def upload_portfolio_positions(name, df, nav=None):
if "model_provider_id" in df.columns:
id_type = "model_provider_id"
else:
id_type = "sedol"
if name not in [p.name for p in get_portfolios()]:
create_portfolio(name)
port_id = get_portfolio_id(name)
for date in list(df.date.unique()):
oper = OpOperation(schema.Mutation)
equities = [
schema.PositionSetEquityInput(
id=schema.PositionSetEquityIdInput(**{id_type: r[1][id_type]}),
economic_exposure=r[1].economic_exposure,
) for r in df.iterrows() if r[1].date == date
]
if nav is not None:
position_set = schema.PositionSetDateInput(date=date,
equities=equities,
equity=nav)
else:
position_set = schema.PositionSetDateInput(date=date,
equities=equities)
oper = OpOperation(schema.Mutation)
oper.upload_position_set_date(portfolio_id=port_id, data=position_set)
oper()
def df_to_position_set(df: pd.DataFrame):
dates = []
for dt in df.date.unique():
equities = [
schema.PositionSetEquityInput(
id=schema.PositionSetEquityIdInput(sedol=row.sedol),
economic_exposure=row.economic_exposure,
) for _, row in df[df.date == dt].iterrows()
]
dates += [schema.PositionSetDateInput(equities=equities, date=dt)]
return schema.PositionSetInput(dates=dates)
def get_market_impact_date(mi_date,
deltas,
nav,
denominator=None,
model_id=DEFAULT_MODEL_ID):
if denominator == None:
denominator = nav
equities = [
schema.PositionSetEquityInput(
id=schema.PositionSetEquityIdInput(model_provider_id=id),
economic_exposure=pct_equity * nav,
) for id, pct_equity in deltas.items()
]
ps_delta = schema.PositionSetDateInput(date=mi_date, equities=equities)
oper = OpOperation(schema.Query)
market_impact = (oper.model(id=model_id).simulation(
position_set=schema.PositionSetInput()).market_impact(
position_set_delta=ps_delta,
equity_id_format="MODEL_PROVIDER_ID",
scale_format="DEFAULT",
))
market_impact.cost()
market_impact.contributors().__fields__("id", "cost")
res = oper()
total_cost = res.model.simulation.market_impact.cost
total_cost_pct = total_cost / denominator
df_total = pd.DataFrame(
data=[[mi_date, total_cost, total_cost_pct]],
columns=["date", "total_cost", "total_cost_pct"],
)
contributors = [(mi_date, c.id, c.cost, c.cost / denominator)
for c in res.model.simulation.market_impact.contributors]
df_contrib = pd.DataFrame(
data=contributors,
columns=["date", "model_provider_id", "cost", "cost_pct"])
return df_total, df_contrib
def load_optimized_portfolio(df,
portfolio_name,
nav,
forecast_horizon,
objective,
model_id=DEFAULT_MODEL_ID):
if portfolio_name not in [p.name for p in op.get_portfolios()]:
op.create_portfolio(portfolio_name)
errors = []
for dt in df.date.unique():
rows = list(row for index, row in df[df.date == dt].iterrows())
oper = op.OpOperation(schema.Query)
long_len = len([r for r in rows if r.expected_return > 0])
if long_len > 0:
long_base = nav / long_len
short_len = len([r for r in rows if r.expected_return <= 0])
if short_len > 0:
short_base = -nav / short_len
equities = [
schema.PositionSetEquityInput(
id=schema.PositionSetEquityIdInput(sedol=row.sedol),
economic_exposure=long_base
if row.expected_return > 0 else short_base,
) for row in rows
]
position_set_dates = [
schema.PositionSetDateInput(equities=equities, date=dt)
]
position_set = schema.PositionSetInput(dates=position_set_dates)
forecast_equities = [
schema.ForecastEquityInput(
id=schema.PositionSetEquityIdInput(sedol=row.sedol),
expected_percent_return=row.expected_return,
) for row in rows
]
forecast = schema.ForecastInput(horizon=forecast_horizon,
equities=forecast_equities)
constants = schema.OptimizationConstantsInput(equity=nav)
optimization = oper.model(id=model_id).optimization(
position_set=position_set,
objective=[objective],
constants=constants,
constraints=constraints,
forecast=forecast,
)
opt_dates = optimization.positions().dates
opt_dates.date()
opt_dates.equities().id().sedol()
opt_dates.equities().__fields__("id", "economic_exposure")
results = None
try:
results = oper()
except op.GqlError as gql_err:
errors.append((dt, gql_err))
values = [[pos_date.date, equity.id.sedol, equity.economic_exposure]
for pos_date in results.model.optimization.positions.dates
for equity in pos_date.equities]
columns = ["date", "sedol", "economic_exposure"]
df_pos = pd.DataFrame(data=values, columns=columns)
upload_portfolio_positions(portfolio_name, df_pos, nav=nav)
return errors