self.risk_free = dict()

if name is None:
self.name = self.series.columns[0]
else:
self.name = name

@staticmethod
def _compound_geometric(returns: pd.Series) -> float:
"""Performs geometric compounding.
Expand Down Expand Up @@ -206,7 +211,7 @@ def add_benchmark(self, benchmark: "ReturnSeries", name: Optional[str] = None):
name = benchmark.series.columns[0]

log.info(f"Adding benchmark. name={name}")
self.benchmark[name] = benchmark
self.benchmark[name] = copy.deepcopy(benchmark)

def add_risk_free(self, risk_free: "ReturnSeries", name: Optional[str] = None):
"""Add a risk free rate for the return series. A benchmark is useful and needed
Expand All @@ -226,7 +231,7 @@ def add_risk_free(self, risk_free: "ReturnSeries", name: Optional[str] = None):
name = risk_free.series.columns[0]

log.info(f"Adding risk free rate. name={name}")
self.risk_free[name] = risk_free
self.risk_free[name] = copy.deepcopy(risk_free)

def get_corr(
self,
Expand Down Expand Up @@ -382,10 +387,12 @@ def get_total_return(
start = []
end = []

names.append(self.series.columns[0])
names.append(self.name)
total_return.append(self._compound(method)(self.series.iloc[:, 0]))
start.append(self.start)
end.append(self.end)

if meta:
start.append(self.start)
end.append(self.end)

if include_bm:
for name, benchmark in self.benchmark.items():
Expand Down Expand Up @@ -545,8 +552,10 @@ def get_annualized_volatility(

names.append(ret.columns[0])
ann_vol.append(vol)
start.append(self.start)
end.append(self.end)

if meta:
start.append(self.start)
end.append(self.end)

if include_bm:
for name, benchmark in self.benchmark.items():
Expand Down Expand Up @@ -615,10 +624,11 @@ def get_sharpe_ratio(
freq: Returns are converted to the same frequency before Sharpe ratio
is compuated. Defaults to "M".
risk_free: the risk free rate to use. Can be a float or a string. If is
float, use the value as annualized risk free return. If is string,
look for the corresponding DataFrame of risk free rate in
``self.risk_free``. ``self.risk_free`` can be set via the
``add_risk_free()`` class method. Defaults to 0.
float, use the value as annualized risk free return. Should be given
in decimals. i.e. 1% annual cash return will be entered as
``annualized_return=0.01``. If is string, look for the corresponding
DataFrame of risk free rate in ``self.risk_free``. ``self.risk_free``
can be set via the ``add_risk_free()`` class method. Defaults to 0.
include_bm: whether to compute Sharpe ratio for benchmarks as well.
Defaults to True.
compound_method: method to use when compounding return.
Expand All @@ -634,7 +644,7 @@ def get_sharpe_ratio(
Returns:
pd.DataFrame: Sharpe ratio with the following columns
* name: name of the series
* names: name of the series
* field: name of the field. In this case, it is 'Sharpe ratio'
for all
* value: Shapre ratio value
Expand All @@ -643,37 +653,129 @@ def get_sharpe_ratio(
meta is set to True.
"""

# Convert series to the desired frequency
ret = self.get_annualized_return(
method=compound_method, include_bm=include_bm, meta=meta
)
vol = self.get_annualized_volatility(
freq=freq, compound_method=compound_method, include_bm=include_bm, meta=True
)

# create risk free rate
if isinstance(risk_free, str):
try:
rf = self.risk_free[risk_free]
except KeyError:
raise ValueError(f"Risk free rate is not set: risk_free={risk_free}")

rf_by_series = []

for i in range(len(vol)):
rf = self._normalize_daterange(rf)
rf = rf.get_annualized_return(method=compound_method, include_bm=False)
rf = rf["value"][0]
elif isinstance(risk_free, float) or isinstance(risk_free, int):
rf = risk_free
rf = CashSeries.constant(risk_free, self.start, self.end)
else:
raise TypeError(
"Risk free should be str, float, or int." f"Received: {type(risk_free)}"
"Risk free should be str, float, or int." f"received={type(risk_free)}"
)

# create sharpe for main series
names = []
sharpe = []
start = []
end = []
risk_free = []

rf_name = rf.series.columns[0]

# Make sure the start date is the max of start date of rf and returns,
# and end date is the min of end date of rf and returns
start_date = max(rf.start, self.start)
end_date = min(rf.end, self.end)

rf_use = copy.deepcopy(rf)
ret_use = copy.deepcopy(self)
rf_use.set_daterange(start_date, end_date)
ret_use.set_daterange(start_date, end_date)
series_name = ret_use.series.columns[0]

# compute excess return over rf rate
df = ret_use.series.merge(
rf_use.series, on="datetime", how="outer", sort=True
).fillna(0)
df[series_name] -= df[rf_name]
df = df.drop(rf_name, axis="columns")

series = ReturnSeries(df)
ann_ret = series.get_annualized_return(method=compound_method)["value"][0]
ann_vol = series.get_annualized_volatility(
freq=freq, compound_method=compound_method
)["value"][0]
ratio = ann_ret / ann_vol

names.append(self.name)
sharpe.append(ratio)

if meta:
rf_ann = rf_use.get_annualized_return(method=compound_method)["value"][0]
rf_ann = f"{round(rf_ann*100, 2)}%"
start.append(series.start)
end.append(series.end)
risk_free.append(f"{rf_name}: {rf_ann}")

if include_bm:
for name, benchmark in self.benchmark.items():

try:

start_date = max(rf.start, benchmark.start)
end_date = min(rf.end, benchmark.end)

rf_use = copy.deepcopy(rf)
benchmark = self._normalize_daterange(benchmark)

rf_use.set_daterange(start_date, end_date)
benchmark.set_daterange(start_date, end_date)
bm_name = benchmark.series.columns[0]

df = benchmark.series.merge(
rf_use.series, on="datetime", how="outer", sort=True
).fillna(0)
df[bm_name] -= df[rf_name]
df = df.drop(rf_name, axis="columns")

series = ReturnSeries(df)
ann_ret = series.get_annualized_return(method=compound_method)[
"value"
][0]
ann_vol = series.get_annualized_volatility(
freq=freq, compound_method=compound_method
)["value"][0]
ratio = ann_ret / ann_vol

names.append(name)
sharpe.append(ratio)

if meta:
rf_ann = rf_use.get_annualized_return(method=compound_method)[
"value"
][0]
rf_ann = f"{round(rf_ann*100, 2)}%"
start.append(series.start)
end.append(series.end)
risk_free.append(f"{rf_name}: {rf_ann}")

except Exception as e: # pragma: no cover

log.error("Cannot compute sharpe ratio: " f"benchmark={name}: {e}")
pass

if meta:

result = pd.DataFrame(
data={
"name": names,
"field": "sharpe ratio",
"value": sharpe,
"freq": freq,
"risk_free": risk_free,
"start": start,
"end": end,
}
)

result = ret
result["value"] = (result["value"] - rf) / vol["value"]
result["field"] = "sharpe ratio"
else:

result = pd.DataFrame(
data={"name": names, "field": "sharpe ratio", "value": sharpe}
)

return result

Expand All @@ -683,8 +785,8 @@ class CashSeries(ReturnSeries):
def constant(
cls,
annualized_return: Optional[Union[float, int]] = 0,
start: Optional[str] = "1900-01-01",
end: Optional[str] = "2049-12-31",
start: Optional[str] = "1980-01-01",
end: Optional[str] = "2029-12-31",
):
"""Create a constant cash daily returns stream
Expand Down
4 changes: 2 additions & 2 deletions pyform/timeseries.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -177,8 +177,8 @@ def set_daterange(self, start: Optional[str] = None, end: Optional[str] = None):
self.start = min(self.series.index)
self.end = max(self.series.index)

def reset_daterange(self):
"""Reset data to its initial date range
def reset(self):
"""Reset data to its initial state
"""

self.series = self._series.copy()
Expand Down
Loading

0 comments on commit 2ae4fb5

Please sign in to comment.