Sleep and training: what actually works (and what does not)

The science, without the bullshit. What you can do tonight, with the numbers behind it.

Why sleep is your most powerful training variable

You track your sets, your kilos, your kilometres, your protein. Sleep, you just endure. Yet it is while you sleep that training turns into adaptation.

A few numbers to set the scene:

Your testosterone. One week at 5 h a night lowers the testosterone of healthy young men by 10 to 15%. That is the equivalent of 10 to 15 years of ageing, in 7 days (Leproult 2011).

Your muscle. One sleepless night cuts muscle protein synthesis (the building of new muscle after a protein meal) by −18%, and raises cortisol by 21% the next day (Lamon 2021). Five nights at 4 h produce the same 18% drop. Good news: training during that period almost cancels the drop (Saner 2020). So a bad night is not a reason to skip the session.

Your cut. This is the study to remember. Same calorie deficit, same weight loss, but one group sleeps 8 h 30 and the other 5 h 30. Result: those sleeping 5 h 30 lose 55% less fat and 60% more muscle. The fat share of the weight lost falls from 48% to 20% (Nedeltcheva 2010). In other words: a badly slept cut is a cut that burns your muscle.

Your appetite. Short sleep raises ghrelin (the hunger hormone) by 28% and appetite by 24%, especially for sweet and fatty food (Spiegel 2004). Conversely, adults sleeping less than 6 h 30 who are helped to sleep 1 h 12 more spontaneously eat 270 kcal less per day, with no dietary instruction at all (Tasali 2022).

Your health. Habitually sleeping less than 6 h raises mortality by about 12% (Cappuccio 2010), stroke risk by 29% (2025 meta-analysis), and multiplies by 4 your chances of catching a cold after exposure to the virus (Prather 2015). The American Heart Association has ranked sleep at the same level as smoking or physical activity among its 8 cardiovascular pillars.

How many hours?

The official recommendations: 7 to 9 h for an adult, at least 7 h on a regular basis (National Sleep Foundation, AASM).

For athletes, the expert consensus in the British Journal of Sports Medicine says 7 to 9 h, up to 10 h during heavy training loads (Walsh 2021). And be warned: athletes sleep on average less than everyone else (6 h 48 in 124 elite athletes measured by sensor), especially when training is before 7 am (Sargent 2014).

Does sleeping more work? College basketball players who extend their sleep to ~10 h for 5 to 7 weeks improve their sprint, their free-throw accuracy (+9%) and their reaction time (Mah 2011). Small study (11 players, no control group), but a review of every intervention tested in athletes concludes that sleep extension is the most effective of them all (Cunha 2023).

Regularity: the finding that changes everything

This is the most important result of recent years.

An Australian team analysed 10 million hours of sensor data from 60,977 adults, then tracked who died within 6 years (Windred 2024). They computed a regularity index: how consistently you are asleep or awake at the same times from one day to the next.

Result: the most regular 80% had 20 to 48% lower mortality than the most irregular 20%. 22 to 57% lower for cardiac and metabolic deaths. And above all: when regularity and duration go into the same model, regularity predicts mortality better than duration.

Why? Your internal clock sets your temperature, your hormones, your digestion and your performance capacity over 24 h. Schedules that shift by 2 h between weekdays and the weekend ("social jetlag") desynchronise it: each hour of shift is associated with a higher BMI and more inflammation (Roenneberg 2012).

Sunlight: the lever nobody sells

Your daytime light sets your clock. Your evening light throws it off.

An international expert panel set precise targets (Brown 2022):

• Daytime: at least 250 melanopic lux (light as the cells of your clock see it). A normal indoor space provides 50 to 150. Outside, even under a grey sky, you are at 1,000 minimum. The only realistic way to hit the target: go outside.
• Evening, from 3 h before bed: 10 lux maximum. A normally lit living room is at 100–300. You are 10 to 30 times too bright.
• Bedroom: less than 1 lux. Complete darkness.

Across 86,772 adults fitted with light sensors, a lot of light at night is associated with 30% more depression, and a lot of light during the day with 19% less depression, independently of sleep (Burns 2023). Every hour spent outdoors is associated with less fatigue and easier sleep onset (Burns 2021, 400,000 people).

And sleeping with a 100 lux light (a bedside lamp), for a single night, raises your night-time heart rate and your insulin resistance the next day (Mason 2022).

Screens: blue light is a false culprit

You have read everywhere that blue light kills your sleep. The science is more nuanced.

What is true: 4 h on a tablet at maximum brightness, 5 evenings in a row, suppresses 55% of your melatonin and delays your clock by 1 h 30 (Chang 2015). But that is a lab protocol, far from your real use.

What is also true:

• In teenagers, what disturbs sleep is screen brightness, not its colour. Filtering blue at equal brightness changes nothing (Duraccio 2022).
• The iPhone's "Night Shift" mode does not improve sleep. The only measurable difference is between "phone" and "no phone" (Duraccio 2021).
• Blue-light blocking glasses: the 2023 Cochrane review finds no evidence of an effect (Singh 2023), and neither does the most recent meta-analysis on objective measures (Luna-Rangel 2025).
• Across 122,000 adults, the phone in the hour before bed costs on average 50 min of sleep per week, mainly because it delays bedtime (Rezaei 2025).

An important nuance: sensitivity to evening light varies by a factor of 50 between individuals (Phillips 2019). Some are disturbed by a night light, others tolerate a ceiling light. "Screens do nothing to me" can be true.

Caffeine: far earlier than you think

The reference meta-analysis compiled 24 studies with objective measures (Gardiner 2023). On average, caffeine taken too late:

• costs you 45 min of sleep in total,
• makes you take 9 min longer to fall asleep,
• adds 12 min of awakenings during the night,
• takes away 11 min of deep sleep (the phase where you secrete growth hormone).

Most important: the researchers computed the delay needed for the effect to disappear. A coffee (107 mg): 8 h 48 before bed. A pre-workout (217 mg): 13 h 12 before.

For an 11 pm bedtime, that gives: last coffee around 2 pm, last dosed pre-workout before 10 am. And that is an average: the half-life of caffeine ranges from 2 to 10 h depending on your genetics, and it doubles on the contraceptive pill.

A meta-analysis targeted at athletes confirms it: caffeine before a late-day session = shorter sleep and longer sleep onset, especially at the effective doses of 3 to 6 mg/kg (Kocak 2025).

Evening training: the myth that will not die

23 studies comparing an evening session with no session at all (Stutz 2019): no difference in sleep onset time, duration or sleep efficiency. Deep sleep even increases slightly. The only exception: an intense effort finished less than 1 h before bed, with the heart still 20 beats above resting.

Data from 4 million nights of WHOOP bands refines the picture (Leota 2025): a very demanding effort finished less than 4 h before bed is associated with later sleep onset, shorter sleep, a faster night-time heart rate and lower heart rate variability. Finished 4 h before or more: no effect. Moderate effort: no effect, whatever the time.

And the background: training regularly is one of the best known "sleeping pills", with better latency, duration and quality in meta-analyses (Kredlow 2015), including for resistance training (Kovacevic 2018).

Alcohol: two drinks are enough

Meta-analysis of 27 studies (Gardiner 2025):

On your tracker, this is the most visible signal: night-time heart rate rises and heart rate variability drops by as much as −40% at high doses (Pietilä 2018, 4,098 nights).

Eating, drinking, temperature, napping

Carbs in the evening: yes. A high-glycaemic-index meal 4 h before bed halves sleep onset time (9 min versus 17) (Afaghi 2007). Small study, but the "no carbs in the evening" dogma has no support on the sleep side, and your glycogen stores will thank you.

Protein at bedtime: yes. 40 g of casein before sleep increases overnight muscle synthesis and, over 12 weeks, mass and strength, without degrading sleep (Snijders 2015).

Water. Sleeping 6 h instead of 8 is associated with a 16–59% higher risk of being poorly hydrated the next day, in 20,000 adults (Rosinger 2019). But drinking a lot in the evening wakes you to urinate, and every awakening costs deep sleep. Hydrate during the day, cut back 2 h before bed. The half-litre before sleep repairs nothing.

Temperature. Your body has to lose about 1 °C to fall asleep. Bedroom at 18–20 °C; above 24 °C, deep and REM sleep decrease (Okamoto-Mizuno 2012). And paradoxically, a hot shower or bath (40–42 °C), 1 to 2 h before bed, cuts sleep onset time by 36%: the heat dilates your vessels and speeds up heat loss afterwards (Haghayegh 2019). A cold shower in the evening does the opposite.

Napping. In athletes, 20 to 90 min between 1 pm and 4 pm improves alertness, reaction and often sprint and strength endurance, especially after a short night (Lastella 2021). Never after 4 pm, never if you already struggle to fall asleep in the evening.

Supplements: what works a little, what does not

Golden rule: always compare against the effect of behaviours. A late coffee costs you 45 min. No supplement gives you back that much.

Melatonin. It is not a sleeping pill, it is a "it is night" signal for your clock. Meta-analysis of 19 trials: −7 min to fall asleep, +8 min of sleep (Ferracioli-Oda 2013). In insomniac adults with no other problem, another meta-analysis finds no effect (Choi 2022). Where it shines: jet lag and late sleepers who want to advance their clock. Dose: 0.5 to 2 mg, taken 3 h before the desired bedtime rather than 30 min (Tomasello 2024). Watch out for the products: across 31 supplements tested, the actual content ranged from −83% to +478% of the stated dose (Erland 2017).

Magnesium. The largest trial to date (155 adults, 250 mg of bisglycinate, 4 weeks) finds a small effect: d = 0.2, the equivalent of a slight noticeable improvement but nothing spectacular (Schuster 2025). The effect is clearer in those who eat few green vegetables, legumes and nuts. In older adults: −17 min to fall asleep, weak evidence (Mah 2021). 200–400 mg in the evening, bisglycinate or citrate, not oxide. Good risk/price ratio if your intake is low.

Glycine. 3 g before bed lowers core temperature and improves subjective quality and next-day fatigue in Japanese trials (Yamadera 2007, Bannai 2012). Problem: small trials, all funded by the manufacturer. Safe and cheap; if you take 10 g of collagen in the evening, you already have 3 g of it.

Ashwagandha. Meta-analysis of 5 trials: moderate effect on sleep, especially at 600 mg and beyond 8 weeks (Cheah 2021). But short studies, funded by the manufacturers, and above all several dozen published cases of liver injury; Denmark has banned it. The worst benefit/risk ratio of the lot. As a last resort, short courses, stop at the first digestive sign.

Tart cherry. +25 min of sleep in 20 people (Howatson 2012), +84 min in 8 insomniac older adults (Losso 2018). Tiny samples, funded by the cherry industry. Its real strength is muscle recovery instead. Watch out for the 25–30 g of sugar per glass.

Valerian, CBD, ZMA: no demonstrated objective effect. CBD can contain THC (anti-doping testing).

Creatine, bonus: it does not disturb sleep, and a large acute dose blunts the cognitive effects of a sleepless night (Gordji-Nejad 2024). For managing a short night, not for sleeping better.

Reading your tracker without driving yourself mad

Watches detect sleep well and wakefulness badly: they overestimate your sleep time by 20 to 40 min. Stage classification (deep, REM) is poor, with agreement with the lab often below 50% (Chinoy 2021).

What is reliable and useful: your bedtimes and wake times (regularity), your night-time heart rate and your heart rate variability. +5 bpm compared with your baseline = alcohol, illness, late session or overreaching. A variability drop of more than 15% over 3 nights = a signal to deload.

What is not: the deep sleep of one isolated night and the overall "score". Look at 7-day trends, never a single night. Obsessing over your score to the point of sleeping badly has a name, orthosomnia (Baron 2017).

See a doctor if you snore with pauses, if you are sleepy despite 7 h, or if you have trouble sleeping more than 3 nights a week for the past 3 months. In that last case, cognitive behavioural therapy for insomnia (CBT-I) beats every sleeping pill and every supplement (Edinger 2021).

Where to start, in order

1. Fixed wake-up time, ±30 min, 7 days a week.
2. At least 7 h 30 in bed, 8 h+ under heavy load.
3. Outside in the morning, dark in the evening, dark bedroom.
4. No caffeine after 2 pm. No pre-workout in the evening.
5. No alcohol in the 3–4 h before bed.
6. Bedroom at 18–20 °C, hot shower 1–2 h before.
7. Intense session finished 2 to 4 h before bed.
8. Phone out of the bedroom.
9. Carbs and protein at dinner, fluids cut back 2 h before.
10. 20 min nap before 3 pm if the night was short.
11. Supplements last: magnesium if your intake is low, melatonin to reset.

By profile

Hypertrophy, high volume. 8 h, non-negotiable. On a cut, below 6 h 30 of sleep, reduce the deficit to 15–20% max and volume by 20%: otherwise you lose muscle instead of fat. Casein at bedtime is fine. Pre-workout in the morning only.

Max strength, powerlifting. Your strength on a single max holds up surprisingly well to a bad night; it is the repeated efforts and the end of the session that suffer (Craven 2022). Do not skip the heavy session, cut the accessories. Evening competition: nap 3–4 h before, caffeine 3 mg/kg 1 h before, day off afterwards.

Hybrid, strength + endurance. Often more than 8 h needed. Endurance is what degrades most after a short night (higher perceived effort, shorter time to exhaustion): put your key endurance sessions after the good nights. Watch night-time heart rate and variability.

Training before 7 am. Go to bed 8 h 30 before you wake. 10,000 lux lamp at breakfast in winter. 20 min nap in the afternoon. Melatonin 0.5 mg around 8 pm if you cannot fall asleep before 10 pm.

Women. In the second half of your cycle, your temperature rises by 0.3–0.5 °C: a cooler bedroom and a hot shower beforehand count more. On the pill, your caffeine lasts twice as long: move your cut-off earlier.

Over 60. Lighter and earlier sleep is normal. Morning light and regular training are your best-proven levers. Magnesium has its best evidence in your age bracket. Get screened for sleep apnoea.

Travel, competition away. More than 5 time zones: melatonin 0.5–3 mg at the local bedtime, sunlight in the morning when heading east. And take heart: the night before a competition is often bad, it is the night before that one that counts.

The most valuable data from your tracker is not your sleep score. It is the time you get up, and how much it is the same every day.

This article presents the current scientific data. For a personalized program, consult a healthcare professional.