sf_quant.research.beta_scale_ports#

sf_quant.research.beta_scale_ports(df: DataFrame, market_col: str = 'bmk_return', target_beta: float = 1.0, lookback: int = 60) → DataFrame#

Scale portfolio returns to a target beta relative to a market benchmark in place.

This function computes rolling beta estimates for each portfolio relative to a market benchmark, then scales the portfolios to achieve a specified target beta. Beta is calculated using rolling covariance and variance over a specified lookback window. Original return columns are replaced with their beta-scaled versions.

Parameters#

dfpl.DataFrame

Portfolio returns data containing:

  • Portfolio columns starting with p_ or named spread.

  • A market benchmark column specified by market_col.

market_colstr, optional

Name of the column containing market benchmark returns (default: “bmk_return”).

target_betafloat, optional

Target beta level for scaled portfolios (default: 1.0). A value of 1.0 means the portfolio moves in line with the market.

lookbackint, optional

Rolling window size in days for beta calculation (default: 60). Typically set to 2-3 months of trading days.

Returns#

pl.DataFrame

The DataFrame with beta-scaled return columns:

  • p_1, p_2, …, p_{num_bins} (float): Beta-scaled returns.

  • Other columns remain unchanged.

Notes#

  • Beta = rolling_cov(portfolio, market) / rolling_var(market)

  • Scaling factor = target_beta / estimated_beta

  • Only columns starting with p_ or named spread are scaled.

  • Original return values are replaced with scaled versions.

  • Early rows (< lookback window) will have null values.

Examples#

>>> import polars as pl
>>> import sf_quant.research as sfr
>>> import datetime as dt
>>> ports = pl.DataFrame({
...     'date': [dt.date(2024, 1, i) for i in range(1, 100)],
...     'p_1': [0.01] * 99,
...     'p_10': [0.02] * 99,
...     'spread': [0.01] * 99,
...     'bmk_return': [0.015] * 99
... })
>>> beta_scaled = sfr.beta_scale_ports(ports, target_beta=0.5, lookback=30)
>>> beta_scaled.columns
['date', 'p_1', 'p_10', 'spread', 'bmk_return']
On this page