The Acute:Chronic Workload Ratio and Its Limits

Few metrics in applied sports science have travelled as far as fast as the acute:chronic workload ratio. Within a few years of its popularisation it appeared on dashboards across professional sport, complete with a “sweet spot” band and a danger zone above it. Within a few more years it had accumulated a serious methodological literature arguing that much of that structure was an artefact.

Our position is neither of the two comfortable ones. ACWR is not a fraud, and it is not a reliable decision rule. It is a ratio with specific statistical properties that you have to understand before the output means anything.

What the ratio is

Take a measure of workload — distance, high-speed distance, session-RPE load, whatever your department uses. Compute an acute value: the sum over a recent short window, classically seven days. Compute a chronic value: the average weekly sum over a longer window, classically twenty-eight days. Divide the first by the second. A value near 1.0 means this week resembles the recent norm; a high value means a spike; a low value means a drop.

The intuition is sound and matches what coaches already believe: sudden increases in exposure relative to what an athlete is accustomed to are associated with elevated injury risk. The problem is not the intuition. It is what happens when you turn the intuition into a ratio and then into a threshold.

The statistical objections, in plain terms

Mathematical coupling

In the classical rolling-average formulation, the acute window is contained inside the chronic window. The same seven days appear in the numerator and in a quarter of the denominator. Numerator and denominator are therefore not independent, and a ratio of coupled quantities has correlation structure that exists for arithmetic reasons rather than physiological ones. Analyses that then relate the ratio to injury outcomes can find associations that are partly artefacts of that coupling.

Spurious correlation with the denominator

Ratios in general behave badly. When you divide by a quantity that itself varies, the resulting variable inherits a relationship with the denominator that is easy to mistake for a relationship with the thing you care about. An athlete with a low chronic load produces volatile ratios from modest absolute changes; an athlete with a high chronic load produces stable ratios from large absolute changes. The ratio has flattened out exactly the information a coach needs.

Threshold arbitrariness

The widely quoted danger threshold around 1.5 emerged from particular datasets in particular sports with particular workload definitions. It is not a physical constant, it does not transfer cleanly across sports, and applying it to a workload measure other than the one it was derived from has no justification at all.

Window choice changes the answer

Seven and twenty-eight days are conventions, not findings. Change to five and twenty-one, or switch from rolling averages to exponentially weighted ones, and both the values and the athletes flagged change. A metric whose output depends materially on an arbitrary parameter choice needs that choice reported alongside it, which it almost never is.

What survives

Strip away the threshold and the ratio and what remains is genuinely useful: the observation that a large, rapid change in exposure relative to an athlete's recent accustomed load is worth noticing. That observation does not require division.

We prefer to present the components directly. Show the acute load. Show the chronic load. Show the week-on-week absolute change. Plot them on the same axis over a season. A coach looking at that picture sees everything the ratio was trying to compress, plus the magnitudes, plus the direction, without the coupling.

• Report the components, not just the quotient. A ratio of 1.6 built from a small chronic base is a different situation from one built on a large base.
• Prefer exponentially weighted windows to rolling averages if you use windows at all — they weight recent exposure more sensibly and reduce the sharp edge as sessions drop out of a rolling window.
• Never apply a population threshold to an individual without knowing that individual's own historical distribution.
• Record the workload definition and window parameters with every stored value, so that a comparison across seasons knows whether it is legitimate.
• Treat the ratio as one input among several, alongside readiness, availability history and what the staff know about the athlete's week.

How the platform handles it

The load and readiness module will compute ACWR — rolling and exponentially weighted, with configurable windows — because clubs ask for it and because refusing to display a metric a department already uses achieves nothing except making our product harder to adopt. What it does not do is put it on the default dashboard or attach a colour-coded threshold band to it.

When you open the ACWR view, the coupling is explained inline, the window parameters are shown, and the acute and chronic components are plotted alongside the ratio. If a value crosses a threshold you have configured, the interface says which athlete, which threshold, and what the underlying absolute change was — because the absolute change is the thing a coach can actually act on.

The broader lesson

ACWR is a good case study in how applied sports science absorbs a metric. A plausible idea gets an easy formula, the formula gets a threshold, the threshold gets a colour, and within a couple of seasons a colour on a dashboard is influencing selection decisions in a way the original evidence never supported. The failure is not in the idea; it is in the compression.

Our general rule follows from that: the more consequential the decision, the more of the underlying quantity should be visible. A number is a summary, and a summary is a set of decisions someone else made about what you do not need to see. In a selection or return-to-play conversation, that is precisely backwards.