Exercise and sleep: when to train, and what a bad night costs

Two questions hide inside that headline, and they are not the same question. One is what training does to your sleep: does an evening session wreck the night? The other is what sleep does to your training: should you even lace up after five hours in bed? The first question gets the magazine coverage. The second is the one you actually ask yourself, at 5am, in a dark hallway, with three kids asleep upstairs.

This page takes both, in that order. What happens to sleep after forty, meaning why it gets lighter and more broken in the first place, belongs in the guide to sleeping better after 40. This one is about timing, and about the trade between sleep and training.

First: exercise is on sleep's side

This is the least disputed part of the subject. Sweden's national health service, 1177, puts it plainly: regular physical activity usually helps you sleep better, and activity outdoors during daylight helps most of all. The mechanism is no mystery. You get genuinely tired by evening, daylight steadies your body clock, and movement is one of the few things that reliably lowers the background tension keeping your head busy when it should be quiet.

So the basic advice is simple: move regularly. The question of what time is a fine adjustment on top of an answer that is already positive. If the choice is between a 9pm session and no session at all, the first is almost always better, both for your sleep in the long run and for everything else.

Timing only becomes worth arguing about once you both train regularly and sleep badly. At that point it is one of the cheapest things to try, because it costs you no training time whatsoever.

Evening training: the advice moved from one hour to four

The old rule said not to train in the hours before bed. Through the 2010s that reversed. Now it is partly reversing again, and it is worth understanding why, because you will run into both answers.

In 2019, researchers at ETH Zurich published a systematic review in Sports Medicine. They screened almost twelve thousand records down to 23 studies, comparing sleep after a single evening session with sleep after an evening without one, in healthy adults. The result cut against the old rule. Evening exercise did not make sleep worse. On average, participants got 1.3 percentage points more deep sleep, less of the lightest sleep stage, and a slightly later start to dream sleep. The authors' own conclusion says the evidence does not support the idea that evening exercise harms sleep, and if anything points the other way.

But that conclusion carries a caveat that almost always falls away when the finding gets repeated: time to fall asleep, total sleep time and sleep efficiency may suffer after vigorous exercise ending an hour or less before bed. So the evening is not the problem. The intensity in that last hour is.

Then came 2025. Nature Communications published an analysis of 14,689 physically active people who wore a biometric tracker for a year, covering 4,084,354 nights between them. That is a body of evidence on a completely different scale from twenty-three laboratory studies. A clear pattern showed up: the later the session and the harder it was, the later people fell asleep, the less they slept, the worse their sleep quality, the higher their resting heart rate overnight, and the lower their heart rate variability. The authors' practical conclusion landed on a number other than one hour. Sessions finishing at least four hours before sleep onset showed no association with any change in sleep at all, regardless of how hard they were.

Why two good studies reach different answers

The tempting move is to pick whichever number you like. The honest one is to notice that the two studies are not measuring the same thing.

The difference in design explains most of it. An experiment can establish that the session caused the difference, but only for that night, in a lab, among a small number of volunteers. An observational study captures real life but can never rule out that people who train late differ in other ways. Shift work, long commutes, evenings with small children, a later body clock: all of it can push the session later and shorten the night, without the session being the cause.

Two more things belong in an honest reading. In the 2019 review, every significant moderating effect vanished when a single study was removed from the calculation, which the authors report themselves. That is a sign the finer associations rest on thin ice. And in the 2025 study, two of the authors are employed by Whoop, the company whose trackers supplied the data. This is openly declared in the paper, and the paper states that the company had no influence on the analysis, but it belongs alongside the four-hour figure whenever that figure gets passed on. The participants are also people who choose to wear an advanced fitness tracker, so they are not a cross-section of the population.

That combination is worth flagging for anyone well into their training. Interval sessions are by definition the hardest thing in your week. If they sit at 8:30pm on a Tuesday, they are the one session type whose timing is worth moving first.

What your watch shows the morning after a late session

The overnight heart rate finding is the most useful part of the 2025 study, because it is something you can see for yourself. Later and harder evening sessions tracked with both a higher resting heart rate overnight and lower heart rate variability. If you wear a watch or a ring, that means the morning after a late hard session often looks worse on screen.

That is expected. It is not an alarm. A hard session leaves your nervous system running higher for hours, sometimes a full day, and that recovery is exactly what the measurement is picking up. The size is known: choosing maximal over light exercise went with an overnight resting heart rate 3.9 beats per minute higher if the session ended two hours before habitual sleep onset, and 9.4 beats higher if it ended two hours after. Note that this is heart rate measured across the night, not a reading taken in your living room in the morning. Where the line between noise and signal sits for you personally is covered in the guide to resting heart rate, and how to read a morning below your own baseline without overreacting is covered in the guide to HRV.

The practical rule is the same whichever number you watch: a single morning is a data point, not a verdict. It is weeks in a row heading the wrong way that mean something.

The other direction: what a bad night does to your session

Now turn the question round. You slept five hours. Should you train?

The best synthesis comes from Sports Medicine in 2022, where researchers pooled 227 measurements from 69 studies. They compared performance after normal sleep with performance after six hours or less. The average came out as a 7.56 percent drop, with an uncertainty range running from 3.13 to 11.9 percent. The effect was significant across every category tested: strength, endurance, speed, skill.

Three caveats belong with that number, and they make it less alarming than it looks.

• The spread between studies was extreme. The measure of how much the studies disagreed came out at 98.1 percent, which is close to the maximum. The 7.56 percent average therefore does not reliably describe what will happen to you tomorrow morning.
• Nine in ten participants were men. Of the roughly eleven hundred participants, 959 were male. How well the figure transfers to women is simply not settled by this body of evidence.
• The sleep loss was severe. The cut-off was six hours or less, and the clearest effects came from genuine sleep deprivation. A night where you got seven hours instead of eight counts as normal sleep here, meaning it sits in the control condition rather than in the sleep loss being tested.

None of that changes the headline: too little sleep makes you worse. But there is a real difference between one bad night and a wrecked week.

The pattern matters more than the hours

The genuinely useful part of the 2022 synthesis is not the average. It is the subgroups. The researchers split the sleep loss by how it happened, not just by how many hours went missing.

Read that middle row again. Shortening the night from the back end, by waking earlier than your body intended, was one of the two patterns that consistently produced worse performance. Shortening it from the front, by going to bed later, did not. The researchers also found performance dropping by roughly 0.4 percent for every hour awake before the test, for those two patterns specifically.

This is the sentence on the page worth remembering, because it points straight into the reader's daily life: the same number of lost hours weighs differently depending on which end of the night you take them from.

The 5am session is a trade, not a free lunch

For an adult with a job, children and a calendar that is already full, early morning is often the one hour nobody else claims. That is a genuinely good reason to train then. The session actually happens, no meeting can move it, and you have done the most important thing of the day before the house wakes up.

But that solution is precisely the one that collides with the table above. A 5am alarm is, in measurement terms, late restriction. It is exactly the pattern that produced consistently worse performance. Here is what the trade looks like:

• Move your bedtime earlier by the same amount and you lose no sleep at all. The morning session is then not a trade-off but simply a shifted day. That is the version that holds up over years.
• Keep the same bedtime and you pay for the session with an hour of sleep, every time, from the end of the night that measured worst. Does it work for a spell? Often. Does it hold for ten weeks running? Rarely.

One honest caveat: the 2022 synthesis does not settle this particular case. It did find late restriction to be a harmful pattern, but it also found that tests performed in the morning were largely unaffected while afternoon tests suffered consistently. The authors' own practical recommendation is in fact to schedule training in the morning when sleep is going to be short anyway. So the two findings pull in different directions for someone who wakes early specifically in order to train, and the study does not resolve it. What remains is the reasoning above: the cost depends on whether you moved your bedtime.

There is a third route that sidesteps the trade entirely, which is to put the session inside a journey you are making anyway. Running to work takes the minutes out of your commute rather than out of your night. More variations on the same idea are in the guide to training when you have no time.

Injury risk: the number everyone quotes

Sooner or later you will meet the claim that sleeping less than eight hours carries 1.7 times the injury risk. The figure is real. It comes from a 2014 study in the Journal of Pediatric Orthopaedics. Four things that rarely travel with the quote are worth knowing.

• The participants were teenagers. 112 school athletes in grades 7 to 12, average age 15. That is not a forty-five-year-old running to work.
• The uncertainty range touches 1.0. The reported interval ran from 1.0 to 3.0. The lower bound sits right at the threshold of no effect at all.
• Sleep was self-reported through a survey, while the injuries were pulled retrospectively from school records.
• School year was the more precisely estimated predictor. Risk rose 1.4 times per grade level, with a narrower uncertainty range than the sleep figure.

The conclusion is not that the number is wrong. It is that the number is weaker and more age-bound than it sounds when quoted as a fact about adults. That chronic sleep loss is bad for a body expected to absorb training load hardly needs this study to prove. That the figure 1.7 applies to you, specifically, is not what was shown.

The decision rule

Most of the above boils down to a handful of decisions. Here they are.

The most common mistake is turning this into an all-or-nothing choice. A short easy session after a bad night is almost always a better decision than either the planned hard session or the sofa. Lowering the ambition is a way of continuing, not a way of giving up. More on holding the habit through weeks when form deserts you is in the guide to staying motivated to exercise.

When to see a doctor

The occasional night with too little sleep is not harmful to a healthy body. Sleep trouble that persists is a different matter, and there is treatment that works.

1177, Sweden's national health service, lists the following as signs of sleep difficulty, and advises seeking help if the trouble does not improve despite changes you have made yourself:

• It takes at least 45 minutes to fall asleep at night.
• You wake once or several times a night and struggle to get back to sleep, or you wake too early and cannot get back to sleep.
• You are tired during the day, function worse in everyday life, feel low or irritable, or struggle to concentrate.
• The trouble persists even after you have changed your habits to sleep better.

Contact your doctor in that case. If you are employed, an occupational health service may also be able to help. In Sweden you can call 1177 for health advice. Heavy daytime sleepiness despite enough hours in bed, loud snoring, or pauses in breathing that someone else has noticed should always be raised with a doctor, since they can point to a condition that needs investigating. If you also wake regularly in the middle of the night, the causes are gathered in the guide to why you wake at 3am.

This page provides general health information. It does not replace assessment by a clinician, and it does not diagnose anything.

Frequently asked questions

Is it bad to exercise in the evening?

For most people, no. A review of 23 studies in healthy adults found that evening sessions did not worsen sleep and in fact produced slightly more deep sleep. The exception is hard training: in that same review, time to fall asleep, total sleep time and sleep efficiency could suffer when a vigorous session ended an hour or less before bed. An easy evening session is rarely the problem.

How many hours before bed should I stop exercising?

It depends which kind of study you ask. Laboratory studies point to only the final hour being sensitive, and then only for hard sessions. A large 2025 study using wearable trackers found instead that sessions ending at least four hours before sleep onset showed no association with any change in sleep, no matter how hard they were. Four hours is the cautious margin, one hour the bare minimum. If you sleep well, you probably need no rule at all.

Should I train if I slept badly?

Usually yes, but easy. The occasional bad night is not dangerous for a healthy body. Lower the ambition instead of cancelling: make it shorter and keep it at conversational pace. What measurably drags performance down is severe sleep loss, not one night that was a bit worse than usual. If you are ill, dizzy or running a fever, do not train at all.

Why is my resting heart rate higher the morning after a late workout?

Because a hard session leaves your nervous system running higher for a while afterwards. In a study covering nearly 4.1 million nights, later and harder evening sessions were linked to both a higher resting heart rate and lower heart rate variability overnight. Choosing maximal over light exercise went with an overnight resting heart rate 3.9 beats per minute higher when the session ended two hours before habitual sleep onset, and 9.4 beats higher when it ended two hours after. A rise after a hard or late session is therefore expected. It is a reason to keep today's session easy, not a reason to worry.

Is a 5am workout good or bad for sleep?

It depends where the hour comes from. Move your bedtime earlier by the same amount and you lose no sleep at all, and the early session is one of the most durable ways to train around a busy family life. Keep the same bedtime and shorten the night from the back end, and you land in precisely the pattern that tracked with consistently worse performance in the research. The session is a trade, not a free lunch.