Recovery is what makes training count
Training is the stimulus. The adaptation happens afterwards, and it depends on sleep, food, and time. This is why a demanding programme layered onto short nights and a stressful quarter often produces less than a modest programme laid onto a stable life.
Recovery is not a reward for hard work. It is part of the work.
Sleep first, and seven hours is the number
The American Academy of Sleep Medicine and the Sleep Research Society ran a formal consensus process and landed on a single recommendation for adults aged 18 to 60: sleep seven or more hours a night on a regular basis. Regularly sleeping less is associated with weight gain, diabetes, hypertension, heart disease and stroke, depression, and higher mortality, and with impaired performance, more errors, and more accidents in the shorter term.
Those are associations drawn largely from observational data, not the measured result of adding an hour to your night. The recommendation is still the clearest number on this page, and it is the one worth protecting before you rearrange a training split.
What actually shifts sleep
A consistent wake time, a dark cool room, and morning daylight are the standard sleep-hygiene levers, and they are more established as a package than any single one of them is on its own. Two of the habits people most often get wrong have been tested directly.
Caffeine, later than you think. In a randomised crossover trial, 400 mg of caffeine taken six hours before bed reduced self-reported sleep by 41 minutes, and objective recording showed more than an hour lost. Twelve healthy adults with a mean age of 29 is a small sample, so treat the size of the number loosely. The direction is the useful part: an afternoon coffee is not neutral by bedtime.
Alcohol, differently than you think. A 2025 meta-analysis of 27 studies found that alcohol delays the onset of REM sleep and shortens it, starting at roughly two standard drinks and worsening with the dose. Only at around five drinks did it shorten the time taken to fall asleep. The pooled effect on total sleep time, sleep efficiency, and time spent awake in the night was too uncertain to call. So the common line that a nightcap fragments your sleep overstates what has been shown; what has been shown is that it degrades REM sleep from a low dose upwards.
Managing load, not just sessions
Your body does not distinguish between training stress and everything else. Deadlines, travel, illness, caring for a sick child, and a heavy week of sessions all draw on the same account.
When life gets loud, reduce training volume before you reduce frequency. Keeping the habit at half the dose is easier to return from than stopping altogether. Plan lighter weeks deliberately every month or two rather than waiting for fatigue to force one.
What wearable data can and cannot tell you
Sleep stages, readiness scores, and heart-rate variability are estimates produced by an algorithm from a small number of signals, and the accuracy is uneven in a way the display does not communicate.
A laboratory comparison of seven consumer devices against polysomnography in 34 healthy adults found that all of them detected sleep well — sensitivity of 0.93 or better — but detected wake poorly, with specificity between 0.18 and 0.54. Two of the devices overestimated total sleep time by 43 to 47 minutes. Sleep staging was, in the authors’ words, mixed and often poor: light, deep, and REM sleep were regularly misclassified.
Two rules keep the numbers honest.
- Do not let a score overrule how you feel. If you feel good and the app says otherwise, train and see how the session goes.
- Watch the direction, not the day. A steady downward drift across two weeks is information. Tuesday being lower than Monday is within the error of the device.
If a device makes you anxious about sleeping, it is costing you more than it returns.
A short readiness check
Before a hard session, ask three questions. Did you sleep close to your normal amount? Is your motivation roughly where it usually is? Does yesterday’s soreness feel like stiffness rather than pain?
Three yeses means proceed as planned. Two means reduce the intensity and keep the session. One or none means make it easy movement and try again tomorrow.
When to take it to a clinician
Poor sleep that persists is a medical question, and no amount of programme adjustment substitutes for having it looked at. Book an appointment if any of these describe you.
- Loud snoring, gasping, or witnessed pauses in your breathing at night, particularly alongside morning headaches or a dry mouth. This is the picture that prompts testing for obstructive sleep apnoea.
- Daytime sleepiness that persists despite adequate time in bed — falling asleep in meetings, or fighting to stay awake while driving. Stop driving and get seen.
- Trouble falling or staying asleep on most nights, for months, with daytime consequences. Persistent insomnia is treatable, and the treatment is not another sleep-hygiene checklist.
- Fatigue that keeps deepening across weeks despite reduced training, especially with unexplained weight change, breathlessness, or low mood.
- A resting heart rate that stays substantially elevated for more than a week or two without an obvious cause.
One point matters more than the rest: the American Academy of Sleep Medicine’s diagnostic guideline makes a strong recommendation that clinical tools, questionnaires, and prediction algorithms should not be used to diagnose sleep apnoea in the absence of a sleep study. Your ring’s apnoea alert is a prompt to book a test, not a result.
What the evidence does not show
- That any of the sleep advice here has been proven to improve your training. The seven-hour recommendation rests mostly on observational health data. Nobody has randomised lifters to sleep durations for a year and measured their squat.
- That a consumer device can stage your sleep. Detecting sleep from wake is something these devices do reasonably well. Telling deep sleep from light sleep is not, and the reported minutes in each stage should not be treated as measurements.
- That heart-rate variability tells you whether to train today. The daily readiness score is a proprietary model, not a validated clinical measure, and it has not been shown to make training decisions better than asking yourself how you feel.
- That the caffeine finding is precise. Twelve people, one dose, one laboratory. The six-hour cut-off is a reasonable default, not an individually calibrated number.
- What alcohol does to your total sleep. The meta-analysis could not resolve it. The REM effect is the part that is established.