Plot rolling return differences against benchmark

Plots rolling annualized tracking differences (CAGR-style or log-return) for selected funds against their benchmark series. The plot uses quantile-based y-limits and formats the y-axis in basis points.

Usage

plot_roll_diffs(
  data,
  n_days,
  funds,
  use_log = FALSE,
  gg_params = NULL,
  title_add = NULL,
  date_brk = NULL,
  qprob = c(0.005, 0.995),
  bmark_type = c("net", "gross")
)

Arguments

data

Input data frame containing a date column and one rolling-difference column per fund (specified by funds).

n_days

Window length (in days) used to compute rolling differences (used for labelling).

funds

Character vector of fund column names to include.

use_log

Logical; if TRUE, labels the plot as log-return differences. If FALSE, labels the plot as CAGR differences.

gg_params

Optional ggplot components to add to the plot.

title_add

Optional title suffix. Can be a single string or a named character vector specifying the title in multiple languages (e.g. c(en=..., bg=...)).

date_brk

Optional date-break specification for the x-axis (e.g. "3 months"). If NULL, it is chosen automatically based on the time span and available data.

qprob

Two-element numeric vector giving lower and upper quantiles used to set the baseline y-axis limits. Defaults to c(0.005, 0.995).

bmark_type

Benchmark type used in the title: "net" or "gross".

Value

A ggplot object.

Details

The function reshapes data to long format and produces a scatter plot coloured by fund. The y-axis limits are primarily determined from quantiles of the rolling differences (as specified by qprob), always including 0.

To avoid clipping recent extremes, the y-limits are expanded (if needed) to also include the full range observed in the most recent 30 days of data, even when those values fall outside the qprob quantiles.

The x-axis breaks are chosen as follows when date_brk is NULL: for spans of up to 3 years, breaks default to "3 months". For longer spans, breaks are anchored to calendar months (semiannual or quarterly depending on span) and are included only when the data range extends beyond the midpoint to the neighboring break.