The signs of accumulated training fatigue, as Zyra Workouts reads them, are four numbers in your log moving together: working-set volume above your usual, sets logged closer and closer to failure, body weight sliding downward, and — if you have connected it — recovery data that has worsened against its own baseline. The app combines them into a single training-load score from 0 to 10 with three bands. This page shows the exact weights and maths, and is equally clear about what the score is not: it is a model of your logged training load, not a diagnosis of overtraining, illness or anything about your health.
Four inputs — volume, effort, weight slope, recovery
Each signal is scaled to a number between 0 (no contribution) and 1 (maximum), then weighted. From the source:
| Signal | Weight | How it is scaled to 0–1 | Where the data comes from |
|---|---|---|---|
| Volume load | 0.30 | this week's working sets ÷ 28-day weekly average ÷ 2, capped at 1 — so exactly your usual volume scores 0.5, double scores 1, half scores 0.25 | your session log |
| Effort | 0.25 | (average RPE − 5) ÷ 5 — RPE 5 scores 0, RPE 7.5 scores 0.5, RPE 10 scores 1 | the RPE on your working sets, averaged over 28 days |
| Weight slope | 0.15 | −(% change per week), capped at 1 — losing 1% a week scores 1; a flat or rising weight scores 0 | your smoothed body-weight trend |
| Recovery | 0.30 | 1 − readiness, where readiness averages HRV vs its baseline, resting heart rate vs its baseline, and sleep hours ÷ 8 | connected health data, if any |
Three design choices are worth spelling out.
Volume is relative, not absolute. Twenty sets a week is a heavy week for one lifter and a light one for another; the model only asks how this week compares with your own last four. The baseline is the working-set count over the previous 28 days divided by four, so a week at your normal volume contributes 0.5 — the middle — and only a jump above normal pushes the signal up.
Effort is the RPE you logged. The scale starts at RPE 5 because anything easier is warm-up territory (the same reasoning behind the RIR clamp at RPE 6). A month of sets at RPE 8 scores 0.6; a month at RPE 9.5 scores 0.9. If you do not log effort, this signal is simply absent.
The slope is one-directional. Only a falling body weight adds to the score. A rising weight scores zero, not negative. The weight is 0.15 — the smallest — because a downward slope has many causes and this model is not the place to interpret them.
Missing data is dropped, not guessed. If you have no connected recovery data, the recovery term is left out and the remaining weights are re-normalised — the score is computed over what exists. With no signals at all there is no score.
In plain terms: how much more than usual, how hard, whether you are getting lighter, and how your recovery data looks — each in proportion.
The bands
The score is 10 × (Σ weight × signal) ÷ (Σ weights present), rounded to one decimal, and banded:
| Score | Band |
|---|---|
| below 4.0 | Low |
| 4.0 to below 7.0 | Moderate |
| 7.0 and above | High |
A worked example with all four signals present. A lifter at exactly their usual volume (0.5), averaging RPE 8 (0.6), body weight flat (0), and recovery data reading somewhat below baseline (recovery signal 0.3):
0.30 × 0.5 + 0.25 × 0.6 + 0.15 × 0 + 0.30 × 0.3 = 0.15 + 0.15 + 0 + 0.09 = 0.39 → 3.9 — Low.
The same lifter after a week at one and a half times their usual sets (volume signal 0.75):
0.225 + 0.15 + 0 + 0.09 = 0.465 → 4.7 — Moderate. One change, one band.
And the same lifter with no recovery data connected, so only volume and effort count:
(0.225 + 0.15) ÷ (0.30 + 0.25) = 0.682 → 6.8 — still Moderate, but near the top of the band, because the two remaining signals carry all the weight.
The tile also shows a seven-day sparkline and a one-word trend — rising, falling or stable — built by holding the slow signals constant and varying the recovery input day by day (sleep if you track it, otherwise HRV). A score that is Moderate and *rising* reads differently from one that is Moderate and *falling*, even at the same number.
What to change first
The model gives you the answer to this question in its weights. Two of the four signals are entirely yours to set, and they carry 0.55 of the total between them:
- Volume (0.30) — the biggest lever you control. Bringing this week's working sets back to your 28-day average moves the volume signal from wherever it climbed back to 0.5. If the volume signal is what spiked, that alone can drop a High week to Moderate.
- Effort (0.25) — a week at a higher RIR target (RPE 8 instead of 9.5) pulls the average down over the following weeks. This is the second of the two deload levers described in training plateau or time to deload, and the exercise-level rules tell you which lift needs it: a Deload needed label there means e1RM falling with RIR averaging 1 or less.
- Recovery (0.30) and weight slope (0.15) are readings, not dials. You cannot set them from the program; you can notice them, and you can look at what changed in the weeks before the score rose.
The single most useful habit is to look at which signal is driving the score before changing anything. A High score from volume alone after a deliberately big week is expected and short-lived; a Moderate score that has been rising for three weeks with effort near 1.0 and e1RMs slipping is the pattern the back-off rule cannot fix one step at a time. This is arithmetic on your log, not a trial result.
What this model is not
It is not a diagnosis. "Overtraining" as a clinical description involves symptoms, history and examination that no app has; Zyra Workouts does not use the word for a reason, and this score does not detect it. It is not a health measurement: the recovery inputs are numbers your devices reported, compared with their own recent average, and the score treats them as training context, nothing more. It is not personal: two people with identical logs get identical scores, whatever else is going on in their lives.
What it is: a consistent, reproducible summary of four things you logged, weighted the same way every day, so that "I think I'm doing more than usual and it's all feeling hard" becomes a number you can watch. If something feels wrong in a way that is not explained by the log — persistent, or not just tired — that is a conversation for a coach or a qualified professional, and the score has nothing to add to it.
The short version
Four logged signals — volume against your 28-day baseline (0.30), average RPE (0.25), body-weight slope (0.15), recovery data (0.30) — scaled 0–1, weighted, and shown out of 10: Low under 4, Moderate to 7, High above. Missing signals are dropped, not assumed. Volume and effort are the levers you hold; recovery and weight are readings. It is a training-load model built from your log, and explicitly not a medical judgement. Zyra Workouts is in development and not yet on the App Store; the rules quoted here are read from its source code and may change before release. Training guidance, never medical advice. Educational overview only — not medical advice.