Load management is now standard vocabulary. A player is rested because his load is high. A training week is periodised to control load. Injury risk is discussed in terms of load spikes.
Underneath all of that is an assumption that load is a measurable quantity with an agreed definition. It isn't, and the most influential attempt to formalise it ran into serious methodological criticism.
External and internal
The standard split is between external load — what the body did, measured by GPS and accelerometers — and internal load, meaning the physiological cost of doing it, measured by heart rate or by the player's own rating of perceived exertion.
Both matter and they can diverge sharply. Two players covering identical distances at identical speeds can experience completely different internal loads depending on fitness, sleep, illness, heat, and psychological stress.
Which of these should drive decisions? There's no consensus. Practitioners generally use both and weight them by judgement, which is reasonable and is not a science.
The acute-chronic workload ratio
For most of the last decade the dominant framework was the acute:chronic workload ratio. Compare a player's load over the past week to his rolling average over the past month. If the ratio is too high — a spike — injury risk rises. Keep it inside a "sweet spot," usually cited as roughly 0.8 to 1.3, and you're safe.
It's an appealing idea, it's intuitive, and it got adopted extremely widely across professional sport.
It has also been subject to substantial methodological criticism. Analysts pointed out problems with how the ratio is constructed — the acute period is contained within the chronic period, which creates mathematical artefacts. Several papers demonstrated that the reported relationships could arise from the statistical properties of the calculation rather than from any physiological effect. There were also issues with how the "sweet spot" thresholds were derived, and concerns about the original analyses.
Some of the original researchers have since acknowledged limitations, and the field has moved towards more careful approaches. But the concept remains embedded in commercial software and in practice, because it's been in the workflow for years and there isn't a clean replacement.
Why replacing it is hard
The underlying problem is that injuries are rare events with many causes and enormous individual variation.
A squad of thirty players might suffer fifteen or twenty significant soft-tissue injuries in a season. That's a tiny number of events to learn from. Any model trying to predict them from dozens of input variables is going to overfit badly, and validating it properly requires more data than any single club will ever have.
Multi-club datasets help and there have been collaborative efforts. But then you're combining data from different systems with different definitions, which reintroduces the standardisation problem.
There's also a confounding issue that's genuinely difficult. Players who are managed carefully get injured less. But they're managed carefully because staff judged them to be at risk, so the intervention and the risk are entangled. Disentangling that from observational data is close to impossible without randomising, which nobody is going to do with professional athletes.
What good practice looks like anyway
Despite all of that, load management as a practice does seem to help, and the reasons are less exotic than the models suggest.
Avoiding large sudden increases. The specific ratio maths may be flawed but the underlying principle — don't triple somebody's running volume in a week — is sound, well-supported, and doesn't require a model.
Maintaining a base. Players with higher chronic training loads tolerate spikes better. This is one of the more robust findings and it argues against excessive rest as much as against overload.
Monitoring individuals against themselves. Trend deviation for a specific player is far more useful than comparison to squad norms.
Listening to the player. Perceived exertion and subjective wellness scores, for all their softness, carry real signal. A player saying he feels heavy is data.
The commercial problem
One thing that muddies all of this: the tools are sold by companies whose product is the metric.
A vendor with a proprietary load score and a risk dashboard has a commercial interest in that score being seen as authoritative. The dashboards present outputs with a confidence the underlying evidence doesn't support — traffic-light indicators, risk percentages, clean thresholds.
Practitioners with a research background see through this. Coaches and directors, who often make the purchasing decision, sometimes don't, and a red indicator on a screen carries a lot of persuasive weight in a conversation about whether a player starts on Saturday.
Where it should go
The honest position is that load management is a reasonable heuristic practice with weak quantitative foundations, and it should be described that way rather than as an evidence-based system.
The most promising direction is probably individualised models — building a picture of a specific player over years rather than trying to find population-level rules. That's slow, it requires long employment relationships, and it doesn't scale into a product, which is exactly why it's underexplored.
In the meantime, an experienced physio who knows a player well is still, as far as I can tell, at least as good a predictor as any dashboard. That's not an argument against the technology. It's an argument for keeping the human in the loop and being honest about what the numbers are worth.