Load Governors: What CTL, ATL, TSB and ACWR Actually Tell a Rower

The context

A load governor is a number that stands in for how hard your training has been, so that a plan can be adjusted before something breaks. It is not a target you chase and not a score you win. It is a dial, like the thermometer on the wall of a boathouse: useful because it tells you that something has changed, and misleading the moment you treat it as a diagnosis.

You want one if you train several times a week and your week-to-week plan actually varies. A rower who rows twice a week, or who does the same three sessions every week, gets almost nothing from a load governor that a training diary does not already tell them. If you are on the erg four, five or six times a week, or building towards a head race, then what you did last week starts to matter more than any single session, and a number that remembers is worth having. Training load is the rare quantity in a plan that you genuinely cannot recall: you can remember how a session felt, but not how hard the last three weeks were.

Your coach app shows four of these numbers: CTL, ATL, TSB and ACWR. They are worth about ten minutes of your attention, and only if you know which of them are measurements and which of them are arithmetic built on top of a measurement.

The core idea

All four quantities are built from one input — a single session's load — and then smoothed, differenced and divided. Take them in the order a reader meets them.

CTL: the background you are used to

Chronic training load is the accumulated load of recent weeks, smoothed so that yesterday counts for more than last month. "Chronic" here means long-standing, not illness. It is the background your body has adapted to: the amount of work you could keep doing without getting worse.

On a dashboard, CTL is the quiet line. It rises when you train more, drifts down when you ease off, and it never moves fast. The question it answers is not "how hard was today" but "what has this training block become".

ATL: the last seven days

Acute training load is the same measurement read through a much shorter window — the last week of sessions, weighted toward the most recent days. If your week included a race, a double, or a run of hard sessions, ATL is where it shows up. If you had three easy days and one long row, ATL barely moves, because a long steady row is one moderate session rather than a spike.

TSB: the gap between them

Training stress balance is CTL minus ATL, and it is the number that most resembles a feeling. Positive means this week is lighter than the background your body is built on: fresher than your norm. Negative means this week is heavier than your body has absorbed, and you are carrying the deficit. A deeply negative TSB is the one that shows up as a warm-up that feels wrong, a race that goes out too hard, or a bad night's sleep after a session that looked fine on paper.

ACWR: the ratio, and the open question

The acute:chronic workload ratio divides ATL by CTL. It is the "are we doing more than we are used to" dial in a single number: near 1.0 means this week looks like the weeks your body has adapted to.

The band that most people know is the one it is supposed to sit in. Ratios of roughly 0.8 to 1.3 have been described as a training "sweet spot", and 1.5 and above as a "danger zone" ([6] Gabbett 2016). Read that claim carefully, because the same review that offers it warns against applying it to individuals: the band was drawn from three team sports — cricket, Australian football and rugby league — and the author says it should be used with caution until more data is available ([6] Gabbett 2016).

The empirical work underneath it is narrower than the reputation. In elite rugby league, 53 players were followed across two seasons, and a very high ratio was where injury risk concentrated: at 2.11 or above, injury risk in the current week was 16.7% and in the week after 11.8%, while players carrying a high background load were comparatively resistant across ratios of 0.85 to 1.35 and then vulnerable again around 1.5 ([7] Hulin 2016). One cohort, one sport, and a step change rather than a smooth curve.

So treat the band as a hypothesis with evidence behind it, not as a settled limit. The literature has not settled the question. A multilevel meta-analysis of team-sport studies found a higher ratio associated with more injuries, but only weakly and with enormous variation between studies — an effect size of 0.35 with heterogeneity near 96% — and concluded that the evidence does not support the ratio as a stand-alone predictor, so it belongs inside a multi-marker, individualised system ([11] Ding 2026). Meanwhile the methodological case against it has not gone away: the ratio adds noise, creates statistical artefacts, and there is no evidence supporting its use in training-load-management systems ([8] Impellizzeri 2020).

A scale you can feel

The numbers are easier to trust when you can see what feeds them. The most common way to score a session is the session rating of perceived exertion: after training you rate the whole session from 1 to 10, and the session's load is that rating multiplied by the minutes you spent on the erg ([4] Halson 2014).

So a long steady row of 60 minutes at an effort of 4 out of 10 scores 240. A hard interval session of eight 500 m pieces, 40 minutes in total, rated 8 out of 10, scores 320. A rest day scores nothing at all.

The how-to

The first thing to get right is that a load number is a proxy, not a measurement of anything physical. It stands in for "how hard was that", and which proxy your software uses changes the answer. The choice is usually made by the app rather than by you.

There are two families of proxy. Internal load is what the athlete experienced: heart rate, blood lactate, and the rating of perceived exertion. External load is the work done, measured independently of how the athlete felt about it, such as distance covered or power on the erg ([3] Impellizzeri 2019). Session-RPE sits in the internal family. A power-based training stress score sits in the external one ([2] Sanders 2017).

The two families disagree, and the disagreement is informative. In the study that introduced session-RPE, the score derived from the athlete's own rating and the score derived from heart rate tracked each other closely across steady and interval work, but the perceived score was consistently the larger of the two ([1] Foster 2001). Same session, same effort, different absolute number, because one method asks what you felt and the other asks what your heart did.

There is a further wrinkle: the best load metric is the one that knows about you. In fifteen competitive road cyclists over a ten-week preseason, every method tested tracked changes in aerobic fitness to some degree, and the strongest relationships belonged to an individualised heart-rate method and the power-based training stress score. The paper's own conclusion is that methods accounting for individual physiology work best ([2] Sanders 2017). A number calibrated on somebody else's physiology will flatter or punish you for the wrong reasons.

What the number cannot tell you is the more important half of the job. It does not know how you slept, whether you are ill, whether the room was thirty degrees, whether you were motivated or merely compliant, or whether you ate. It does not know that the last two sessions were technically poor. The monitoring literature is blunt about this: there is still no single definitive marker of training load, and very few of the candidates have strong evidence behind them ([4] Halson 2014). In rowing specifically, the recommendation has long been to watch stress and recovery at the same time rather than stress alone ([5] Mäestu 2005).

One practical consequence follows. Because a load number is derived, it can move because the pipeline changed rather than the athlete: a different monitor, a different question on the app, a session that never got logged. Any of those shifts your ATL without you having done anything different, so the number is worth reading alongside the training diary, never instead of it.

The trade-offs

Each of the four numbers has a specific weakness, and the weakness is different for each.

ACWR is fragile as a statistic. The ratio divides one number by another, and the two are the same underlying data smoothed over different windows. A mathematical coupling between numerator and denominator produces correlations that look like findings and are not, and the ratio's own properties — it spikes when the chronic value happens to be low, for instance — mean it adds noise rather than removing it ([8] Impellizzeri 2020). Reviews of how the ratio has been used in injury research have catalogued the problems with the computation itself: the initial-load artefact of exponential weighting, the choice of time windows, discretisation of continuous data, sparse weeks, and the fact that injury itself changes how an athlete trains ([9] Wang 2020). A clinical commentary on the same literature put it more bluntly: load measures cannot tell you whether a change in load is raising or lowering injury risk, so practitioners should go on using their knowledge of the athlete ([10] Impellizzeri 2020).

CTL and ATL are not two measurements. They are one series of sessions read through two different filters. That is not wrong — the filters are what let you compare this week with the block you have been in — but it means the four numbers are not four independent pieces of evidence. When CTL and ATL move together, you have learned almost nothing new.

A load says nothing about quality. A session in which you dragged the handle badly for an hour and a session in which you moved well for the same length and the same effort score the same. Nothing in session-RPE, in heart-rate zones or in a power trace tells your coach that the three hardest pieces of the week were rushed and that the technique fell apart in the last repetition. Rowing-specific monitoring has long argued that stress has to be read alongside recovery and alongside the performance itself, precisely because the load number is blind to both ([5] Mäestu 2005).

A number can move because the measurement changed. This is the failure mode that catches people who trust a dashboard more than they trust their diary. The app changes what it asks you, the monitor stops pairing, a session goes unlogged, and the load series changes shape. The governor then appears to have opinions about a week you did not train differently. Any change in the pipeline should be treated as a break in the series rather than a change in the athlete.

What to do tomorrow

Pick one week. For every session, write down two numbers: how many minutes you were on the erg, and how hard the whole session felt out of ten. Multiply them. That is the raw input behind every number your coach app shows, and you can now check the app against the week you actually had.

Then do one thing with it. Compare the week you just logged with the week before it and ask a single question: did the change come from more sessions, from longer sessions, or from harder ones? Those three changes need three different responses, and no single load number can tell them apart for you.

That is the whole job. A governor is a memory aid that lets you and your coach talk about training volume in the same units, and it is worth exactly as much as the conversation you have with it.