quantlab.portfolio.base.PortfolioContext#

class quantlab.portfolio.base.PortfolioContext(timestamp, predictions, tradable, current_weights, returns=None, factors=None, staleness=None, risk_exposures=None, risk_estimate=None)[source]#

Bases: object

Everything a portfolio construction rule may read at one bar.

Nothing in it reaches past the bar it describes, so a rule handed a context cannot look ahead.

Attributes:

timestamppd.Timestamp: The bar the rule decides on; the weights it returns fill at the next bar.
predictionsxr.Dataset: Every label’s prediction at the bar: one variable per label name, in the predictor’s label order, on symbol.
tradablexr.DataArray: Booleans on symbol: whether the symbol can be traded at the bar, judged from nothing later than the bar (the price dataset’s tradable_bars: by default, a real fill price at the bar). A rule treats a symbol without a finite prediction of the label it reads as not selectable too.
current_weightsxr.DataArray: The weights currently held, on symbol, valued at this bar’s valuation price: what the orders of the earlier rebalances left, rejected orders and delisting settlements included; 0.0 where nothing is held, all 0.0 before the first rebalance.
returnsxr.DataArray or None: The trailing window of one-bar returns ending at the bar, on (timestamp, symbol), of the rule’s declared lookback_bars length (no bars for a rule that needs none), read from its last declared history_bars raw valuation prices up to and including the bar. Each return is computed from the last valuation price known at its bar within those, so a halt shows as zero returns and then the whole gap on the bar the symbol trades again; NaN before a symbol’s first price in them and on that price’s bar. None in a context built by hand for a rule that reads none.
stalenessxr.DataArray or None: Bars since each symbol’s last real valuation price within the rule’s last declared history_bars bars, on symbol: 0 when it has one at the bar, NaN when it has none in those bars. None without prices.
factorsxr.Dataset or None: The values at the bar of the Factor panels the rule declares (InputDeclaration.factors; a benchmark beta, for example): one variable per factor name, on symbol, NaN where a symbol has none. None when the rule declares none.
risk_exposuresxr.Dataset or None: The exposures at the bar of the factor risk model the rule declares (InputDeclaration.risk_model), as FactorRiskModel.exposures gives them (read from the exposures factor’s store or computed, per the model’s exposure_data_strategy): its exposure_names and estimation-universe flag, on symbol, NaN where a symbol has none. None when the rule declares no risk model.
risk_estimatexr.Dataset or None: The estimate store’s row at the bar of the same factor risk model, its forecast made at the bar of the next bar’s risk (quantlab.risk.base.ESTIMATE_VARIABLES): factor_covariance on (factor_i, factor_j) and specific_risk on symbol, NaN where a symbol has none. With risk_exposures it is everything FactorRiskModel.forecast reads. None when the rule declares no risk model or the store has no row at the bar.

Examples

>>> import numpy as np, xarray as xr
>>> symbols = ["AAA", "BBB", "CCC"]
>>> context = PortfolioContext(
...     timestamp=pd.Timestamp("2024-01-02"),
...     predictions=xr.Dataset({"ret": ("symbol", [0.3, 0.1, 0.2])}, coords={"symbol": symbols}),
...     tradable=xr.DataArray([True, True, False], dims="symbol", coords={"symbol": symbols}),
...     current_weights=xr.DataArray([0.0, 0.4, 0.6], dims="symbol", coords={"symbol": symbols}),
... )
>>> context.symbols.tolist()
['AAA', 'BBB', 'CCC']
>>> context.locked.values
array([False, False,  True])

property locked#

Booleans on symbol: held and not tradable, so kept at the current weight.

Examples

>>> context.locked.values
array([False, False,  True])

property symbols#

The symbols of the bar, the axis the returned weights must be on.

Examples

>>> context.symbols.tolist()
['AAA', 'BBB', 'CCC']