Why a New Training Block Can Feel Worse Before It Feels Better

The first fortnight of a training block is where most people quit. You add a session, the erg gets heavier rather than lighter, the 2K splits slip a few seconds, and the logbook starts telling a story that has an obvious villain: the plan. The story is almost always wrong, and understanding why is more useful than any threshold number — including the one almost every club reaches for first.

Two clocks, not one

The explanation that has held up since 1980 is that training load drives two responses at once, and they run on different clocks ([1] Banister). Fitness accumulates slowly. Fatigue accumulates quickly and is cleared faster than fitness accumulates. Net performance is the difference between the two, which means raising the load can make you slower for one to four weeks while you are in fact getting fitter.

That is the whole model, and it is genuinely useful for one purpose: explaining a dip you have already decided to accept. It is not good at predicting a date when you will get fast, because the constants that describe how quickly your fitness builds are specific to the individual and the sport — and the constants are not stable between studies, or even between athletes.

Worth knowing where it came from, because the origin limits how far it travels. The original work quantified training impulse in swimmers, using a zone-weighted internal-load measure ([1] Banister). It was not fitted on rowers, and it was not fitted on session-RPE. Every rowing-specific number that follows from it is an extrapolation.

The rowing evidence for shape rather than timing is decent, though. In eighteen trained rowers over twelve weeks, a low-intensity distribution produced a gain in power at lactate threshold of 23.5 ± 12.2 W, against 5.1 ± 5.0 W for a mixed-intensity group doing comparable work ([9] Ingham). Both groups got fitter. One of them spent most of the block feeling like it was going backwards, and that is the trade the low-intensity group accepted.

The ratio that does not survive

If fitness and fatigue run on separate clocks, the obvious move is to compress them into one number: divide this week's load by the last four weeks' average and see whether you are ahead or behind. That index — the acute:chronic workload ratio — became the default answer, and it has the citation count to match.

It does not hold up.

Start with the paper that made it famous. Fifty-three elite rugby league players across two seasons found that the highest injury risk sat at very high ratios, and — this is the part usually dropped — that a high chronic workload made players more resistant to injury across the ordinary range of ratios, while the risk concentrated in sudden spikes ([2] Hulin). The conclusion is about the shape of a jump relative to a well-built base, not about crossing a line on a dial. Read as a threshold, it is not what the data said.

Then the methodology. Ratios built by dividing a recent week against an average that already contains that week are partly self-referential, so an apparent relationship to outcome can appear whether or not a real one exists. Add to that the fact that systematic reviews of the index disagree materially about how good their own underlying studies are ([8] MacMillan), and the review that settles it: there is no evidence supporting the use of this ratio in load-management systems or for recommendations aimed at reducing injury risk, and the statistical properties of the ratio make it an inaccurate metric that adds noise and creates artifacts ([3] Impellizzeri).

So the number is not a slightly imperfect threshold. The case for using it to change a training decision has not been made.

What survives is the part that was never in doubt: a large jump in load carries risk. That is checkable by eye. A rower coming back from two weeks off will trip any ratio above 1.5 through arithmetic alone, and treating that as a physiological emergency is precisely the error this literature warns against.

What the load number is actually measuring

Before blaming the number, be clear about what went into it. The original model weighted training by heart-rate zone, because heart rate was what the swimmers' laboratory could measure cheaply ([1] Banister). Most club logs do something close to that by multiplying a session-RPE score by its duration.

That method is not neutral. It counts what the athlete felt, which means it already contains the fatigue it is meant to be measuring — so a week where you are already tired registers as more load, which raises the ratio, which is then read as evidence that you did too much. The signal and the noise arrive together.

There is a second consequence, and it is the reason the four-week window exists at all. A ratio cannot distinguish "this week was hard" from "the last month was light." Those look identical on the dial but call for opposite responses: one means ease off, the other means you have built a base worth using. Asking which week you mean is not pedantry; it is the question the index cannot answer for you.

The taper, where the model is directly visible

Taper is the one situation where the two-clock model stops being a metaphor. Fatigue clears faster than fitness decays, so removing load should convert accumulated work into performance — and unlike most of the model, this has actually been meta-analysed.

Across twenty-seven studies, the best-evidenced taper is two weeks long, with training volume reduced 41–60%, and training intensity and frequency left alone ([10] Bosquet). Cut the volume. Keep the intensity. The athletes who improved did not taper by rowing easier.

That is a much narrower prescription than "rest a bit before the race", and it is the most directly actionable thing in this article.

Three things called "overreaching"

The word covers at least three states with different timelines, and conflating them is why people get the response wrong.

Functional overreaching is a planned, beneficial state that resolves within days of a taper, and endurance athletes have a reasonably identifiable marker set for it ([6] Roete). Non-functional overreaching is fatigue that does not resolve and should not have been planned. Overtraining syndrome is a clinical label applied after other causes are excluded — and the research base underneath the diagnostic language is much weaker than the language implies ([7] Weakley). A training log cannot confirm it, and a rower whose fatigue survives a genuine taper with normal food and normal sleep needs a clinician, not another protocol change.

There is one measurement that separates the first two in the laboratory, and it runs opposite to intuition. Thirty-one endurance athletes were taken through an overload block and then a two-week taper. The ten who developed functional overreaching showed heart-rate recovery getting faster by 8 ± 5 bpm, alongside falling lactate and catecholamines, all of which returned to baseline after the taper ([4] Aubry). Most rowers expect slower recovery to mean more fatigue. In functional overreaching, recovery speeds up.

Two caveats keep this honest. First, a recovery number is only meaningful above a minimum test intensity — take it after an easy row and it is not a fatigue signal ([5] Le Meur). Second, the same study is explicit that a faster recovery is not systematically associated with better performance ([4] Aubry). It marks the state; it does not promise the result.

The thing that looks identical and is not

Fatigue that will not shift, a poor block of sleep, and under-fuelling produce nearly the same logbook: bad sessions, low motivation, a resting heart rate sitting higher than usual. The overreaching literature keeps running into the third of these because energy availability turns up inside almost every overtraining study it reviews — the states overlap so heavily that they are hard to separate.

That matters practically because the interventions diverge. The response to a genuine load spike is to ease off. The response to under-fuelling is to eat more while keeping the load, and easing off just delays the problem. A rower who has been dieting hard through a training block and then reads a fatigue signal as an instruction to deload will make both things worse at once.

This is also the reason the clinical state gets treated as a last resort rather than a diagnosis. Overtraining syndrome is a label applied after other causes are excluded ([7] Weakley), and a training log cannot do that exclusion. Persistent fatigue that survives a real two-week taper, with normal sleep and adequate food, is the point where the next step is a clinician rather than another protocol change.

What to do with this article

Expect the dip. If a block feels bad for one to three weeks and the taper improves it, that is the model working, not failing. Stop reading splits as a verdict on the plan for the first month of a block.

Change one thing at a time. Load, intensity and frequency move together in most real weeks, which is exactly the confound that made the ratio look causal. If you want to know what a change did, hold the other two still for long enough to see it.

Taper for two weeks and cut volume by roughly half. Leave intensity and frequency alone ([10] Bosquet). This is the highest-confidence recommendation in the article and the easiest to apply.

Use a moving average rather than a single test. One bad 2K on a bad morning carries almost no information; the three-week trend carries most of it. The drop that matters is a sustained one, and the threshold for acting on it is a trend you can see without a spreadsheet.

Ask what else changed before blaming the load. Under-fuelling, poor sleep and a hard block all look alike from inside the logbook, and they need different responses.