Strength survives the night. Skill doesn't.#
Pool 227 outcome measures from 69 publications and acute sleep loss costs roughly 7.56% of physical performance (95% CI −11.9 to −3.13)1. That is the number everyone quotes, and on its own it is close to useless: the heterogeneity statistic underneath it is I² = 98.1%, a formal way of saying those 227 results do not belong in the same bucket. The average is real arithmetic over a set of things that shouldn't be averaged.
Split them by what was being asked of the body and the picture becomes usable, because the damage is wildly uneven.
| What was tested | Outcomes | Mean change | 95% CI |
|---|---|---|---|
| Skill (accuracy, technical execution) | 13 | −20.9% | −27.0 to −14.9 |
| Strength-endurance | 9 | −9.85% | −19.6 to −0.13 |
| Anaerobic power | 58 | −6.26% | −9.10 to −3.41 |
| High-intensity interval exercise | 27 | −6.15% | −10.5 to −1.77 |
| Endurance | 22 | −5.55% | −8.12 to −2.99 |
| Speed/power endurance | 32 | −2.90% | −4.97 to −0.82 |
| Maximal strength | 66 | −2.85% | −4.47 to −1.23 |
Seven-fold, top to bottom. A maximal lift is brief, externally paced, and heavily motivated — someone counts down, you brace, it's over in three seconds — and it comes through a bad night nearly intact. A tennis serve, a free throw, a technical lift at speed asks for continuously updated motor accuracy over minutes, and it falls apart. The strength-endurance interval is worth reading rather than skimming: its upper bound sits at −0.13%, meaning nine outcomes were barely enough to separate that category from no effect at all.
The narrative review literature landed in the same place before anyone had the numbers to pool. Surveying the field, Fullagar et al. (2015) concluded that some maximal physical efforts may be maintained after sleep loss while sport-specific performance typically declines. That is the shape of the whole problem: the measurement you're most likely to take — the weight on the bar — is the one least sensitive to the thing you're trying to detect.
The same volume, moved slower, at a higher price#
The most instructive trial on this ran nine nights, not one. Ten resistance-trained women aged 18 to 35 completed two arms in randomized crossover — nine nights at five hours in bed against nine nights at seven or more, separated by a six-week washout — and lifted throughout2.
Their volume load — sets times reps times load, the number a training log records — fell by less than 1%. Trivially. If you had watched only the logbook you would have concluded that nine nights of five-hour sleep does nothing to resistance training.
Everything else disagreed. Mean concentric velocity fell by up to 15% on lower-body compound lifts, while upper-body lifts showed no significant change. Intraset velocity loss on back squats widened by up to 7%. Session perceived exertion rose 11%. Salivary cortisol rose 42%. Total training distress rose 84%.
The logbook said the training was fine. The bar speed, the cortisol, and the athletes themselves said it cost nearly twice as much to produce.
That dissociation between quantity and quality is the finding, and it explains a common experience: you drag yourself in after four hours' sleep, complete the session as written, and conclude sleep doesn't matter much for training. The completed session is genuine. So is the fact that you generated it more slowly, under more strain, and enjoyed it considerably less — and velocity is the variable that tracks adaptation in trained lifters, not tonnage. Note also which lifts moved: squats and their relatives, not the press. Whatever sleep loss takes, it takes more of it from the biggest, most coordinated efforts.
One honest limit on this trial: n = 10, all female, all trained. It is a small study with a long protocol, which is the right trade-off for this question but not a licence to treat 15% as a general constant.
Effort inflates before output falls#
Rating of perceived exertion is the most reliably moved variable in this entire literature. Across 45 studies and 670 participants, sleep deprivation raised RPE with a pooled effect of SMD 0.51 (95% CI 0.20 to 0.82)3. In athletes specifically it was SMD 0.39 (95% CI 0.11 to 0.66). Every subgroup — morning testing, afternoon testing, early-night deprivation, late-night deprivation — pushed the same direction.
That matters more than it sounds. RPE is not a soft outcome sitting beside the hard ones; in this context it is the earliest one. Work that feels harder gets abandoned sooner, cut shorter, or reduced in load at the next session — none of which shows up as a performance decrement in the session you measured. It shows up two weeks later as a training block you quietly downgraded.
The same analysis found the deficits concentrated differently by training status: athletes were significantly impaired across aerobic endurance, explosive power, maximum force, speed and skill control, while non-athletes showed the clearest hit to aerobic endurance (SMD −1.02; 95% CI −1.84 to −0.21). Being fitter does not buy you immunity here. If anything it means you have more finely tuned capacities available to lose.
Every hour awake takes about 0.4% off the top#
Sleep loss is a dose, and it has a clock attached. Craven's analysis found performance declining by roughly 0.4% for every hour awake before exercise, in both total-deprivation and late-restriction protocols. That converts a vague warning into a schedule.
It also produces the single most actionable result in the topic. Morning exercise after sleep loss came in at −5.42% (95% CI −9.66 to −1.17); afternoon and evening exercise at −8.31% (95% CI −13.2 to −3.37)1. The RPE data points the same way: perceived exertion rose more in afternoon tests (SMD 0.72) than morning ones (SMD 0.44)3.
So if you slept badly and the session is non-negotiable, move it earlier. You are not fresher at 7am in any restorative sense — you have simply accumulated fewer waking hours since the deficit began. The arithmetic of 0.4% an hour is ours to apply, but the two endpoints are measured, and they differ by about three percentage points of performance across a working day.
Which half of the night you lose matters less than the total. Both early-night and late-night restriction produced significant impairments, with late restriction the worse of the two (−7.39%; 95% CI −10.1 to −4.66). There is no clever way to take the loss.
The experiment run in reverse#
The famous result in this field goes the other direction. Eleven players on the Stanford men's varsity basketball team kept their normal schedule for two to four weeks, then spent five to seven weeks aiming for at least ten hours in bed nightly4. Objective sleep rose by 110.9 ± 79.7 minutes a night. A timed 282-foot sprint fell from 16.2 ± 0.61 seconds to 15.5 ± 0.54. Free-throw accuracy rose 9%; three-point accuracy rose 9.2%. Reaction time, sleepiness, vigor and fatigue all improved.
Notice what improved most: shooting. The same dimension that collapsed hardest under deprivation is the one that responded hardest to extension. Skill is the sleep-sensitive axis in both directions, and it is the axis nobody tracks.
The design deserves plain statement, because this study gets cited as if it were a randomized trial. It is eleven athletes, no control group, and a pre-post comparison running five to seven weeks deep into a season in which the players were also training and playing. Getting better at basketball over two months is not a competing explanation anyone can rule out here. The direction is credible and consistent with the deprivation literature; the 9% is not a number to promise anyone.
Where the missing calories actually go#
The intuitive conclusion — a tired session burns fewer calories — turns out to be the weakest claim in this article. Knowles' lifters moved essentially the same volume load, and volume load is close to a proxy for the mechanical work performed. Slower, harder, more distressed, but not measurably less work done. If you want the energy cost of a lifting session, sleep is not the variable that moves it; the numbers themselves are modest and widely overestimated with or without a good night.
Where sleep loss visibly subtracts movement is outside the gym. In fifteen healthy men, two nights of restricted sleep significantly reduced free-living daytime physical activity (P = 0.008) and shifted its intensity downward, with less time spent in intense activity (P = 0.046) — while food intake, hunger, ghrelin and leptin all stayed put6. The workout survives. The incidental movement in the other twenty-three hours is what quietly goes, which is the same accounting problem that makes exercise a smaller calorie lever than it feels like.
There is one distortion worth flagging. Perceived exertion rises after sleep loss for the same external work — so the internal signals a session feels like it deserves credit for are inflated exactly when output is flat or falling. That is a bad combination for anyone estimating burn from how hard a session felt, and a reminder that wearable calorie figures carry error in both directions.
The practical position is unglamorous. Train after a bad night if you want to — the evidence does not support skipping, and the overnight repair machinery is a separate argument from the session itself. Just train earlier, expect to grade the day on completion rather than on numbers, and don't read a maintained volume load as evidence that nothing happened. And keep the size of this in proportion: sleep's larger effect on body composition arrives through appetite, not through the gym, which is the case made in the sleep and weight loss pillar and traced calorie by calorie in how sleep loss drives weight gain.
FAQ#
Should I skip the gym after a bad night's sleep?#
Probably not. Maximal strength is the least affected capacity measured — down 2.85% (95% CI −4.47 to −1.23) across 66 outcomes1 — and nine nights of five-hour sleep barely changed training volume in trained lifters2. The session will happen. Expect it to feel about 11% harder and to move slower, and consider it a maintenance day rather than a day to test a personal record.
What kind of training suffers most when you're sleep deprived?#
Anything requiring accuracy. Skill and technical-execution outcomes fell 20.9% (95% CI −27.0 to −14.9), against 2.85% for maximal strength1. Sports with a heavy motor-precision component — shooting, serving, complex lifts performed at speed — take a far bigger hit than grinding through a set of heavy squats does.
Is it better to train in the morning or the evening when you haven't slept?#
Morning, by roughly three percentage points of performance. Post-deprivation exercise came in at −5.42% in the morning against −8.31% in the afternoon or evening, consistent with a decline of about 0.4% per hour awake1. Perceived exertion follows the same pattern, rising more in afternoon tests than morning ones3.
Sources#
- Craven J, McCartney D, Desbrow B, Sabapathy S, Bellinger P, Roberts L, Irwin C. Effects of acute sleep loss on physical performance: a systematic and meta-analytical review. Sports Med. 2022;52(11):2669-2690.
- Knowles OE, Drinkwater EJ, Roberts SSH, Alexander SE, Abbott G, Garnham A, Lamon S, Aisbett B. Sustained sleep restriction reduces resistance exercise quality and quantity in females. Med Sci Sports Exerc. 2022;54(12):2167-2177.
- Kong Y, Yu B, Guan G, Wang Y, He H. Effects of sleep deprivation on sports performance and perceived exertion in athletes and non-athletes: a systematic review and meta-analysis. Front Physiol. 2025;16:1544286.
- Mah CD, Mah KE, Kezirian EJ, Dement WC. The effects of sleep extension on the athletic performance of collegiate basketball players. Sleep. 2011;34(7):943-950.
- Fullagar HHK, Skorski S, Duffield R, Hammes D, Coutts AJ, Meyer T. Sleep and athletic performance: the effects of sleep loss on exercise performance, and physiological and cognitive responses to exercise. Sports Med. 2015;45(2):161-186.
- Schmid SM, Hallschmid M, Jauch-Chara K, Wilms B, Benedict C, Lehnert H, Born J, Schultes B. Short-term sleep loss decreases physical activity under free-living conditions but does not increase food intake under time-deprived laboratory conditions in healthy men. Am J Clin Nutr. 2009;90(6):1476-1482.
