def forecast_turnover(self, instrument_code, rule_variation_name):
"""
Get the annualised turnover for a forecast/rule combination
:param instrument_code: instrument to get values for
:type instrument_code: str
:param rule_variation_name: rule to get values for
:type rule_variation_name: str
:returns: float
"""
use_pooled_turnover = str2Bool(
self.parent.config.forecast_cost_estimates['use_pooled_turnover'])
if use_pooled_turnover:
instrument_code_list = self.has_same_rules_as_code(instrument_code)
else:
instrument_code_list = [instrument_code]
turnover_for_SR = self.forecast_turnover_for_list(
instrument_code_list, rule_variation_name)
return turnover_for_SR
def _use_fixed_weights(self):
if str2Bool(self.parent.config.use_forecast_scale_estimates):
fixed_flavour = True
else:
fixed_flavour = False
return fixed_flavour
def forecast_scalar(cs_forecasts, window=250000, min_periods=500, backfill=True):
"""
Work out the scaling factor for xcross such that T*x has an abs value of 10 (or whatever the average absolute forecast is)
:param cs_forecasts: forecasts, cross sectionally
:type cs_forecasts: pd.DataFrame TxN
:param span:
:type span: int
:param min_periods:
:returns: pd.DataFrame
"""
backfill = str2Bool(backfill) # in yaml will come in as text
# We don't allow this to be changed in config
target_abs_forecast = system_defaults['average_absolute_forecast']
# Take CS average first
# we do this before we get the final TS average otherwise get jumps in
# scalar when new markets introduced
if cs_forecasts.shape[1] == 1:
x = cs_forecasts.abs().iloc[:, 0]
else:
x = cs_forecasts.ffill().abs().median(axis=1)
# now the TS
avg_abs_value = x.rolling(window=window, min_periods=min_periods).mean()
scaling_factor = target_abs_forecast / avg_abs_value
if backfill:
scaling_factor = scaling_factor.fillna(method="bfill")
return scaling_factor
def __init__(self, log=logtoscreen("optimiser"), frequency="W", date_method="expanding",
rollyears=20, method="bootstrap", cleaning=True,
cost_multiplier=1.0, apply_cost_weight=True,
ann_target_SR=TARGET_ANN_SR, equalise_gross=False,
**passed_params):
cleaning=str2Bool(cleaning)
optimise_params=copy(passed_params)
ann_dict=dict(D=BUSINESS_DAYS_IN_YEAR, W=WEEKS_IN_YEAR, M=MONTHS_IN_YEAR, Y=1.0)
annualisation=ann_dict.get(frequency, 1.0)
period_target_SR=ann_target_SR/(annualisation**.5)
moments_estimator=momentsEstimator(optimise_params, annualisation, ann_target_SR)
optimiser=optimiserWithParams(method, optimise_params, moments_estimator)
setattr(self, "optimiser", optimiser)
setattr(self, "log", log)
setattr(self, "frequency", frequency)
setattr(self, "method", method)
setattr(self, "equalise_gross", equalise_gross)
setattr(self, "cost_multiplier", cost_multiplier)
setattr(self, "annualisation", annualisation)
setattr(self, "period_target_SR", period_target_SR)
setattr(self, "date_method", date_method)
setattr(self, "rollyears", rollyears)
setattr(self, "cleaning", cleaning)
setattr(self, "apply_cost_weight", apply_cost_weight)
def _use_fixed_weights(self):
if str2Bool(self.parent.config.use_forecast_scale_estimates):
fixed_flavour = True
else:
fixed_flavour = False
return fixed_flavour
def forecast_scalar(xcross, window=250000, min_periods=500, backfill=True):
"""
Work out the scaling factor for xcross such that T*x has an abs value of 10
:param x:
:type x: pd.DataFrame TxN
:param span:
:type span: int
:param min_periods:
:returns: pd.DataFrame
"""
backfill = str2Bool(backfill) # in yaml will come in as text
# We don't allow this to be changed in config
target_abs_forecast = system_defaults['average_absolute_forecast']
# Take CS average first
# we do this before we get the final TS average otherwise get jumps in
# scalar
if xcross.shape[1] == 1:
x = xcross.abs().iloc[:, 0]
else:
x = xcross.ffill().abs().median(axis=1)
# now the TS
avg_abs_value = x.rolling(window=window, min_periods=min_periods).mean()
scaling_factor = target_abs_forecast / avg_abs_value
if backfill:
scaling_factor = scaling_factor.fillna(method="bfill")
return scaling_factor
def get_SR_transaction_cost_for_instrument_forecast(
self, instrument_code: str, rule_variation_name: str) -> float:
"""
Get the SR cost for a forecast/rule combination
:param instrument_code: instrument to get values for
:type instrument_code: str
:param rule_variation_name: rule to get values for
:type rule_variation_name: str
:returns: float
KEY OUTPUT
"""
use_pooled_costs = str2Bool(
self.config.forecast_cost_estimates["use_pooled_costs"])
if use_pooled_costs:
SR_cost = self._get_SR_cost_for_rule_with_pooled_costs(
instrument_code, rule_variation_name)
else:
SR_cost = self._get_SR_cost_of_rule_for_individual_instrument(
instrument_code, rule_variation_name)
return SR_cost
def __init__(self, data, frequency="W",
date_method="expanding", rollyears=20,
dict_group=dict(), boring_offdiag=0.99, cleaning=True, **kwargs):
cleaning=str2Bool(cleaning)
group_dict=group_dict_from_natural(dict_group)
data=df_from_list(data)
column_names=list(data.columns)
data=data.resample(frequency, how="last")
fit_dates = generate_fitting_dates(data, date_method=date_method, rollyears=rollyears)
size=len(column_names)
corr_with_no_data=boring_corr_matrix(size, offdiag=boring_offdiag)
corr_list=[]
print(__file__ + ":" + str(inspect.getframeinfo(inspect.currentframe())[:3][1]) + ":" +"Correlation estimate")
for fit_period in fit_dates:
print(__file__ + ":" + str(inspect.getframeinfo(inspect.currentframe())[:3][1]) + ":" +"Estimating from %s to %s" % (fit_period.period_start, fit_period.period_end))
if fit_period.no_data:
corr_with_nan=boring_corr_matrix(size, offdiag=np.nan, diag=np.nan)
corrmat=corr_with_nan
else:
data_for_estimate=data[fit_period.fit_start:fit_period.fit_end]
corrmat=correlation_single_period(data_for_estimate, **kwargs)
if cleaning:
current_period_data=data[fit_period.fit_start:fit_period.fit_end]
must_haves=must_have_item(current_period_data)
corrmat=clean_correlation(corrmat, corr_with_no_data, must_haves)
corr_list.append(corrmat)
setattr(self, "corr_list", corr_list)
setattr(self, "columns", column_names)
setattr(self, "fit_dates", fit_dates)
def forecast_scalar(xcross, window=250000, min_periods=500, backfill=True):
"""
Work out the scaling factor for xcross such that T*x has an abs value of 10
:param x:
:type x: pd.DataFrame TxN
:param span:
:type span: int
:param min_periods:
:returns: pd.DataFrame
"""
backfill=str2Bool(backfill) ## in yaml will come in as text
##We don't allow this to be changed in config
target_abs_forecast = system_defaults['average_absolute_forecast']
## Take CS average first
## we do this before we get the final TS average otherwise get jumps in scalar
if xcross.shape[1]==1:
x=xcross.abs().iloc[:,0]
else:
x=xcross.ffill().abs().median(axis=1)
## now the TS
avg_abs_value=pd.rolling_mean(x, window=window, min_periods=min_periods)
scaling_factor=target_abs_forecast/avg_abs_value
if backfill:
scaling_factor=scaling_factor.fillna(method="bfill")
return scaling_factor
def forecast_turnover(self, instrument_code, rule_variation_name):
"""
Get the annualised turnover for a forecast/rule combination
:param instrument_code: instrument to get values for
:type instrument_code: str
:param rule_variation_name: rule to get values for
:type rule_variation_name: str
:returns: float
"""
use_pooled_turnover = str2Bool(
self.parent.config.forecast_cost_estimates['use_pooled_turnover'])
if use_pooled_turnover:
instrument_code_list = self.has_same_rules_as_code(instrument_code)
else:
instrument_code_list = [instrument_code]
turnover_for_SR = self.forecast_turnover_for_list(
instrument_code_list, rule_variation_name)
return turnover_for_SR
def __init__(self, log=logtoscreen("optimiser"), frequency="W", date_method="expanding",
rollyears=20, method="bootstrap", cleaning=True,
cost_multiplier=1.0, apply_cost_weight=True,
ann_target_SR=TARGET_ANN_SR, equalise_gross=False,
**passed_params):
cleaning=str2Bool(cleaning)
optimise_params=copy(passed_params)
ann_dict=dict(D=BUSINESS_DAYS_IN_YEAR, W=WEEKS_IN_YEAR, M=MONTHS_IN_YEAR, Y=1.0)
annualisation=ann_dict.get(frequency, 1.0)
period_target_SR=ann_target_SR/(annualisation**.5)
moments_estimator=momentsEstimator(optimise_params, annualisation, ann_target_SR)
optimiser=optimiserWithParams(method, optimise_params, moments_estimator)
setattr(self, "optimiser", optimiser)
setattr(self, "log", log)
setattr(self, "frequency", frequency)
setattr(self, "method", method)
setattr(self, "equalise_gross", equalise_gross)
setattr(self, "cost_multiplier", cost_multiplier)
setattr(self, "annualisation", annualisation)
setattr(self, "period_target_SR", period_target_SR)
setattr(self, "date_method", date_method)
setattr(self, "rollyears", rollyears)
setattr(self, "cleaning", cleaning)
setattr(self, "apply_cost_weight", apply_cost_weight)
def correlation_single_period(data_for_estimate,
using_exponent=True, min_periods=20, ew_lookback=250,
floor_at_zero=True):
"""
We generate a correlation from eithier a pd.DataFrame, or a list of them if we're pooling
It's important that forward filling, or index / ffill / diff has been done before we begin
also that we're on the right time frame, eg weekly if that's what we're doing
:param data_for_estimate: Data to get correlations from
:type data_for_estimate: pd.DataFrame
:param using_exponent: Should we use exponential weighting?
:type using_exponent: bool
:param ew_lookback: Lookback, in periods, for exp. weighting
:type ew_lookback: int
:param min_periods: Minimum periods before we get a correlation
:type min_periods: int
:param floor_at_zero: remove negative correlations before proceeding
:type floor_at_zero: bool or str
:returns: 2-dim square np.array
"""
# These may come from config as str
using_exponent = str2Bool(using_exponent)
if using_exponent:
# If we stack there will be duplicate dates
# So we massage the span so it's correct
# This assumes the index is at least daily and on same timestamp
# This is an artifact of how we prepare the data
dindex = data_for_estimate.index
dlenadj = float(len(dindex)) / len(set(list(dindex)))
# Usual use for IDM, FDM calculation when whole data set is used
corrmat = pd.ewmcorr(
data_for_estimate,
span=int(
ew_lookback *
dlenadj),
min_periods=min_periods)
# only want the final one
corrmat = corrmat.values[-1]
else:
# Use normal correlation
# Usual use for bootstrapping when only have sub sample
corrmat = data_for_estimate.corr(min_periods=min_periods)
corrmat = corrmat.values
if floor_at_zero:
corrmat[corrmat < 0] = 0.0
return corrmat
def calculation_of_raw_forecast_weights(self, instrument_code):
"""
returns the forecast weights for a given instrument code
Checks to see if there are pooled forecasts
"""
## Get some useful stuff from the config
## do we pool our estimation?
pooling_returns = str2Bool(self.parent.config.forecast_weight_estimate["pool_gross_returns"])
pooling_costs = str2Bool(self.parent.config.forecast_cost_estimates["use_pooled_costs"])
if (pooling_returns & pooling_costs):
return self.calculation_of_pooled_raw_forecast_weights(instrument_code)
else:
## could still be using pooled returns
return self.calculation_of_raw_forecast_weights_for_instrument(instrument_code)
def run_on_completion_only(self, process_name, method_name):
this_method_dict = self.get_method_configuration_for_process_name(
process_name, method_name)
run_on_completion_only = this_method_dict.get("run_on_completion_only",
False)
run_on_completion_only = str2Bool(run_on_completion_only)
return run_on_completion_only
def correlation_single_period(data_for_estimate,
using_exponent=True,
min_periods=20,
ew_lookback=250,
floor_at_zero=True):
"""
We generate a correlation from either a pd.DataFrame, or a list of them if we're pooling
It's important that forward filling, or index / ffill / diff has been done before we begin
also that we're on the right time frame, eg weekly if that's what we're doing
:param data_for_estimate: Data to get correlations from
:type data_for_estimate: pd.DataFrame
:param using_exponent: Should we use exponential weighting?
:type using_exponent: bool
:param ew_lookback: Lookback, in periods, for exp. weighting
:type ew_lookback: int
:param min_periods: Minimum periods before we get a correlation
:type min_periods: int
:param floor_at_zero: remove negative correlations before proceeding
:type floor_at_zero: bool or str
:returns: 2-dim square np.array
"""
# These may come from config as str
using_exponent = str2Bool(using_exponent)
if using_exponent:
# If we stack there will be duplicate dates
# So we massage the span so it's correct
# This assumes the index is at least daily and on same timestamp
# This is an artifact of how we prepare the data
dindex = data_for_estimate.index
dlenadj = float(len(dindex)) / len(set(list(dindex)))
# Usual use for IDM, FDM calculation when whole data set is used
corrmat = data_for_estimate.ewm(
span=int(ew_lookback * dlenadj),
min_periods=min_periods).corr(pairwise=True)
# only want the final one
corrmat = corrmat.values[-1]
else:
# Use normal correlation
# Usual use for bootstrapping when only have sub sample
corrmat = data_for_estimate.corr(min_periods=min_periods)
corrmat = corrmat.values
if floor_at_zero:
corrmat[corrmat < 0] = 0.0
return corrmat
def visible_on_lan() -> bool:
config = get_control_config()
visible = config.get_element_or_missing_data("dashboard_visible_on_lan")
if visible is missing_data:
return False
visible = str2Bool(visible)
return visible
def equalise_estimates_from_lists(
mean_list: list,
stdev_list: list,
equalise_SR: bool = True,
ann_target_SR: float = 0.5,
equalise_vols: bool = True,
) -> list:
equalise_vols = str2Bool(equalise_vols)
equalise_SR = str2Bool(equalise_SR)
if equalise_vols:
mean_list, stdev_list = vol_equaliser(
mean_list=mean_list, stdev_list=stdev_list
)
if equalise_SR:
mean_list = SR_equaliser(stdev_list, target_SR=ann_target_SR)
return mean_list, stdev_list
def calculation_of_raw_forecast_weights(self, instrument_code):
"""
returns the forecast weights for a given instrument code
Checks to see if there are pooled forecasts
"""
# Get some useful stuff from the config
# do we pool our estimation?
pooling_returns = str2Bool(
self.parent.config.forecast_weight_estimate["pool_gross_returns"])
pooling_costs = str2Bool(
self.parent.config.forecast_cost_estimates["use_pooled_costs"])
if (pooling_returns & pooling_costs):
return self.calculation_of_pooled_raw_forecast_weights(
instrument_code)
else:
# could still be using pooled returns
return self.calculation_of_raw_forecast_weights_for_instrument(
instrument_code)
def _get_instrument_code_depending_on_pooling_status(
instrument_code: str, forecast_scalar_config: dict) -> str:
# this determines whether we pool or not
pool_instruments = str2Bool(forecast_scalar_config.pop("pool_instruments"))
if pool_instruments:
# pooled, same for all instruments
instrument_code_to_pass = ALL_KEYNAME
else:
instrument_code_to_pass = copy(instrument_code)
return instrument_code_to_pass
def __init__(self, data, log=logtoscreen("optimiser"), frequency="W", date_method="expanding",
rollyears=20,
dict_group=dict(), boring_offdiag=0.99, cleaning=True, **kwargs):
cleaning=str2Bool(cleaning)
## grouping dictionary, convert to faster, algo friendly, form
group_dict=group_dict_from_natural(dict_group)
data=df_from_list(data)
column_names=list(data.columns)
data=data.resample(frequency, how="last")
### Generate time periods
fit_dates = generate_fitting_dates(data, date_method=date_method, rollyears=rollyears)
size=len(column_names)
corr_with_no_data=boring_corr_matrix(size, offdiag=boring_offdiag)
## create a list of correlation matrices
corr_list=[]
print(__file__ + ":" + str(inspect.getframeinfo(inspect.currentframe())[:3][1]) + ":" +"Correlation estimate")
## Now for each time period, estimate correlation
for fit_period in fit_dates:
print(__file__ + ":" + str(inspect.getframeinfo(inspect.currentframe())[:3][1]) + ":" +"Estimating from %s to %s" % (fit_period.period_start, fit_period.period_end))
if fit_period.no_data:
## no data to fit with
corr_with_nan=boring_corr_matrix(size, offdiag=np.nan, diag=np.nan)
corrmat=corr_with_nan
else:
data_for_estimate=data[fit_period.fit_start:fit_period.fit_end]
corrmat=correlation_single_period(data_for_estimate,
**kwargs)
if cleaning:
current_period_data=data[fit_period.fit_start:fit_period.fit_end]
must_haves=must_have_item(current_period_data)
# means we can use earlier correlations with sensible values
corrmat=clean_correlation(corrmat, corr_with_no_data, must_haves)
corr_list.append(corrmat)
setattr(self, "corr_list", corr_list)
setattr(self, "columns", column_names)
setattr(self, "fit_dates", fit_dates)
def forecast_turnover(self, instrument_code: str,
rule_variation_name: str) -> float:
use_pooled_turnover = str2Bool(
self.parent.config.forecast_cost_estimates["use_pooled_turnover"]
)
if use_pooled_turnover:
else:
instrument_code_list = [instrument_code]
return turnover_for_SR
def forecast_turnover(self, instrument_code: str,
rule_variation_name: str) -> float:
use_pooled_turnover = str2Bool(
self.config.forecast_cost_estimates["use_pooled_turnover"])
if use_pooled_turnover:
turnover = self._forecast_turnover_pooled(instrument_code,
rule_variation_name)
else:
turnover = self._forecast_turnover_for_individual_instrument(
instrument_code, rule_variation_name)
return turnover
def __init__(self,
data_for_correlation,
ew_lookback: int = 250,
min_periods: int = 20,
cleaning: bool = True,
floor_at_zero: bool = True,
length_adjustment: int = 1,
**_ignored_kwargs):
cleaning = str2Bool(cleaning)
floor_at_zero = str2Bool(floor_at_zero)
self._cleaning = cleaning
self._floor_at_zero = floor_at_zero
correlation_calculations = exponentialCorrelationResults(
data_for_correlation,
ew_lookback=ew_lookback,
min_periods=min_periods,
length_adjustment=length_adjustment)
self._correlation_calculations = correlation_calculations
self._data_for_correlation = data_for_correlation
def __init__(self,
data_as_df,
length_of_data=1,
ew_lookback=250,
boring_offdiag=0.99,
cleaning=True,
floor_at_zero=True,
**kwargs):
"""
Create an object to calculate correlations
We set up one of these with a set of data and parameters, and then call repeatedly
:param data_as_df: The dataframe of correlations
:param boring_offdiag: The off diagonal element to put into the matrix if data is absent
:param cleaning: Should we include fake values in the matrix so we don't need a warm up period?
:param floor_at_zero: Should we remove negative correlations?
:param ew_lookback: Lookback to use if exponential calculation used
:param length_of_data: Original length of data passed in (to correct for stacking of dataframe)
:return: np.array of correlation matrix
"""
self.cleaning = str2Bool(cleaning)
self.floor_at_zero = str2Bool(floor_at_zero)
# correct the lookback if we're jamming stuff together
self.ew_lookback_corrected = length_of_data * ew_lookback
size = data_as_df.shape[1]
self.corr_with_no_data = boring_corr_matrix(size, offdiag=np.nan)
self.corr_for_cleaning = boring_corr_matrix(size,
offdiag=boring_offdiag)
self.kwargs = kwargs
self.data_as_df = data_as_df
def __init__(self,
data_as_df,
length_of_data=1,
ew_lookback=250,
boring_offdiag=0.99,
cleaning=True,
floor_at_zero=True,
**kwargs):
"""
Create an object to calculate correlations
We set up one of these with a set of data and parameters, and then call repeatedly
:param data_as_df: The dataframe of correlations
:param boring_offdiag: The off diagonal element to put into the matrix if data is absent
:param cleaning: Should we include fake values in the matrix so we don't need a warm up period?
:param floor_at_zero: Should we remove negative correlations?
:param ew_lookback: Lookback to use if exponential calculation used
:param length_of_data: Original length of data passed in (to correct for stacking of dataframe)
:return: np.array of correlation matrix
"""
self.cleaning = str2Bool(cleaning)
self.floor_at_zero = str2Bool(floor_at_zero)
## correct the lookback if we're jamming stuff together
self.ew_lookback_corrected = length_of_data * ew_lookback
size = data_as_df.shape[1]
self.corr_with_no_data = boring_corr_matrix(size, offdiag=np.nan)
self.corr_for_cleaning = boring_corr_matrix(
size, offdiag=boring_offdiag)
self.kwargs = kwargs
self.data_as_df = data_as_df
def get_costs(self, instrument_code):
"""
Get the relevant kinds of cost for an instrument
:param instrument_code: instrument to value for
:type instrument_code: str
:returns: 2 tuple
"""
use_SR_costs = str2Bool(self.parent.config.use_SR_costs)
if use_SR_costs:
return (self.get_SR_cost(instrument_code), None)
else:
return (None, self.get_cash_costs(instrument_code))
def get_costs(self, instrument_code):
"""
Get the relevant kinds of cost for an instrument
:param instrument_code: instrument to value for
:type instrument_code: str
:returns: 2 tuple
"""
use_SR_costs = str2Bool(self.parent.config.use_SR_costs)
if use_SR_costs:
return (self.get_SR_cost(instrument_code), None)
else:
return (None, self.get_cash_costs(instrument_code))
def correlation_single_period(data_for_estimate,
using_exponent=True, min_periods=20, ew_lookback=250,
floor_at_zero=True):
using_exponent=str2Bool(using_exponent)
if using_exponent:
dindex=data_for_estimate.index
dlenadj=float(len(dindex))/len(set(list(dindex)))
corrmat=pd.ewmcorr(data_for_estimate, span=int(ew_lookback*dlenadj), min_periods=min_periods)
corrmat=corrmat.values[-1]
else:
corrmat=data_for_estimate.corr(min_periods=min_periods)
corrmat=corrmat.values
if floor_at_zero:
corrmat[corrmat<0]=0.0
return corrmat
def get_forecast_correlation_matrices_from_code_list(self, codes_to_use):
"""
Returns a correlationList object which contains a history of correlation matricies
:param codes_to_use:
:type str:
:returns: correlation_list object
>>> from systems.tests.testdata import get_test_object_futures_with_rules_and_capping_estimate
>>> from systems.basesystem import System
>>> (accounts, fcs, rules, rawdata, data, config)=get_test_object_futures_with_rules_and_capping_estimate()
>>> system=System([rawdata, rules, fcs, accounts, ForecastCombineEstimated()], data, config)
>>> ans=system.combForecast.get_forecast_correlation_matrices("EDOLLAR")
>>> ans.corr_list[-1]
array([[ 1. , 0.1168699 , 0.08038547],
[ 0.1168699 , 1. , 0.86907623],
[ 0.08038547, 0.86907623, 1. ]])
>>> print(ans.columns)
['carry', 'ewmac16', 'ewmac8']
"""
# Get some useful stuff from the config
corr_params = copy(self.parent.config.forecast_correlation_estimate)
# do we pool our estimation?
pooling = str2Bool(corr_params.pop("pool_instruments"))
# which function to use for calculation
corr_func = resolve_function(corr_params.pop("func"))
self.log.terse(
"Calculating forecast correlations over %s" %
", ".join(codes_to_use))
forecast_data = [
self.get_all_forecasts(
instr_code,
self.check_for_cheap_enough_rules(instr_code)) for instr_code in codes_to_use]
# if we're not pooling passes a list of one
forecast_data = [forecast_ts.ffill()
for forecast_ts in forecast_data]
return corr_func(forecast_data, log=self.log.setup(
call="correlation"), **corr_params)
def forecast_scalar(cs_forecasts,
window=250000,
min_periods=500,
backfill=True):
"""
Work out the scaling factor for xcross such that T*x has an abs value of 10 (or whatever the average absolute forecast is)
:param cs_forecasts: forecasts, cross sectionally
:type cs_forecasts: pd.DataFrame TxN
:param span:
:type span: int
:param min_periods:
:returns: pd.DataFrame
"""
backfill = str2Bool(backfill) # in yaml will come in as text
# We don't allow this to be changed in config
target_abs_forecast = get_default_config_key_value(
"average_absolute_forecast")
if target_abs_forecast is missing_data:
raise Exception(
"average_absolute_forecast not defined in system defaults file")
# Remove zeros/nans
copy_cs_forecasts = copy(cs_forecasts)
copy_cs_forecasts[copy_cs_forecasts == 0.0] = np.nan
# Take CS average first
# we do this before we get the final TS average otherwise get jumps in
# scalar when new markets introduced
if copy_cs_forecasts.shape[1] == 1:
x = copy_cs_forecasts.abs().iloc[:, 0]
else:
x = copy_cs_forecasts.ffill().abs().median(axis=1)
# now the TS
avg_abs_value = x.rolling(window=window, min_periods=min_periods).mean()
scaling_factor = target_abs_forecast / avg_abs_value
if backfill:
scaling_factor = scaling_factor.fillna(method="bfill")
return scaling_factor
def correlation_calculator(data_for_estimate,
using_exponent=True,
min_periods=20,
ew_lookback=250):
"""
We generate a correlation from a pd.DataFrame, which could have been stacked up
:param data_for_estimate: simData to get correlations from
:type data_for_estimate: pd.DataFrame
:param using_exponent: Should we use exponential weighting? If not every item is weighted equally
:type using_exponent: bool
:param ew_lookback: Lookback, in periods, for exp. weighting
:type ew_lookback: int
:param min_periods: Minimum periods before we get a correlation
:type min_periods: int
:returns: 2-dim square np.array
"""
# These may come from config as str
using_exponent = str2Bool(using_exponent)
if using_exponent:
# If we have stacked there will be duplicate dates
# So we massage the span so it's correct
# This assumes the index is at least daily and on same timestamp
# This is an artifact of how we prepare the data
# Usual use for IDM, FDM calculation when whole data set is used
corrmat = data_for_estimate.ewm(
span=ew_lookback, min_periods=min_periods).corr(pairwise=True)
number_of_items=data_for_estimate.shape[1]
# only want the final one
corrmat = corrmat.iloc[-number_of_items:,].values
else:
# Use normal correlation
# Usual use for bootstrapping when only have sub sample
corrmat = data_for_estimate.corr(min_periods=min_periods)
corrmat = corrmat.values
return corrmat
def correlation_calculator(data_for_estimate,
using_exponent=True,
min_periods=20,
ew_lookback=250):
"""
We generate a correlation from a pd.DataFrame, which could have been stacked up
:param data_for_estimate: simData to get correlations from
:type data_for_estimate: pd.DataFrame
:param using_exponent: Should we use exponential weighting? If not every item is weighted equally
:type using_exponent: bool
:param ew_lookback: Lookback, in periods, for exp. weighting
:type ew_lookback: int
:param min_periods: Minimum periods before we get a correlation
:type min_periods: int
:returns: 2-dim square np.array
"""
# These may come from config as str
using_exponent = str2Bool(using_exponent)
if using_exponent:
# If we have stacked there will be duplicate dates
# So we massage the span so it's correct
# This assumes the index is at least daily and on same timestamp
# This is an artifact of how we prepare the data
# Usual use for IDM, FDM calculation when whole data set is used
corrmat = data_for_estimate.ewm(
span=ew_lookback, min_periods=min_periods).corr(pairwise=True)
number_of_items = data_for_estimate.shape[1]
# only want the final one
corrmat = corrmat.iloc[-number_of_items:, ].values
else:
# Use normal correlation
# Usual use for bootstrapping when only have sub sample
corrmat = data_for_estimate.corr(min_periods=min_periods)
corrmat = corrmat.values
return corrmat
def get_forecast_correlation_matrices(
self, instrument_code: str) -> CorrelationList:
"""
Returns a correlationList object which contains a history of correlation matricies
:param instrument_code:
:type str:
:returns: correlation_list object
>>> from systems.tests.testdata import get_test_object_futures_with_rules_and_capping_estimate
>>> from systems.basesystem import System
>>> (accounts, fcs, rules, rawdata, data, config)=get_test_object_futures_with_rules_and_capping_estimate()
>>> system=System([rawdata, rules, fcs, accounts, ForecastCombineEstimated()], data, config)
>>> ans=system.combForecast.get_forecast_correlation_matrices("EDOLLAR")
>>> ans.corr_list[-1]
array([[ 1. , 0.1168699 , 0.08038547],
[ 0.1168699 , 1. , 0.86907623],
[ 0.08038547, 0.86907623, 1. ]])
>>> print(ans.columns)
['carry', 'ewmac16', 'ewmac8']
"""
# Get some useful stuff from the config
corr_params = copy(self.config.forecast_correlation_estimate)
# do we pool our estimation?
pooling = str2Bool(corr_params.pop("pool_instruments"))
if pooling:
# find set of instruments with same trading rules as I have
if self._use_estimated_weights():
codes_to_use = self.has_same_cheap_rules_as_code(
instrument_code)
else:
codes_to_use = self.has_same_rules_as_code(instrument_code)
else:
codes_to_use = [instrument_code]
correlation_list = (
self.get_forecast_correlation_matrices_from_instrument_code_list(
codes_to_use))
return correlation_list
def forecast_scalar(
cs_forecasts: pd.DataFrame,
target_abs_forecast: float = 10.0,
window: int = 250000, ## JUST A VERY LARGE NUMBER TO USE ALL DATA
min_periods=500, # MINIMUM PERIODS BEFORE WE ESTIMATE A SCALAR,
backfill=True ## BACKFILL OUR FIRST ESTIMATE, SLIGHTLY CHEATING, BUT...
) -> pd.Series:
"""
Work out the scaling factor for xcross such that T*x has an abs value of 10 (or whatever the average absolute forecast is)
:param cs_forecasts: forecasts, cross sectionally
:type cs_forecasts: pd.DataFrame TxN
:param span:
:type span: int
:param min_periods:
:returns: pd.DataFrame
"""
backfill = str2Bool(backfill) # in yaml will come in as text
# Remove zeros/nans
copy_cs_forecasts = copy(cs_forecasts)
copy_cs_forecasts[copy_cs_forecasts == 0.0] = np.nan
# Take CS average first
# we do this before we get the final TS average otherwise get jumps in
# scalar when new markets introduced
if copy_cs_forecasts.shape[1] == 1:
x = copy_cs_forecasts.abs().iloc[:, 0]
else:
x = copy_cs_forecasts.ffill().abs().median(axis=1)
# now the TS
avg_abs_value = x.rolling(window=window, min_periods=min_periods).mean()
scaling_factor = target_abs_forecast / avg_abs_value
if backfill:
scaling_factor = scaling_factor.fillna(method="bfill")
return scaling_factor
def get_forecast_correlation_matrices_from_instrument_code_list(
self, codes_to_use: list) -> CorrelationList:
"""
Returns a correlationList object which contains a history of correlation matricies
:param codes_to_use:
:type str:
:returns: correlation_list object
>>> from systems.tests.testdata import get_test_object_futures_with_rules_and_capping_estimate
>>> from systems.basesystem import System
>>> (accounts, fcs, rules, rawdata, data, config)=get_test_object_futures_with_rules_and_capping_estimate()
>>> system=System([rawdata, rules, fcs, accounts, ForecastCombineEstimated()], data, config)
>>> ans=system.combForecast.get_forecast_correlation_matrices("EDOLLAR")
>>> ans.corr_list[-1]
array([[ 1. , 0.1168699 , 0.08038547],
[ 0.1168699 , 1. , 0.86907623],
[ 0.08038547, 0.86907623, 1. ]])
>>> print(ans.columns)
['carry', 'ewmac16', 'ewmac8']
"""
# Get some useful stuff from the config
corr_params = copy(self.config.forecast_correlation_estimate)
# do we pool our estimation?
# not used here since we've looked at this already
_pooling_already_decided = str2Bool(
corr_params.pop("pool_instruments"))
# which function to use for calculation
corr_func = resolve_function(corr_params.pop("func"))
self.log.terse("Calculating forecast correlations over %s" %
", ".join(codes_to_use))
forecast_data = self.get_all_forecasts_for_a_list_of_instruments(
codes_to_use)
correlation_list = corr_func(forecast_data, **corr_params)
return correlation_list
def _system_init(self, system):
"""
When we add this stage object to a system, this code will be run
It will determine if we use an estimate or a fixed class of object
"""
if str2Bool(system.config.use_instrument_weight_estimates):
fixed_flavour=False
else:
fixed_flavour=True
if fixed_flavour:
self.__class__= PortfoliosFixed
self.__init__()
setattr(self, "parent", system)
else:
self.__class__= PortfoliosEstimated
self.__init__()
setattr(self, "parent", system)
def _system_init(self, system):
"""
When we add this stage object to a system, this code will be run
It will determine if we use an estimate or a fixed class of object
"""
if str2Bool(system.config.use_forecast_scale_estimates):
fixed_flavour=False
else:
fixed_flavour=True
if fixed_flavour:
self.__class__=ForecastScaleCapFixed
self.__init__()
setattr(self, "parent", system)
else:
self.__class__=ForecastScaleCapEstimated
self.__init__()
setattr(self, "parent", system)
def _system_init(self, system):
"""
When we add this stage object to a system, this code will be run
It will determine if we use an estimate or a fixed class of object
"""
if str2Bool(system.config.use_forecast_scale_estimates):
fixed_flavour = False
else:
fixed_flavour = True
if fixed_flavour:
self.__class__ = ForecastScaleCapFixed
self.__init__()
setattr(self, "parent", system)
else:
self.__class__ = ForecastScaleCapEstimated
self.__init__()
setattr(self, "parent", system)
def _system_init(self, system):
"""
When we add this stage object to a system, this code will be run
It will determine if we use an estimate or a fixed class of object
"""
if str2Bool(system.config.use_instrument_weight_estimates):
fixed_flavour = False
else:
fixed_flavour = True
if fixed_flavour:
self.__class__ = PortfoliosFixed
self.__init__()
setattr(self, "parent", system)
else:
self.__class__ = PortfoliosEstimated
self.__init__()
setattr(self, "parent", system)
def get_optimal_weights_with_fixed_contract_values(
self,
relevant_date: datetime.datetime = arg_not_supplied,
previous_weights: portfolioWeights = arg_not_supplied
) -> portfolioWeights:
covariance_matrix = self.get_covariance_matrix(
relevant_date=relevant_date)
risk_aversion = self.risk_aversion_coefficient()
expected_returns = self.get_implied_expected_returns(relevant_date)
per_contract_value = self.get_per_contract_value(relevant_date)
max_portfolio_weights = self.get_maximum_portfolio_weight_at_date(
relevant_date)
original_portfolio_weights = self.original_portfolio_weights_for_relevant_date(
relevant_date)
max_risk_as_variance = self.get_max_risk_as_variance()
costs = self.get_costs_per_contract_as_proportion_of_capital_all_instruments(
)
use_process_pool = str2Bool(
self.config.small_system['use_process_pool'])
## split up a bit??
optimal_weights = optimise_with_fixed_contract_values(
per_contract_value=per_contract_value,
expected_returns=expected_returns,
risk_aversion=risk_aversion,
covariance_matrix=covariance_matrix,
max_portfolio_weights=max_portfolio_weights,
original_portfolio_weights=original_portfolio_weights,
max_risk_as_variance=max_risk_as_variance,
costs=costs,
previous_weights=previous_weights,
use_process_pool=use_process_pool)
return optimal_weights
def calculation_of_raw_forecast_weights(self, instrument_code):
"""
Estimate the forecast weights for this instrument
We store this intermediate step to expose the calculation object
:param instrument_code:
:type str:
:returns: TxK pd.DataFrame containing weights, columns are trading rule variation names, T covers all
"""
def _calculation_of_raw_forecast_weights(system, NotUsed1, NotUsed2, this_stage,
codes_to_use, weighting_func, **weighting_params):
this_stage.log.terse("Calculating raw forecast weights over %s" % ", ".join(codes_to_use))
if hasattr(system, "accounts"):
pandl_forecasts=[this_stage.pandl_for_instrument_rules_unweighted(code)
for code in codes_to_use]
else:
error_msg="You need an accounts stage in the system to estimate forecast weights"
this_stage.log.critical(error_msg)
output=weighting_func(pandl_forecasts, log=self.log.setup(call="weighting"), **weighting_params)
return output
## Get some useful stuff from the config
weighting_params=copy(self.parent.config.forecast_weight_estimate)
## do we pool our estimation?
pooling=str2Bool(weighting_params.pop("pool_instruments"))
## which function to use for calculation
weighting_func=resolve_function(weighting_params.pop("func"))
if pooling:
## find set of instruments with same trading rules as I have
codes_to_use=self._has_same_rules_as_code(instrument_code)
instrument_code_ref=ALL_KEYNAME
## We
label='_'.join(codes_to_use)
else:
codes_to_use=[instrument_code]
label=instrument_code
instrument_code_ref=instrument_code
##
## label: how we identify this thing in the cache
## instrument_code_ref: eithier the instrument code, or 'all markets' if pooling
## _get_raw_forecast_weights: function to call if we don't find in cache
## self: this_system stage object
## codes_to_use: instrument codes to get data for
## weighting_func: function to call to calculate weights
## **weighting_params: parameters to pass to weighting function
##
raw_forecast_weights_calcs = self.parent.calc_or_cache_nested(
'calculation_of_raw_forecast_weights', instrument_code_ref, label,
_calculation_of_raw_forecast_weights,
self, codes_to_use, weighting_func, **weighting_params)
return raw_forecast_weights_calcs
def __init__(self, data, log=logtoscreen("optimiser"), frequency="W", date_method="expanding", rollyears=20,
dict_group=dict(), boring_offdiag=0.99, cleaning=True, **kwargs):
"""
We generate a correlation from eithier a pd.DataFrame, or a list of them if we're pooling
Its important that forward filling, or index / ffill / diff has been done before we begin
:param data: Data to get correlations from
:type data: pd.DataFrame or list if pooling
:param frequency: Downsampling frequency. Must be "D", "W" or bigger
:type frequency: str
:param date_method: Method to pass to generate_fitting_dates
:type date_method: str
:param roll_years: If date_method is "rolling", number of years in window
:type roll_years: int
:param dict_group: dictionary of groupings; used to replace missing values
:type dict_group: dict
:param boring_offdiag: Value used in creating 'boring' matrix, for when no data
:type boring_offdiag: float
:param **kwargs: passed to correlation_single_period
:returns: CorrelationList
"""
cleaning=str2Bool(cleaning)
## grouping dictionary, convert to faster, algo friendly, form
group_dict=group_dict_from_natural(dict_group)
data=df_from_list(data)
column_names=list(data.columns)
data=data.resample(frequency, how="last")
### Generate time periods
fit_dates = generate_fitting_dates(data, date_method=date_method, rollyears=rollyears)
size=len(column_names)
corr_with_no_data=boring_corr_matrix(size, offdiag=boring_offdiag)
## create a list of correlation matrices
corr_list=[]
log.terse("Correlation estimate")
## Now for each time period, estimate correlation
for fit_period in fit_dates:
log.msg("Fitting from %s to %s" % (fit_period.period_start, fit_period.period_end))
if fit_period.no_data:
## no data to fit with
corr_with_nan=boring_corr_matrix(size, offdiag=np.nan, diag=np.nan)
corrmat=corr_with_nan
else:
data_for_estimate=data[fit_period.fit_start:fit_period.fit_end]
corrmat=correlation_single_period(data_for_estimate,
**kwargs)
if cleaning:
# means we can use earlier correlations with sensible values
corrmat=clean_correlation(corrmat, corr_with_no_data, boring_offdiag)
corr_list.append(corrmat)
setattr(self, "corr_list", corr_list)
setattr(self, "columns", column_names)
setattr(self, "fit_dates", fit_dates)
def calculation_of_raw_forecast_weights_for_instrument(self, instrument_code):
"""
Does an optimisation for a single instrument
We do this if we can't do the special case of a pooled optimisation
Estimate the forecast weights for this instrument
We store this intermediate step to expose the calculation object
:param instrument_code:
:type str:
:returns: TxK pd.DataFrame containing weights, columns are trading rule variation names, T covers all
"""
def _calculation_of_raw_forecast_weights(system, instrument_code, this_stage,
codes_to_use, weighting_func, pool_costs, **weighting_params):
this_stage.log.terse("Calculating raw forecast weights for %s, over %s" % (instrument_code, ", ".join(codes_to_use)))
rule_list = self.apply_cost_weighting(instrument_code)
weight_func=weighting_func(log=self.log.setup(call="weighting"), **weighting_params)
if weight_func.need_data():
## returns a list of accountCurveGroups
pandl_forecasts=[this_stage.get_returns_for_optimisation(code)
for code in codes_to_use]
## the current curve is special
pandl_forecasts_this_code=this_stage.get_returns_for_optimisation(instrument_code)
## have to decode these
## returns two lists of pd.DataFrames
(pandl_forecasts_gross, pandl_forecasts_costs) = decompose_group_pandl(pandl_forecasts, pandl_forecasts_this_code, pool_costs=pool_costs)
## The weighting function requires two lists of pd.DataFrames, one gross, one for costs
weight_func.set_up_data(data_gross = pandl_forecasts_gross, data_costs = pandl_forecasts_costs)
else:
## in the case of equal weights, don't need data
forecasts = this_stage.get_all_forecasts(instrument_code, rule_list)
weight_func.set_up_data(weight_matrix=forecasts)
SR_cost_list = [this_stage.get_SR_cost_for_instrument_forecast(instrument_code, rule_variation_name)
for rule_variation_name in rule_list]
weight_func.optimise(ann_SR_costs=SR_cost_list)
return weight_func
## Get some useful stuff from the config
weighting_params=copy(self.parent.config.forecast_weight_estimate)
## do we pool our estimation?
pooling_returns = str2Bool(self.parent.config.forecast_weight_estimate["pool_gross_returns"])
pool_costs = str2Bool(self.parent.config.forecast_cost_estimates["use_pooled_costs"])
## which function to use for calculation
weighting_func=resolve_function(weighting_params.pop("func"))
if pooling_returns:
## find set of instruments with same trading rules as I have
codes_to_use=self.has_same_cheap_rules_as_code(instrument_code)
else:
codes_to_use=[instrument_code]
##
## _get_raw_forecast_weights: function to call if we don't find in cache
## self: this_system stage object
## codes_to_use: instrument codes to get data for
## weighting_func: function to call to calculate weights
## **weighting_params: parameters to pass to weighting function
##
raw_forecast_weights_calcs = self.parent.calc_or_cache(
'calculation_of_raw_forecast_weights', instrument_code,
_calculation_of_raw_forecast_weights,
self, codes_to_use, weighting_func, pool_costs, **weighting_params)
return raw_forecast_weights_calcs
pass
def get_forecast_correlation_matrices(self, instrument_code):
"""
Returns a correlationList object which contains a history of correlation matricies
:param instrument_code:
:type str:
:returns: correlation_list object
>>> from systems.tests.testdata import get_test_object_futures_with_rules_and_capping_estimate
>>> from systems.basesystem import System
>>> (accounts, fcs, rules, rawdata, data, config)=get_test_object_futures_with_rules_and_capping_estimate()
>>> system=System([rawdata, rules, fcs, accounts, ForecastCombineEstimated()], data, config)
>>> ans=system.combForecast.get_forecast_correlation_matrices("EDOLLAR")
>>> ans.corr_list[-1]
array([[ 1. , 0.1168699 , 0.08038547],
[ 0.1168699 , 1. , 0.86907623],
[ 0.08038547, 0.86907623, 1. ]])
>>> print(ans.columns)
['carry', 'ewmac16', 'ewmac8']
"""
def _get_forecast_correlation_matrices(system, NotUsed1, NotUsed2, this_stage,
codes_to_use, corr_func, **corr_params):
print(__file__ + ":" + str(inspect.getframeinfo(inspect.currentframe())[:3][1]) + ":" +"Calculating forecast correlations over %s" % ", ".join(codes_to_use))
forecast_data=[this_stage.get_all_forecasts(instr_code, this_stage.apply_cost_weighting(instr_code)) for instr_code in codes_to_use]
## if we're not pooling passes a list of one
forecast_data=[forecast_ts.ffill() for forecast_ts in forecast_data]
return corr_func(forecast_data, log=self.log.setup(call="correlation"), **corr_params)
## Get some useful stuff from the config
corr_params=copy(self.parent.config.forecast_correlation_estimate)
## do we pool our estimation?
pooling=str2Bool(corr_params.pop("pool_instruments"))
## which function to use for calculation
corr_func=resolve_function(corr_params.pop("func"))
if pooling:
## find set of instruments with same trading rules as I have
codes_to_use=self.has_same_cheap_rules_as_code(instrument_code)
instrument_code_ref=ALL_KEYNAME
## We
label='_'.join(codes_to_use)
else:
codes_to_use=[instrument_code]
label=instrument_code
instrument_code_ref=instrument_code
##
## label: how we identify this thing in the cache
## instrument_code_ref: eithier the instrument code, or 'all markets' if pooling
## _get_forecast_correlation_matrices: function to call if we don't find in cache
## self: this_system stage object
## codes_to_use: instrument codes
## func: function to call to calculate correlations
## **corr_params: parameters to pass to correlation function
##
forecast_corr_list = self.parent.calc_or_cache_nested(
'get_forecast_correlation_matrices', instrument_code_ref, label,
_get_forecast_correlation_matrices,
self, codes_to_use, corr_func, **corr_params)
return forecast_corr_list
def use_estimated_instrument_div_mult(self):
"""
It will determine if we use an estimate or a fixed class of object
"""
return str2Bool(self.parent.config.use_instrument_div_mult_estimates)
def __init__(self,
log=logtoscreen("optimiser"),
frequency="W",
date_method="expanding",
rollyears=20,
method="bootstrap",
cleaning=True,
cost_multiplier=1.0,
apply_cost_weight=True,
ann_target_SR=TARGET_ANN_SR,
equalise_gross=False,
**passed_params):
"""
Set up optimiser
:param frequency: Downsampling frequency. Must be "D", "W" or bigger
:type frequency: str
:param date_method: Method to pass to generate_fitting_dates
:type date_method: str
:param roll_years: If date_method is "rolling", number of years in window
:type roll_years: int
:param method: Method used for fitting, one of 'bootstrap', 'shrinkage', 'one_period'
:type method: str
:param equalise_gross: Should we equalise expected gross returns so that only costs affect weightings?
:type equalise_gross: bool
:param cost_multiplier: Multiply costs by this number
:type cost_multiplier: float
:param apply_cost_weight: Should we adjust our weightings to reflect costs?
:type apply_cost_weight: bool
:param *_estimate_params: dicts of **kwargs to pass to moments estimation, and optimisation functions
:returns: pd.DataFrame of weights
"""
## Because interaction of parameters is complex, display warnings
display_warnings(log, cost_multiplier, equalise_gross,
apply_cost_weight, method, **passed_params)
cleaning = str2Bool(cleaning)
optimise_params = copy(passed_params)
## annualisation
ann_dict = dict(D=BUSINESS_DAYS_IN_YEAR,
W=WEEKS_IN_YEAR,
M=MONTHS_IN_YEAR,
Y=1.0)
annualisation = ann_dict.get(frequency, 1.0)
period_target_SR = ann_target_SR / (annualisation**.5)
## A moments estimator works out the mean, vol, correlation
## Also stores annualisation factor and target SR (used for shrinkage and equalising)
moments_estimator = momentsEstimator(optimise_params, annualisation,
ann_target_SR)
## The optimiser instance will do the optimation once we have the appropriate data
optimiser = optimiserWithParams(method, optimise_params,
moments_estimator)
setattr(self, "optimiser", optimiser)
setattr(self, "log", log)
setattr(self, "frequency", frequency)
setattr(self, "method", method)
setattr(self, "equalise_gross", equalise_gross)
setattr(self, "cost_multiplier", cost_multiplier)
setattr(self, "annualisation", annualisation)
setattr(self, "period_target_SR", period_target_SR)
setattr(self, "date_method", date_method)
setattr(self, "rollyears", rollyears)
setattr(self, "cleaning", cleaning)
setattr(self, "apply_cost_weight", apply_cost_weight)
def __init__(self, log=logtoscreen("optimiser"), frequency="W", date_method="expanding",
rollyears=20, method="bootstrap", cleaning=True,
cost_multiplier=1.0, apply_cost_weight=True,
ann_target_SR=TARGET_ANN_SR, equalise_gross=False,
**passed_params):
"""
Set up optimiser
:param frequency: Downsampling frequency. Must be "D", "W" or bigger
:type frequency: str
:param date_method: Method to pass to generate_fitting_dates
:type date_method: str
:param roll_years: If date_method is "rolling", number of years in window
:type roll_years: int
:param method: Method used for fitting, one of 'bootstrap', 'shrinkage', 'one_period'
:type method: str
:param equalise_gross: Should we equalise expected gross returns so that only costs affect weightings?
:type equalise_gross: bool
:param cost_multiplier: Multiply costs by this number
:type cost_multiplier: float
:param apply_cost_weight: Should we adjust our weightings to reflect costs?
:type apply_cost_weight: bool
:param *_estimate_params: dicts of **kwargs to pass to moments estimation, and optimisation functions
:returns: pd.DataFrame of weights
"""
## Because interaction of parameters is complex, display warnings
display_warnings(log, cost_multiplier, equalise_gross, apply_cost_weight, method, **passed_params)
cleaning=str2Bool(cleaning)
optimise_params=copy(passed_params)
## annualisation
ann_dict=dict(D=BUSINESS_DAYS_IN_YEAR, W=WEEKS_IN_YEAR, M=MONTHS_IN_YEAR, Y=1.0)
annualisation=ann_dict.get(frequency, 1.0)
period_target_SR=ann_target_SR/(annualisation**.5)
## A moments estimator works out the mean, vol, correlation
## Also stores annualisation factor and target SR (used for shrinkage and equalising)
moments_estimator=momentsEstimator(optimise_params, annualisation, ann_target_SR)
## The optimiser instance will do the optimation once we have the appropriate data
optimiser=optimiserWithParams(method, optimise_params, moments_estimator)
setattr(self, "optimiser", optimiser)
setattr(self, "log", log)
setattr(self, "frequency", frequency)
setattr(self, "method", method)
setattr(self, "equalise_gross", equalise_gross)
setattr(self, "cost_multiplier", cost_multiplier)
setattr(self, "annualisation", annualisation)
setattr(self, "period_target_SR", period_target_SR)
setattr(self, "date_method", date_method)
setattr(self, "rollyears", rollyears)
setattr(self, "cleaning", cleaning)
setattr(self, "apply_cost_weight", apply_cost_weight)
def calculation_of_pooled_raw_forecast_weights(self, instrument_code):
"""
Estimate the forecast weights for this instrument
We store this intermediate step to expose the calculation object
:param instrument_code:
:type str:
:returns: TxK pd.DataFrame containing weights, columns are trading rule variation names, T covers all
"""
def _calculation_of_pooled_raw_forecast_weights(system, instrument_code_ref, this_stage,
codes_to_use, weighting_func, **weighting_params):
print(__file__ + ":" + str(inspect.getframeinfo(inspect.currentframe())[:3][1]) + ":" +"Calculating pooled raw forecast weights over instruments: %s" % instrument_code_ref)
rule_list = self.apply_cost_weighting(instrument_code)
weight_func=weighting_func(log=self.log.setup(call="weighting"), **weighting_params)
if weight_func.need_data():
## returns a list of accountCurveGroups
## cost pooling will already have been applied
pandl_forecasts=[this_stage.get_returns_for_optimisation(code)
for code in codes_to_use]
## have to decode these
## returns two lists of pd.DataFrames
(pandl_forecasts_gross, pandl_forecasts_costs) = decompose_group_pandl(pandl_forecasts, pool_costs=True)
## The weighting function requires two lists of pd.DataFrames, one gross, one for costs
weight_func.set_up_data(data_gross = pandl_forecasts_gross, data_costs = pandl_forecasts_costs)
else:
## in the case of equal weights, don't need data
forecasts = this_stage.get_all_forecasts(instrument_code, rule_list)
weight_func.set_up_data(weight_matrix=forecasts)
SR_cost_list = [this_stage.get_SR_cost_for_instrument_forecast(instrument_code, rule_variation_name)
for rule_variation_name in rule_list]
weight_func.optimise(ann_SR_costs=SR_cost_list)
return weight_func
## Get some useful stuff from the config
weighting_params=copy(self.parent.config.forecast_weight_estimate)
## do we pool our estimation?
pooling_returns = str2Bool(weighting_params.pop("pool_gross_returns"))
pooling_costs = self.parent.config.forecast_cost_estimates['use_pooled_costs']
assert pooling_returns and pooling_costs
## which function to use for calculation
weighting_func=resolve_function(weighting_params.pop("func"))
codes_to_use=self.has_same_cheap_rules_as_code(instrument_code)
instrument_code_ref ="_".join(codes_to_use) ## ensures we don't repeat optimisation
##
## _get_raw_forecast_weights: function to call if we don't find in cache
## self: this_system stage object
## codes_to_use: instrument codes to get data for
## weighting_func: function to call to calculate weights
## **weighting_params: parameters to pass to weighting function
##
raw_forecast_weights_calcs = self.parent.calc_or_cache(
'calculation_of_raw_forecast_weights', instrument_code_ref,
_calculation_of_pooled_raw_forecast_weights,
self, codes_to_use, weighting_func, **weighting_params)
return raw_forecast_weights_calcs
def __init__(self, data, log=logtoscreen("optimiser"), frequency="W", date_method="expanding",
rollyears=20, fit_method="bootstrap", cleaning=True,
**passed_params):
"""
Optimise weights over some returns data
:param data: Returns data
:type data: pd.DataFrame or list if pooling
:param frequency: Downsampling frequency. Must be "D", "W" or bigger
:type frequency: str
:param date_method: Method to pass to generate_fitting_dates
:type date_method: str
:param roll_years: If date_method is "rolling", number of years in window
:type roll_years: int
:param fit_method: Method used for fitting, one of 'bootstrap', 'shrinkage', 'one_period'
:type fit_method: str
:param cleaning: Should we clean correlations so can use incomplete data?
:type cleaning: bool
:param *_estimate_params: dicts of **kwargs to pass to moments estimation, and optimisation functions
:returns: pd.DataFrame of weights
"""
cleaning=str2Bool(cleaning)
optimise_params=copy(passed_params)
## A moments estimator works out the mean, vol, correlation
moments_estimator=momentsEstimator(optimise_params)
## The optimiser instance will do the optimation once we have the appropriate data
optimiser=optimiserWithParams(optimise_params, moments_estimator)
## annualisation
ann_dict=dict(D=BUSINESS_DAYS_IN_YEAR, W=WEEKS_IN_YEAR, M=MONTHS_IN_YEAR, Y=1.0)
annualisation=ann_dict.get(frequency, 1.0)
## de-pool pooled data
data=df_from_list(data)
## resample, indexing before and differencing after (returns, remember)
data=data.cumsum().resample(frequency, how="last").diff()
## account for change in frequency
data=data*annualisation
fit_dates = generate_fitting_dates(data, date_method=date_method, rollyears=rollyears)
setattr(self, "fit_dates", fit_dates)
## Now for each time period, estimate weights
## create a list of weight vectors
weight_list=[]
## create a class object for each period
opt_results=[]
log.terse("Optimising...")
for fit_period in fit_dates:
log.msg("Optimising for data from %s to %s" % (str(fit_period.period_start), str(fit_period.period_end)))
## Do the optimisation for one period, using a particular optimiser instance
results_this_period=optSinglePeriod(self, data, fit_period, optimiser, cleaning)
opt_results.append(results_this_period)
weights=results_this_period.weights
## We adjust dates slightly to ensure no overlaps
dindex=[fit_period.period_start+datetime.timedelta(days=1),
fit_period.period_end-datetime.timedelta(days=1)]
## create a double row to delineate start and end of test period
weight_row=pd.DataFrame([weights]*2, index=dindex, columns=data.columns)
weight_list.append(weight_row)
## Stack everything up
weight_df=pd.concat(weight_list, axis=0)
setattr(self, "results", opt_results)
setattr(self, "weights", weight_df)
def get_forecast_scalar(self, instrument_code, rule_variation_name):
"""
Get the scalar to apply to raw forecasts
If not cached, these are estimated from past forecasts
If configuration variable pool_forecasts_for_scalar is "True", then we do this across instruments.
:param instrument_code:
:type str:
:param rule_variation_name:
:type str: name of the trading rule variation
:returns: float
>>> from systems.tests.testdata import get_test_object_futures_with_rules
>>> from systems.basesystem import System
>>> (rules, rawdata, data, config)=get_test_object_futures_with_rules()
>>> system1=System([rawdata, rules, ForecastScaleCapEstimated()], data, config)
>>>
>>> ## From default
>>> system1.forecastScaleCap.get_forecast_scalar("EDOLLAR", "ewmac8").tail(3)
scale_factor
2015-12-09 5.849888
2015-12-10 5.850474
2015-12-11 5.851091
>>> system1.forecastScaleCap.get_capped_forecast("EDOLLAR", "ewmac8").tail(3)
ewmac8
2015-12-09 0.645585
2015-12-10 -0.210377
2015-12-11 0.961821
>>>
>>> ## From config
>>> scale_config=dict(pool_instruments=False)
>>> config.forecast_scalar_estimate=scale_config
>>> system3=System([rawdata, rules, ForecastScaleCapEstimated()], data, config)
>>> system3.forecastScaleCap.get_forecast_scalar("EDOLLAR", "ewmac8").tail(3)
scale_factor
2015-12-09 5.652174
2015-12-10 5.652833
2015-12-11 5.653444
>>>
"""
def _get_forecast_scalar(
system, Not_Used, rule_variation_name, this_stage, instrument_list,
scalar_function, forecast_scalar_config):
"""
instrument_list contains multiple things, pools everything across all instruments
"""
print(__file__ + ":" + str(inspect.getframeinfo(inspect.currentframe())[:3][1]) + ":" +"Getting forecast scalar for %s over %s" % (rule_variation_name, ", ".join(instrument_list)))
## Get forecasts for each instrument
forecast_list=[
this_stage.get_raw_forecast(instrument_code, rule_variation_name)
for instrument_code in instrument_list]
cs_forecasts=pd.concat(forecast_list, axis=1)
scaling_factor=scalar_function(cs_forecasts, **forecast_scalar_config)
return scaling_factor
## Get some useful stuff from the config
forecast_scalar_config=copy(self.parent.config.forecast_scalar_estimate)
# The config contains 'func' and some other arguments
# we turn func which could be a string into a function, and then
# call it with the other ags
scalarfunction = resolve_function(forecast_scalar_config.pop('func'))
## this determines whether we pool or not
pool_instruments=str2Bool(forecast_scalar_config.pop("pool_instruments"))
if pool_instruments:
## pooled, same for all instruments
instrument_code_key=ALL_KEYNAME
instrument_list=self.parent.get_instrument_list()
else:
## not pooled
instrument_code_key=instrument_code
instrument_list=[instrument_code]
forecast_scalar = self.parent.calc_or_cache_nested(
"get_forecast_scalar", instrument_code_key, rule_variation_name,
_get_forecast_scalar, self, instrument_list, scalarfunction, forecast_scalar_config)
return forecast_scalar
def calculation_of_raw_forecast_weights(self, instrument_code):
"""
Estimate the forecast weights for this instrument
We store this intermediate step to expose the calculation object
:param instrument_code:
:type str:
:returns: TxK pd.DataFrame containing weights, columns are trading rule variation names, T covers all
"""
def _calculation_of_raw_forecast_weights(system, instrument_code, this_stage,
codes_to_use, weighting_func, pool_costs=False, **weighting_params):
this_stage.log.terse("Calculating raw forecast weights over %s" % ", ".join(codes_to_use))
if hasattr(system, "accounts"):
## returns a list of accountCurveGroups
pandl_forecasts=[this_stage.pandl_for_instrument_rules_unweighted(code)
for code in codes_to_use]
## the current curve is special
pandl_forecasts_this_code=this_stage.pandl_for_instrument_rules_unweighted(instrument_code)
## have to decode these
## returns two lists of pd.DataFrames
(pandl_forecasts_gross, pandl_forecasts_costs) = decompose_group_pandl(pandl_forecasts, pandl_forecasts_this_code, pool_costs=pool_costs)
else:
error_msg="You need an accounts stage in the system to estimate forecast weights"
this_stage.log.critical(error_msg)
## The weighting function requires two lists of pd.DataFrames, one gross, one for costs
output=weighting_func(pandl_forecasts_gross, pandl_forecasts_costs,
log=self.log.setup(call="weighting"), **weighting_params)
return output
## Get some useful stuff from the config
weighting_params=copy(self.parent.config.forecast_weight_estimate)
## do we pool our estimation?
pooling=str2Bool(weighting_params.pop("pool_instruments"))
## which function to use for calculation
weighting_func=resolve_function(weighting_params.pop("func"))
if pooling:
## find set of instruments with same trading rules as I have
codes_to_use=self._has_same_rules_as_code(instrument_code)
else:
codes_to_use=[instrument_code]
##
## _get_raw_forecast_weights: function to call if we don't find in cache
## self: this_system stage object
## codes_to_use: instrument codes to get data for
## weighting_func: function to call to calculate weights
## **weighting_params: parameters to pass to weighting function
##
raw_forecast_weights_calcs = self.parent.calc_or_cache(
'calculation_of_raw_forecast_weights', instrument_code,
_calculation_of_raw_forecast_weights,
self, codes_to_use, weighting_func, **weighting_params)
return raw_forecast_weights_calcs
def _get_forecast_scalar_estimated(self, instrument_code,
rule_variation_name):
"""
Get the scalar to apply to raw forecasts
If not cached, these are estimated from past forecasts
If configuration variable pool_forecasts_for_scalar is "True", then we
do this across instruments.
:param instrument_code:
:type str:
:param rule_variation_name:
:type str: name of the trading rule variation
:returns: float
>>> from systems.tests.testdata import get_test_object_futures_with_rules
>>> from systems.basesystem import System
>>> (rules, rawdata, data, config)=get_test_object_futures_with_rules()
>>> system1=System([rawdata, rules, ForecastScaleCapEstimated()], data, config)
>>>
>>> ## From default
>>> system1.forecastScaleCap.get_forecast_scalar("EDOLLAR", "ewmac8").tail(3)
scale_factor
2015-12-09 5.849888
2015-12-10 5.850474
2015-12-11 5.851091
>>> system1.forecastScaleCap.get_capped_forecast("EDOLLAR", "ewmac8").tail(3)
ewmac8
2015-12-09 0.645585
2015-12-10 -0.210377
2015-12-11 0.961821
>>>
>>> ## From config
>>> scale_config=dict(pool_instruments=False)
>>> config.forecast_scalar_estimate=scale_config
>>> system3=System([rawdata, rules, ForecastScaleCapEstimated()], data, config)
>>> system3.forecastScaleCap.get_forecast_scalar("EDOLLAR", "ewmac8").tail(3)
scale_factor
2015-12-09 5.652174
2015-12-10 5.652833
2015-12-11 5.653444
>>>
"""
# Get some useful stuff from the config
forecast_scalar_config = copy(
self.parent.config.forecast_scalar_estimate)
# this determines whether we pool or not
pool_instruments = str2Bool(
forecast_scalar_config.pop("pool_instruments"))
if pool_instruments:
# pooled, same for all instruments
instrument_code_to_pass = ALL_KEYNAME
else:
instrument_code_to_pass = copy(instrument_code)
scaling_factor = self._get_forecast_scalar_estimated_from_instrument_list(
instrument_code_to_pass, rule_variation_name,
forecast_scalar_config)
forecast = self.get_raw_forecast(instrument_code, rule_variation_name)
scaling_factor = scaling_factor.reindex(forecast.index, method="ffill")
return scaling_factor
def _use_estimated_weights(self):
return str2Bool(self.parent.config.use_forecast_scale_estimates)
def __init__(self,
data,
identifier=None,
parent = None,
frequency="W",
date_method="expanding",
rollyears=20,
method="bootstrap",
cleaning=True,
cost_multiplier=1.0,
apply_cost_weight=True,
ann_target_SR=TARGET_ANN_SR,
equalise_gross=False,
pool_gross_returns=False,
use_pooled_costs=False,
use_pooled_turnover=None, # not used
**passed_params):
"""
Set up optimiser
:param data: A dict of account curve dataframes, if not pooled data will only have one entry
:type data: dict
:param identifier: A dictionary key which refers to the element in data which we're optimising for
:type identifier: str
:param parent: The parent object, probably a system stage, which we get the log attribute from
:type parent: systemStage
:param frequency: Downsampling frequency. Must be "D", "W" or bigger
:type frequency: str
:param date_method: Method to pass to generate_fitting_dates
:type date_method: str
:param roll_years: If date_method is "rolling", number of years in window
:type roll_years: int
:param method: Method used for fitting, one of 'bootstrap', 'shrinkage', 'one_period'
:type method: str
:param cleaning: Do we clean weights so we don't need a warmup period?
:type cleaning: bool
:param equalise_gross: Should we equalise expected gross returns so that only costs affect weightings?
:type equalise_gross: bool
:param pool_gross_returns: Should we pool gross returns together?
:type pool_gross_returns: bool
:param use_pooled_costs: Should we pool costs together?
:type use_pooled_costs: bool
:param cost_multiplier: Multiply costs by this number
:type cost_multiplier: float
:param apply_cost_weight: Should we adjust our weightings to reflect costs?
:type apply_cost_weight: bool
:param *_estimate_params: dicts of **kwargs to pass to moments estimation, and optimisation functions
:returns: pd.DataFrame of weights
"""
if parent is None:
log = logtoscreen("optimiser")
else:
log = parent.log
setattr(self, "log", log)
# Because interaction of parameters is complex, display warnings
self.display_warnings(cost_multiplier, equalise_gross,
apply_cost_weight, method, **passed_params)
cleaning = str2Bool(cleaning)
optimise_params = copy(passed_params)
# annualisation
ANN_DICT = dict(
D=BUSINESS_DAYS_IN_YEAR, W=WEEKS_IN_YEAR, M=MONTHS_IN_YEAR, Y=1.0)
annualisation = ANN_DICT.get(frequency, 1.0)
self.set_up_data(data, frequency=frequency, equalise_gross=equalise_gross,
cost_multiplier=cost_multiplier, annualisation=annualisation,
ann_target_SR=TARGET_ANN_SR,
use_pooled_costs=use_pooled_costs,
pool_gross_returns=pool_gross_returns,
identifier=identifier)
# A moments estimator works out the mean, vol, correlation
# Also stores annualisation factor and target SR (used for shrinkage
# and equalising)
moments_estimator = momentsEstimator(optimise_params, annualisation,
ann_target_SR)
# The optimiser instance will do the optimation once we have the
# appropriate data
optimiser = optimiserWithParams(method, optimise_params,
moments_estimator)
setattr(self, "optimiser", optimiser)
setattr(self, "frequency", frequency)
setattr(self, "method", method)
setattr(self, "equalise_gross", equalise_gross)
setattr(self, "cost_multiplier", cost_multiplier)
setattr(self, "annualisation", annualisation)
setattr(self, "date_method", date_method)
setattr(self, "rollyears", rollyears)
setattr(self, "cleaning", cleaning)
setattr(self, "apply_cost_weight", apply_cost_weight)
def use_estimated_instrument_div_mult(self):
"""
It will determine if we use an estimate or a fixed class of object
"""
return str2Bool(self.parent.config.use_instrument_div_mult_estimates)
def __init__(self, data_gross, data_costs, log=logtoscreen("optimiser"), frequency="W", date_method="expanding",
rollyears=20, fit_method="bootstrap", cleaning=True, equalise_gross=False,
cost_multiplier=1.0, apply_cost_weight=True, ceiling_cost_SR=0.13,
ann_target_SR=TARGET_ANN_SR,
**passed_params):
"""
Optimise weights over some returns data
:param data_gross: Returns data for gross returns
:type data_gross: pd.DataFrame or list if pooling
:param data_net: Returns data for costs
:type data_net: pd.DataFrame or list if pooling
:param frequency: Downsampling frequency. Must be "D", "W" or bigger
:type frequency: str
:param date_method: Method to pass to generate_fitting_dates
:type date_method: str
:param roll_years: If date_method is "rolling", number of years in window
:type roll_years: int
:param fit_method: Method used for fitting, one of 'bootstrap', 'shrinkage', 'one_period'
:type fit_method: str
:param equalise_gross: Should we equalise expected gross returns so that only costs affect weightings?
:type equalise_gross: bool
:param cost_multiplier: Multiply costs by this number
:type cost_multiplier: float
:param apply_cost_weight: Should we adjust our weightings to reflect costs?
:type apply_cost_weight: bool
:param ceiling_cost_SR: What is the maximum SR cost beyond which I don't allocate to an asset. Set to 999 to avoid using.
:type ceiling_cost_SR: float
:param *_estimate_params: dicts of **kwargs to pass to moments estimation, and optimisation functions
:returns: pd.DataFrame of weights
"""
## Because interaction of parameters is complex, display warnings
display_warnings(log, cost_multiplier, equalise_gross, apply_cost_weight, **passed_params)
cleaning=str2Bool(cleaning)
optimise_params=copy(passed_params)
## annualisation
ann_dict=dict(D=BUSINESS_DAYS_IN_YEAR, W=WEEKS_IN_YEAR, M=MONTHS_IN_YEAR, Y=1.0)
annualisation=ann_dict.get(frequency, 1.0)
period_target_SR=ann_target_SR/(annualisation**.5)
ceiling_cost_SR_period=ceiling_cost_SR/(annualisation**.5)
## A moments estimator works out the mean, vol, correlation
## Also stores annualisation factor and target SR (used for shrinkage and equalising)
moments_estimator=momentsEstimator(optimise_params, annualisation, ann_target_SR)
## The optimiser instance will do the optimation once we have the appropriate data
optimiser=optimiserWithParams(optimise_params, moments_estimator)
## resample, indexing before and differencing after (returns, remember)
data_gross = [data_item.cumsum().resample(frequency, how="last").diff() for
data_item in data_gross]
data_costs = [data_item.cumsum().resample(frequency, how="last").diff() for
data_item in data_costs]
## stack de-pool pooled data
data_gross=df_from_list(data_gross)
data_costs=df_from_list(data_costs)
## net gross and costs
if equalise_gross:
log.terse("Setting all gross returns to be identical - optimisation driven only by costs")
if cost_multiplier!=1.0:
log.terse("Using cost multiplier on optimisation of %.2f" % cost_multiplier)
data = work_out_net(data_gross, data_costs, annualisation=annualisation,
equalise_gross=equalise_gross, cost_multiplier=cost_multiplier,
ceiling_cost_ann_SR=ceiling_cost_SR,
period_target_SR=period_target_SR)
fit_dates = generate_fitting_dates(data, date_method=date_method, rollyears=rollyears)
setattr(self, "fit_dates", fit_dates)
## Now for each time period, estimate weights
## create a list of weight vectors
weight_list=[]
## create a class object for each period
opt_results=[]
log.terse("Optimising...")
for fit_period in fit_dates:
log.msg("Optimising for data from %s to %s" % (str(fit_period.period_start), str(fit_period.period_end)))
## Do the optimisation for one period, using a particular optimiser instance
results_this_period=optSinglePeriod(self, data, fit_period, optimiser, cleaning)
opt_results.append(results_this_period)
weights=results_this_period.weights
## We adjust dates slightly to ensure no overlaps
dindex=[fit_period.period_start+datetime.timedelta(days=1),
fit_period.period_end-datetime.timedelta(days=1)]
## create a double row to delineate start and end of test period
weight_row=pd.DataFrame([weights]*2, index=dindex, columns=data.columns)
weight_list.append(weight_row)
## Stack everything up
raw_weight_df=pd.concat(weight_list, axis=0)
if apply_cost_weight:
log.terse("Applying cost weighting to optimisation results")
weight_df = apply_cost_weighting(raw_weight_df, data_gross, data_costs, annualisation)
else:
weight_df =raw_weight_df
setattr(self, "results", opt_results)
setattr(self, "weights", weight_df)
setattr(self, "raw_weights", raw_weight_df)