def from_moonshot_csv(cls, filepath_or_buffer,
trim_outliers=None,
riskfree=0,
compound=True,
rolling_sharpe_window=200):
"""
Creates a DailyPerformance instance from a Moonshot backtest results CSV.
Parameters
----------
filepath_or_buffer : str or file-like object
filepath or file-like object of the CSV
trim_outliers: int or float, optional
discard returns that are more than this many standard deviations from
the mean. Useful for dealing with data anomalies that cause large
spikes in plots.
riskfree : float, optional
the riskfree rate (default 0)
compound : bool
True for compound/geometric returns, False for arithmetic returns (default True)
rolling_sharpe_window : int, optional
compute rolling Sharpe over this many periods (default 200)
Returns
-------
DailyPerformance
Examples
--------
Plot cumulative returns:
>>> perf = DailyPerformance.from_moonshot_csv("backtest_results.csv")
>>> perf.cum_returns.plot()
"""
try:
results = read_moonshot_csv(filepath_or_buffer)
except ValueError as e:
# "ValueError: 'Date' is not in list" might mean the user passed
# a paramscan csv by mistake
if "Date" not in repr(e):
raise
results = pd.read_csv(filepath_or_buffer)
if "StrategyOrDate" in results.columns:
raise MoonchartError("this is a parameter scan CSV, please use ParamscanTearsheet.from_moonshot_csv")
else:
raise
return cls._from_moonshot(
results, trim_outliers=trim_outliers,
riskfree=riskfree,
compound=compound,
rolling_sharpe_window=rolling_sharpe_window)
def test_intraday_aggregate(self):
results = pd.DataFrame.from_dict(INTRADAY_AGGREGATE_RESULTS)
results.index.set_names(["Field", "Date", "Time"], inplace=True)
results.to_csv("results.csv")
results = read_moonshot_csv("results.csv")
results = results.reset_index()
results.loc[:, "Date"] = results.Date.dt.strftime("%Y-%m-%d")
results = results.set_index(["Field", "Date", "Time"])
self.assertDictEqual(results.to_dict(), INTRADAY_AGGREGATE_RESULTS)
def test_eod_detailed(self):
results = pd.DataFrame.from_dict(EOD_DETAILED_RESULTS)
results.index.set_names(["Field", "Date"], inplace=True)
results.to_csv("results.csv")
results = read_moonshot_csv("results.csv")
results = results.reset_index()
results.loc[:, "Date"] = results.Date.dt.strftime("%Y-%m-%d")
results = results.set_index(["Field", "Date"])
self.assertDictEqual(results.to_dict(), EOD_DETAILED_RESULTS)
def test_intraday_detailed(self):
results = pd.DataFrame.from_dict(INTRADAY_DETAILED_RESULTS)
results.index.set_names(["Field", "Date", "Time"], inplace=True)
results.to_csv("results.csv")
results = read_moonshot_csv("results.csv")
results = results.reset_index()
results.loc[:, "Date"] = results.Date.dt.strftime("%Y-%m-%d")
results = results.set_index(["Field", "Date", "Time"])
results = results.where(results.notnull(), None)
self.assertDictEqual(results.to_dict(), INTRADAY_DETAILED_RESULTS)
def test_aggregate_intraday_to_daily(self):
results = pd.DataFrame.from_dict(INTRADAY_AGGREGATE_RESULTS)
results.index.set_names(["Field", "Date", "Time"], inplace=True)
results.to_csv("results.csv")
intraday_results = read_moonshot_csv("results.csv")
daily_results = intraday_to_daily(intraday_results)
daily_results = daily_results.reset_index()
daily_results.loc[:,
"Date"] = daily_results.Date.dt.strftime("%Y-%m-%d")
daily_results = daily_results.set_index(["Field", "Date"])
self.assertDictEqual(
daily_results.to_dict(),
{
'fx-revert': {
# max
('AbsExposure', '2018-12-18'): 0.0,
('AbsExposure', '2018-12-19'): 1.0,
# max
('AbsWeight', '2018-12-18'): 0.0,
('AbsWeight', '2018-12-19'): 1.0,
# last
('Benchmark', '2018-12-18'): 1.13606,
('Benchmark', '2018-12-19'): 1.142185,
# sum
('Commission', '2018-12-18'): 0.00022,
('Commission', '2018-12-19'): 0.0001,
# mean
('NetExposure', '2018-12-18'): 0.0,
('NetExposure', '2018-12-19'): 1.5,
# sum
('Return', '2018-12-18'): 0.0,
('Return', '2018-12-19'): 0.0005820400950750315,
# sum
('Slippage', '2018-12-18'): 4e-05,
('Slippage', '2018-12-19'): 5.000000000000001e-05,
# max
('TotalHoldings', '2018-12-18'): 0.0,
('TotalHoldings', '2018-12-19'): 5.0,
# sum
('Turnover', '2018-12-18'): 0.1,
('Turnover', '2018-12-19'): 0.30000000000000004
}
})
def from_moonshot_csv(filepath_or_buffer, **kwargs):
"""
Creates a full tear sheet from a moonshot backtest results CSV.
Additional kwargs are passed to :class:`pyfolio.create_full_tear_sheet`.
Parameters
----------
filepath_or_buffer : str or file-like object
filepath or file-like object of the CSV
Returns
-------
None
"""
results = read_moonshot_csv(filepath_or_buffer)
return from_moonshot(results, **kwargs)
def from_pnl_csv(cls, filepath_or_buffer,
trim_outliers=None,
riskfree=0,
compound=True,
rolling_sharpe_window=200):
"""
Creates a DailyPerformance instance from a PNL CSV.
Parameters
----------
filepath_or_buffer : str or file-like object
filepath or file-like object of the CSV
trim_outliers: int or float, optional
discard returns that are more than this many standard deviations from the mean
riskfree : float, optional
the riskfree rate (default 0)
compound : bool
True for compound/geometric returns, False for arithmetic returns (default True)
rolling_sharpe_window : int, optional
compute rolling Sharpe over this many periods (default 200)
Returns
-------
DailyPerformance
Examples
--------
Plot cumulative returns:
>>> perf = DailyPerformance.from_pnl_csv("pnl.csv")
>>> perf.cum_returns.plot()
"""
results = read_moonshot_csv(filepath_or_buffer)
return cls._from_pnl(
results, trim_outliers=trim_outliers,
riskfree=riskfree,
compound=compound,
rolling_sharpe_window=rolling_sharpe_window)
def test_detailed_intraday_to_daily(self):
results = pd.DataFrame.from_dict(INTRADAY_DETAILED_RESULTS)
results.index.set_names(["Field", "Date", "Time"], inplace=True)
results.to_csv("results.csv")
intraday_results = read_moonshot_csv("results.csv")
daily_results = intraday_to_daily(intraday_results)
daily_results = daily_results.reset_index()
daily_results.loc[:,
"Date"] = daily_results.Date.dt.strftime("%Y-%m-%d")
daily_results = daily_results.set_index(["Field", "Date"])
daily_results = daily_results.where(daily_results.notnull(), None)
self.assertDictEqual(
daily_results.to_dict(),
{
'EUR.USD(12087792)': { # max
('AbsExposure', '2018-12-18'): 0.0,
('AbsExposure', '2018-12-19'): 0.2,
# max
('AbsWeight', '2018-12-18'): 0.0,
('AbsWeight', '2018-12-19'): 0.2,
# last
('Benchmark', '2018-12-18'): 1.13606,
('Benchmark', '2018-12-19'): 1.142185,
# sum
('Commission', '2018-12-18'): 1e-05,
('Commission', '2018-12-19'): 2e-05,
# mean
('NetExposure', '2018-12-18'): 0.0,
('NetExposure', '2018-12-19'): 0.20000000000000004,
# sum
('Return', '2018-12-18'): 0.0,
('Return', '2018-12-19'): 0.00035730412741801225,
# sum
('Slippage', '2018-12-18'): 0.0,
('Slippage', '2018-12-19'): 4e-05,
# max
('TotalHoldings', '2018-12-18'): 0.0,
('TotalHoldings', '2018-12-19'): 2.0,
# sum
('Turnover', '2018-12-18'): 0.0,
('Turnover', '2018-12-19'): 0.30000000000000004,
# mean
('Weight', '2018-12-18'): 0.0,
('Weight', '2018-12-19'): 0.15
},
'GBP.USD(12087797)': { # max
('AbsExposure', '2018-12-18'): 0.0,
('AbsExposure', '2018-12-19'): 0.2,
# max
('AbsWeight', '2018-12-18'): 0.0,
('AbsWeight', '2018-12-19'): 0.2,
# last
('Benchmark', '2018-12-18'): None,
('Benchmark', '2018-12-19'): None,
# sum
('Commission', '2018-12-18'): 0.0,
('Commission', '2018-12-19'): 3.0000000000000004e-05,
# mean
('NetExposure', '2018-12-18'): 0.0,
('NetExposure', '2018-12-19'): 0.20000000000000004,
# sum
('Return', '2018-12-18'): 0.0,
('Return', '2018-12-19'): 0.0007114899464124138,
# sum
('Slippage', '2018-12-18'): 0.0,
('Slippage', '2018-12-19'): 3.0000000000000004e-05,
# max
('TotalHoldings', '2018-12-18'): 0.0,
('TotalHoldings', '2018-12-19'): 1.0,
# sum
('Turnover', '2018-12-18'): 0.0,
('Turnover', '2018-12-19'): 0.0,
# mean
('Weight', '2018-12-18'): 0.0,
('Weight', '2018-12-19'): 0.20000000000000004
}
})
def from_moonshot_csv(cls,
filepath_or_buffer,
start_date=None,
end_date=None,
trim_outliers=None,
how_to_aggregate=None,
riskfree=0,
compound=True,
rolling_sharpe_window=200):
"""
Creates a DailyPerformance instance from a Moonshot backtest results CSV.
Parameters
----------
filepath_or_buffer : str or file-like object
filepath or file-like object of the CSV
start_date : str (YYYY-MM-DD), optional
truncate at this start date (otherwise include entire date range)
end_date : str (YYYY-MM-DD), optional
truncate at this end date (otherwise include entire date range)
trim_outliers: int or float, optional
discard returns that are more than this many standard deviations from
the mean. Useful for dealing with data anomalies that cause large
spikes in plots.
how_to_aggregate : dict, optional
a dict of {fieldname: aggregation method} specifying how to aggregate
fields from intraday to daily. See the docstring for
`moonchart.utils.intraday_to_daily` for more details.
riskfree : float, optional
the riskfree rate (default 0)
compound : bool
True for compound/geometric returns, False for arithmetic returns (default True)
rolling_sharpe_window : int, optional
compute rolling Sharpe over this many periods (default 200)
Returns
-------
DailyPerformance
Examples
--------
Plot cumulative returns:
>>> perf = DailyPerformance.from_moonshot_csv("backtest_results.csv")
>>> perf.cum_returns.plot()
"""
try:
results = read_moonshot_csv(filepath_or_buffer)
except ValueError as e:
# "ValueError: 'Date' is not in list" might mean the user passed
# a paramscan csv by mistake
if "Date" not in repr(e):
raise
results = pd.read_csv(filepath_or_buffer)
if "StrategyOrDate" in results.columns:
raise MoonchartError(
"this is a parameter scan CSV, please use ParamscanTearsheet.from_moonshot_csv"
)
else:
raise
return cls._from_moonshot_or_pnl(
results,
start_date=start_date,
end_date=end_date,
trim_outliers=trim_outliers,
how_to_aggregate=how_to_aggregate,
riskfree=riskfree,
compound=compound,
rolling_sharpe_window=rolling_sharpe_window)
