You slept four hours, maybe five, with the last stretch spent staring at the ceiling doing the math on how tired you'll be. The group ride still leaves at 6 a.m. By the second climb your legs feel like they're full of wet sand — heavier, slower, unwilling. You brace for a bad number on the file.

The bad number doesn't show up.

That mismatch is the whole story of a 2025 trial: put trained cyclists through a full night of zero sleep, then back on the bike the next day, and the ride feels dramatically worse — but for a fixed, known effort, the watts mostly don't move (Gattoni et al. 2025). What a bad night changes isn't your engine. It's the dashboard reading the engine.

Section 01What actually happens after one bad night

Twenty-six male amateur cyclists and triathletes (age 30.5 ± 8.8 years, VO2peak 55.3 ± 4.9 ml/kg/min) were split into a sleep-deprivation group and a control group and tested across three consecutive days. Day one was baseline. On day two, the sleep-deprivation group stayed awake for 25 hours straight while the control group slept normally. On day three, everyone slept a normal night, and both groups were retested. Each test day used the same two efforts: a 40-minute constant-load ride at a moderate intensity, then a 20-minute time trial.

During the 40-minute ride, RPE (p = 0.023) and Feeling Scale mood scores (p = 0.013) were both significantly worse after the sleepless night than after normal sleep. Heart rate and blood lactate during that same ride showed no significant change, and the 20-minute time trial that followed showed no significant sleep-related drop in work output — the power was still there (Gattoni et al. 2025). One further night of ordinary sleep was enough to bring the perceptual measures back toward baseline.

So the ride felt worse, cyclists said they felt worse, and the number on the head unit barely blinked.

Section 02Why the numbers don't match the feeling

This isn't a fluke of one study design — it's a known feature of how effort gets regulated, and it has a name: the psychobiological model of endurance performance. The short version is that perceived effort, not the muscle itself, is very often what decides when an all-out effort ends.

The clearest demonstration of this predates Gattoni's cycling trial by over a decade. In a randomized crossover study, 16 subjects cycled to exhaustion at 80% of peak power after either 90 minutes of a demanding cognitive task (mental fatigue) or 90 minutes of watching neutral documentaries (control). After the mental-fatigue treatment, riders lasted just 640 ± 316 seconds before quitting, versus 754 ± 339 seconds after the relaxed control session — about 15% less time (p = 0.003) — even though heart rate, blood lactate, and muscle activation stayed largely unaffected between conditions. What did move was perceived effort: riders in the mentally fatigued state rated the same physical task as significantly harder (p = 0.007), and because RPE climbed at a similar rate in both conditions over time, the fatigued group simply hit their perceived-exertion ceiling sooner and stopped (Marcora et al. 2009).

That's the mechanism. And it explains something that looks, at first glance, like a contradiction between Gattoni's cycling data and a similar-sounding sleep study. In a separate trial, twelve active men who stayed awake all night lasted 1137 ± 253 seconds in an open-ended cycling time-to-exhaustion test, versus 1236 ± 282 seconds after a normal night — about 8% less time (p = 0.013). Muscle activation fatigued by roughly the same amount in both conditions — a peripheral drop of about 7% (p = 0.003) and a cortical trend of about 5% (p = 0.059), both driven by the exercise itself rather than by the missed sleep. Sleep deprivation added no extra measurable neuromuscular fatigue on top of that, which is why the authors concluded the earlier quitting point wasn't explained by the muscle or the nervous system giving out any faster (Temesi et al. 2013).

Put the two sleep studies side by side and the pattern is consistent, not contradictory. It comes down to what kind of task you give someone. A fixed 20-minute time trial is a closed-loop task — a rider paces against a known distance and a known endpoint, and effort gets rationed across that endpoint the same way regardless of how rough the perceptual signal is running. An open-ended ride-to-exhaustion test is an open-loop task with no fixed endpoint — the rider just goes until perceived effort maxes out, and a sleepless night makes that ceiling arrive sooner without the muscles actually failing any faster.

Section 03The protocol for a bad-night ride

1. Keep the number, drop the story.

   If the session calls for 250 watts, hold 250 watts. A rough night makes 250 watts feel like 280. The data says it's still 250.
2. Pace to output on fixed efforts, to feel on open ones.

   A 20-minute time trial or a prescribed interval set has a known endpoint — ride the number, not the story your legs are telling you. An unstructured all-day ride or a race decided by a late, unplanned move behaves more like an open-loop task — treat that one with more caution on a short-sleep day.
3. Expect the RPE mismatch — don't fight it.

   Perceived effort runs a gear or two hot on zero sleep. That's the documented pattern, not a sign that something's wrong with your legs.
4. Sleep once, normally, and move on.

   A single night of ordinary sleep was enough to bring mood and effort perception back toward baseline in Gattoni's cohort. One rough night doesn't need a week of banked recovery to pay off.

Section 04Where this gets misread

Section 05Applying it with HELIOS

HELIOS's skin-temperature and HRV sensors pick up a short night the same way a lab would — as a recovery flag by morning. ULTRA doesn't use that flag to pre-emptively soften the day's power targets. The Gattoni data is the reason why: for a single acute night, the damage shows up in perception and mood, not in the muscle's ability to produce watts. So the daily call holds the session as planned and tells you to expect it to feel harder — instead of guessing at a discount that the physiology doesn't actually ask for.

Section 06Counterpoint: when a bad night actually does bite

Every part of this needs a boundary drawn around it. Gattoni's result comes from a single acute night of total deprivation paired with a short, fixed 20-minute effort — close to a best-case scenario for "the power's still there." Real-world sleep loss rarely arrives as one clean sleepless night; it more often compounds across a training block or a travel-heavy race week. The broader review literature is honest that this more realistic scenario — partial restriction spread over several nights — is less studied and less clear-cut than the acute, total-deprivation picture. What sleep loss reliably does impair, per the wider literature, is cognitive function — slower, less accurate performance — reason enough on its own not to treat repeated short nights as a non-issue just because one acute night held up here (Fullagar et al. 2015).

The type of effort matters too. A fixed-duration effort lets a rider pace against a known endpoint regardless of how rough perceived effort feels. An open-ended effort ridden until failure doesn't have that anchor — and that's exactly the scenario where a night of total deprivation cost real time in the lab, not just a worse feeling: 1137 ± 253 seconds versus 1236 ± 282 seconds, about 8% less (p = 0.013) (Temesi et al. 2013). An unstructured all-day ride or a race decided by a late, unplanned move sits closer to that open-ended category than to a paced time trial, and deserves more caution on a short-sleep day. There's a sharper edge underneath all of it, too: if perceived effort is running hot and unreliable on a bad night, the riders most likely to misjudge that day badly are the ones pacing by feel instead of by a number.

Section 07Bottom line

One sleepless night before a key session or a short race is not, by itself, a reason to rewrite your targets — the data says the watts are very likely still in your legs, even though the ride will feel considerably harder and your mood will take a real hit. Pace to the number, not the story, and treat one ordinary night of recovery sleep as enough to reset the perceptual damage.

This is a single-night, single-effort finding, though — not a licence to ignore sleep. Chronic restriction and long, feel-paced events are a different and tougher problem, and they deserve a different plan.