What MyNextRow's Coach Actually Reads From Your Logbook

You log a session. The monitor writes down the distance, the time, the average split, the stroke rate, the heart rate and the damper setting. You can also type a line into the note field. Then a few days later a coach talks about your fitness and your fatigue, and it is not obvious where any of that came from.

This article is the plumbing: what MyNextRow's coach actually reads out of your logbook, in what order, and — the more useful half — what it never reads at all.

The short version: it reads your sessions, your goal, and what you tell it in chat. Everything else in this article is detail on those three.

The context

Start with what a logbook is. The Concept2 Logbook API returns a fixed set of fields for every session: the date, the workout type, the machine type, the distance, the time, the average split, average and maximum heart rate, stroke rate, drag factor, calories, and, for an online workout, the piece-by-piece splits ([1] Concept2 2016). Every one of those is either measured by the machine or typed into a form. Not one of them is a feeling.

That is the boundary the whole article sits on. A logbook is a record of what happened, not a diary of how it felt.

MyNextRow syncs those sessions, derives the training-load metrics from them, and derives nothing else ([2] MyNextRow 2026). Everything that is not in the synced data has to arrive another way, and there are only two other ways: the goal you typed at onboarding, and the conversation.

Which is worth stating plainly, because the failure mode is not that the coach is careless. It is that a logbook has no field for the things that actually decide whether you should train today. There is no field for how hard it felt, none for soreness, none for sleep, none for the meeting that ran late, and none for the fact that Tuesday never happened.

The core idea

The coach reads the trend, not the session. MyNextRow hands the model your ten most recent sessions in full detail — date, type, distance, time, average split, watts, stroke rate, heart rate, drag factor, your note if you wrote one, and up to ten intervals from each — and separately a longer series of sixty sessions that it reduces to a load curve. That curve is split in two: a chronic load built over roughly six weeks, and an acute load built over one week. The distance between the two is what decides whether the coach adds volume, holds it, or backs off.

That is the fitness-versus-fatigue model in the form it was first proposed: each training session contributes an impulse that builds fitness, and a fatigue effect that decays faster than the fitness it leaves behind, so performance is the balance between the two ([3] Banister 1980). The consequence for a rower is that the single session is not the signal. A tidy 30-minute piece inside a bad fortnight is not a good fortnight, and the coach is reading the fortnight.

So which numbers change the advice? The ones that describe a pattern: how much you have done recently, how hard that volume has been relative to your own reference pace, how much you have accumulated, and how long it has been since your last hard session. Which ones do not? One good split. One fast 5K. One Saturday that felt great.

There is a second input inside the sessions, and it is the one people forget: your reference pace. The coach predicts your 2K from your most recent credible piece, checks that prediction against your other recent pieces, and attaches a confidence label to it. Below roughly 1,500 metres or above 30,000 metres a piece does not count at all, and if your best qualifying piece is nine months old the confidence label drops accordingly. That label is the difference between "here is your training zone" and "here is a guess with a warning attached".

And a third input that is not in the logbook at all: the goal. The sentence you wrote at onboarding, in your own words, and the category you picked — race, fitness, recovery or technique — go into every exchange alongside the sessions. Goal-setting research treats goals as mediators of incentives rather than expressions of preference ([10] Locke 2002), which is why the app asks for one instead of inferring it. The goal sits in the room when the prescription is chosen.

The how-to

Now the part that is actually useful: making the logbook worth reading.

Log a session-RPE. This is the one number the machine cannot produce for you. After each session, rate how hard it was on a zero-to-ten scale and put the number in your note field. Multiply it by the minutes and you have a load figure that does not depend on how the monitor counted the split. The method is old and well tested: session RPE multiplied by duration tracks training load closely enough to be used for monitoring across sports ([4] Foster 2001). It is also the kind of measure that monitoring reviews keep putting near the top, because the athlete's own report of fatigue, soreness, mood and sleep catches what the monitor misses. In one systematic review of athlete monitoring, subjective self-reported measures were more sensitive and more consistent than the objective measures they were compared against ([5] Saw 2016).

Write what changed, not what you did. "Felt strong, left shoulder aching" is worth more to the coach than "30 minutes at 2:05". The split is already in the data. The note field is the only channel through which the internal side of a session reaches the model.

Say why you skipped a session. The coach can see that you have not trained for three days — it has explicit rules for how much recovery time you have had — but it cannot see why. If the gap was a sprained wrist rather than a wedding, the right next session is different, and only a note can tell it. Missing sessions do not vanish from the load curve either: they decay it, which is exactly the wrong signal if the reason was illness.

Keep the log continuous through a bad week. A trend cannot be read through a hole, and a bad week is precisely when the trend is worth having. If you are ill or injured, say so in the next note. The coach reads the note and cannot infer it.

Expect the first weeks to be thin. The recent-history window fills after about ten logged sessions. The chronic load needs roughly six weeks of history before it means much, and the ratio between the two needs at least four. Before that, the coach is estimating, and it says so.

The trade-offs

Now the honest part, which is the part a reader is actually asking about.

It cannot see the sessions you did not log. This is the largest limit and it is structural rather than fixable: your logbook is a record of what you chose to record.

It cannot see how hard the session felt. There is no effort field in the Concept2 logbook and the app does not collect one. Your note field is the only route for that information.

It cannot see the sore shoulder, the bad night, or the deadline at work. Not unless you say so, and only in the conversation. This matters more than it sounds, because no single load number describes fatigue. Monitoring reviews are blunt about it: there is no single definitive marker, and the way to detect a meaningful change is to combine the load with the context around it ([6] Halson 2014).

It cannot see the boat. Everything in your logbook was recorded on an ergometer. In a study of forty-nine junior elite rowers who did both, ergometer speed and single-scull speed on the water correlated at R² of 28.9% — a real relationship, and a long way from one ([9] Nevill 2010). If you row on the water, your logbook is describing a machine, and the coach is reading a machine.

The number it prints is a guard rail, not a verdict. The acute-to-chronic load ratio behind its safety rules comes from a cohort of elite rugby league players, and even there the result was not the tidy one usually quoted. Players carrying a high chronic load were more resistant to injury than players carrying a low one across a range of moderate ratios, and less resistant when the acute load spiked ([7] Hulin 2016). The methodological critique since is blunt: the causal link to injury was never established, the statistical properties make it an inaccurate metric, and it adds noise and creates statistical artefacts ([8] Impellizzeri 2020). Read the bands as floors and ceilings, not as findings about your ribs.

It will not explain itself as cleanly as you would like. Separating what a model read from what it inferred is a research problem, not a formatting choice. The review the field keeps returning to argues that explainable AI should be built on the social-science research into how people define and evaluate explanations, rather than on the researcher's intuition of what a good explanation is ([11] Miller 2019).

And your response to the plan is yours. Studies of individual response to training programmes find large differences between people in how much they improve, with measurement error and within-athlete variability contributing much of the spread ([12] Hecksteden 2015). A coach that has read your logbook for two weeks is reading a trend that has not yet established what you do with it.

Where the privacy line sits. How MyNextRow handles the workout data it syncs from Concept2, the training-load metrics it derives from it, and the conversations you have with the coach is described in its privacy policy ([2] MyNextRow 2026). This article covers only what is read and how it shapes the advice.

What to do tomorrow

Two things, and they are the two that change what the coach can see.

1. Rate your next session out of ten and put the number in the note field. One digit. It is the only channel for the internal cost of the session, and it is the cheapest thing you will ever add to a logbook.
2. Write one sentence in your next note about what actually happened. Not the split — the split is already there. How it went, what changed, what you skipped.

Everything else in this article is secondary to those two lines, because everything else is already in the log and is already being read.