Shindo The one-sentence answer#
A training readiness score is a daily estimate of how much training load your body is currently able to absorb, computed by comparing this morning’s recovery signals against your own recent baseline — not against anybody else’s.
Everything interesting is in the second half of that sentence.
What goes into it#
Nearly every implementation blends some subset of four things:
Heart-rate variability. The variation in time between consecutive heartbeats, usually measured overnight, used as an indirect readout of autonomic nervous system state. Higher generally means more parasympathetic — more recovered.
Sleep. Duration, and some notion of quality or stage distribution. The better implementations score this against your sleep need rather than against eight hours.
Resting heart rate. Usually the lowest sustained rate during sleep. A rise of several beats above your own norm is a general stress signal.
Current training load. How much fatigue you are carrying from recent work — often expressed as form, or fitness minus fatigue. This is the input that separates a training readiness score from a general wellness score, and some apps omit it entirely.
Some products add skin temperature, respiratory rate, blood oxygen or subjective stress. These add sensitivity to illness in particular, but they are refinements rather than the core.
Why the baseline matters more than the measurement#
This is the part most explanations skip, and it is the whole game.
Heart-rate variability varies enormously between people. A well-trained 25-year-old might sit around 120 ms; an equally fit 50-year-old might sit at 35 ms. Neither number means anything on its own, and comparing yours to somebody else’s is meaningless.
What carries information is how far today sits from your own recent distribution. A drop of 20% below your personal norm is a signal regardless of whether your norm is 35 or 120. This is why a readiness score needs a few weeks of your data before it means anything, and why the first fortnight with any of these apps is the least reliable period, not the most.
The same logic applies to sleep. Scoring against a fixed eight hours permanently penalises a genuine six-hour sleeper and permanently flatters a nine-hour one. Scoring against personal need means only a deviation registers.
Why two apps give different numbers on the same data#
A frequent and reasonable complaint. The reasons are mundane:
- Different input sets. An app that ignores training load will call you recovered the morning after a hard century ride. One that includes form will not.
- Different weightings. How much is sleep worth against HRV? There is no settled answer, and each vendor picked one.
- Different baseline windows. Seven days, 30 days, 60 days — a short window adapts fast and is noisy; a long one is stable and slow to notice a real change.
- Different handling of missing data. Some apps substitute a neutral value when a night is missing, which drags the score toward the middle. Others leave the metric out and redistribute its weight.
- Different sensors. A wrist optical sensor and a chest strap do not measure HRV equally well, and where in the night the sample is taken changes the number.
None of this makes the scores useless. It does mean the absolute number is not comparable across apps, and that switching apps resets your baseline. Track the trend within one system.
The four questions worth asking of any readiness score#
1. Does it show you the drivers? A score with no explanation cannot be checked against how you actually feel, and cannot be overruled intelligently. “62” is a horoscope. “62, because your HRV is 18% below baseline and you are carrying −24 form” is a claim you can evaluate.
2. Does it include training load? If it only looks at recovery signals, it is a wellness score, not a training readiness score. The distinction matters most in exactly the situation you care about — the day after a hard session.
3. What does it do with missing data? If a night without the watch produces a confident number anyway, something was invented. The honest behaviour is to say the metric is unavailable.
4. Is it deterministic? The same inputs should always produce the same output. If a language model is generating the number rather than describing it, the score can differ between runs on identical data, which makes it unfalsifiable.
When to override it#
Readiness scores are genuinely useful and they are also routinely wrong, because they cannot see your life. Override the score when:
- You feel fine and the number is low, and there is an obvious explanation it cannot see — you drank two glasses of wine, you slept in a hot room, you had a late meal, you flew yesterday. All of these suppress HRV without meaning you cannot train.
- You feel awful and the number is high. Subjective state carries real information that no wearable measures. The score is not more expert than you are about you.
- It is race day. Race off the plan, not off a score.
Conversely, take it seriously when a low score persists for several days, when it comes with an elevated resting heart rate, or when it agrees with the fact that you feel terrible. The single most valuable thing these scores do is catch the slow slide into non-functional overreaching, which is precisely the thing that is invisible day to day.
What a readiness score cannot do#
It cannot diagnose. A suppressed HRV reading is a signal that something is off, and it is genuinely blind to what: hard training, a developing infection, work stress, alcohol, heat and poor sleep all look similar from the wrist.
It also cannot see technique, terrain, motivation, or whether the session that is scheduled is the right session. It answers one narrow question — how much load can you take — and it should be treated as one input among several, one of which is you.
Shindo computes readiness from sleep, heart-rate variability, resting heart rate and current form, always shows the four drivers behind the number, and leaves a metric blank rather than inventing a value when data is missing. The methodology is published in full.
Read next