Copy page

• Copy page
• View source

sigmaquant.performance.rolling.rolling_metric

sigmaquant.performance.rolling.rolling_metric(

returns,

metric,

*,

window,

min_periods=None,

**kwargs,

)

Compute a rolling metric on a sequence of returns.

The input is internally converted to a pandas Series. If the input does not provide a datetime index, a default integer index is used.

Parameters:

• returns – Sequence of periodic returns. If a pandas Series is provided, its index is preserved. Otherwise, a default integer index is assigned.

• metric –

  Metric function to apply. The function must have the signature:

  metric(returns: Iterable[float], **kwargs) -> float

• window – Size of the rolling window, expressed in number of observations.

• min_periods – Minimum number of observations required to compute the metric. If None, defaults to window.

• **kwargs – Additional keyword arguments passed directly to metric.

Returns:

Time series of rolling metric values.

Return type:

pandas.Series

Notes

• The rolling operation is index-based, not time-based.

• If the input does not carry a datetime index, temporal semantics (e.g. resampling, calendar alignment) are not available.

• NaN handling is delegated to the metric function.