Skip to contents

ACWR (development version)

ACWR 0.2.0

This is a bug-fix release that changes numerical results.

Results differ from 0.1.0 whenever sessions are not daily

In 0.1.0, RAC() and RAU() averaged the load per session inside each window, and EWMA() decayed once per session, ignoring the dates. Rest days were therefore not counted as zero load, and results were only correct when there was exactly one session every day. RAC() and RAU() also failed with “NA/NaN argument” when a week number was missing.

From 0.2.0, each athlete’s sessions are turned into a daily series from the first to the last date: sessions on the same day are summed and days without a session count as a load of 0. With exactly one session every day, EWMA results are unchanged and the RAC and RAU ratios are unchanged from day 28 on (except for rounding), but RAC/RAU acute and chronic loads are now 7 times larger (weekly sums instead of daily averages) and the ratios are NA during the first 27 days. Whenever sessions are not daily, all results differ.

For example, with the same weekly loads but 3 sessions per week for three weeks and 6 in the fourth, the coupled ratio (RAC) at the last session is now 1.45 (it was 0.93), the uncoupled ratio (RAU) 1.70 (was 1.31) and EWMA 1.20 (was 0.93).

Changes to the calculations

  • Acute and chronic loads of RAC() and RAU() are now weekly-equivalent sums over calendar days, not averages per session:
    • acute: sum of the daily load over the last 7 days (including today);
    • RAC chronic: sum over the last 28 days / 4 (coupled);
    • RAU chronic: sum over days 8 to 28 before today / 3 (uncoupled). This follows the definitions of coupled and uncoupled ratios in Windt & Gabbett (2019) doi:10.1136/bjsports-2017-098925 (most recent week divided by the average of the last 4 weeks, or of the 3 preceding weeks). Summing over calendar days counts rest days as zero load.
  • EWMA() applies the decay to every calendar day, including rest days with zero load, as in Williams et al. (2017) doi:10.1136/bjsports-2016-096589. Its acute and chronic loads are on the scale of a daily load.
  • RAC() and RAU() return NA until 28 days of history exist (acute loads, until 7 days). The new argument partial = TRUE uses the days available instead. ACWR() also gains partial.
  • A missing load (NA) makes every window containing its day NA, with one warning that gives the number of affected days. Because EWMA is recursive, every later day is NA for EWMA; impute missing loads before calling the functions.
  • The ratio is NA when the chronic load is 0.
  • Values are no longer rounded inside the functions.

Interface

  • All functions return one value per input row, in the input order. ACWR() now returns the rows in the original order of db (0.1.0 returned them grouped by athlete). It also no longer fails when no athlete has the ID 1.
  • EWMA() gains training_dates = NULL. If it is not given, sessions are assumed to be on consecutive days and a message is shown.
  • weeks (in ACWR(), RAC() and RAU()) and days (in ACWR()) are deprecated: they are kept for compatibility but no longer used in the calculation, and are now optional.
  • Inputs are validated with clear error messages: dates must be Date or POSIXct, loads must be numeric and non-negative, columns must exist in db. Unsorted dates are sorted internally.
  • The internal function training_blocks() has been removed.
  • plot_ACWR():
    • the x-axis is a continuous calendar, with one slot per day from the first to the last day, so rest days and weeks without training are visible (bar of height 0); axis labels are shown every 7 days;
    • several sessions on the same day are shown as one bar (summed load);
    • the ACWR line joins the sessions that have a value and is not drawn where the ACWR is NA;
    • day must hold day numbers (whole numbers) or dates (Date or POSIXct).

Other

  • New dataset training_load_rest: 3 athletes with rest days, a double session day and a week without training.
  • Examples now run.
  • README: new section on the limitations of the ACWR.
  • DESCRIPTION: fixed DOI and spelling of Gabbett.

ACWR 0.1.0

CRAN release: 2022-03-01

  • First CRAN release.