Signs of Overtraining in Runners: What Your Training Load Can and Can't Tell You

The defining sign of overtraining is performance that stays down even after you rest. Heavy legs, poor sleep, a short temper and one cold after another are the company it keeps. Short-term overload that clears in 1–2 easy weeks is called overreaching, and it is part of normal training. Your watch's training load can warn you that fatigue is piling up, but no load number can tell you which of the two you are in.

What are the signs of overtraining in runners?

The one sign that matters is a drop in performance that persists despite rest. Everything else is supporting evidence, because each symptom alone has a dozen ordinary causes. The joint European and American consensus on overtraining defines the problem this way and diagnoses it by exclusion (Meeusen et al. 2013).

The symptoms that cluster around it:

How you feel is the earliest warning you have. A systematic review of 56 studies found that simple questions about mood, soreness, sleep and energy responded to rising training load more reliably than blood hormones, immune markers or heart rate measures (Saw et al. 2016).

Better athletes lean on feel more, not less. In a 2026 survey of 666 endurance athletes, coaches and scientists, heart rate was the most used intensity tool at every level (86% of runners), yet perceived effort rose from 55% of recreational athletes to about 75% of national-level and better ones. More than half of those higher-level respondents combined all three sources, pace or power, heart rate and feel, against 32% of recreational athletes (Seiler-Viken et al. 2026, a preprint not yet peer reviewed). A watch number is one input; how you feel is another, and the athletes who have been at it longest check one against the other.

Overreaching vs overtraining syndrome: what's the difference?

Overreaching and overtraining syndrome sit on one scale, and the difference is how long recovery takes. The consensus names 3 stages (Meeusen et al. 2013):

• Functional overreaching. Performance dips after a hard block, then rebounds within days to about 2 weeks, often above where it started. This is the goal of a training block, not a problem.
• Non-functional overreaching. The dip lasts weeks to months and there is no rebound. The training cost you more than it gave.
• Overtraining syndrome. The dip lasts months or longer, with the full symptom picture. A doctor diagnoses it after ruling out everything else.

The consensus itself calls these boundaries vague: nobody can tell on day 10 which stage you are in, only in hindsight. That is why the next question matters more than the label.

Am I overtraining or just tired?

Cut your training and watch what your performance does. Run easy, or rest, for 1–2 weeks. If your legs, mood and a familiar test effort come back within that time, you were overreached, and the block probably did its job. If 2 easy weeks change nothing, you are past functional overreaching, and it is time to see a doctor rather than to train through it (Meeusen et al. 2013).

A doctor will look for the things that produce the same picture:

• iron deficiency or anemia;
• thyroid problems;
• a lingering viral illness;
• not eating enough for the training you do (next section).

A lab protocol exists that tries to separate the stages directly (2 maximal efforts 4 hours apart, with hormones measured after each), but it is a research tool, not a clinic test. For a runner, the practical test is the easy fortnight and an honest look at what you eat.

Underfueling looks like overtraining

Eating too little for your training produces almost the same symptoms as overtraining, and it turns up again and again in overload studies. When sports scientists reviewed 21 overtraining studies that also tracked diet, 18 of them (86%) showed energy or carbohydrate intake falling during the overload (Stellingwerff et al. 2021).

The condition is called relative energy deficiency in sport (RED-S). It is measured as energy availability: the calories you eat minus the calories your training burns, per kilogram of fat-free mass. About 45 kcal per kg of fat-free mass per day is the healthy reference, and hormones start to suffer toward 30 and below (Loucks et al. 2011). The International Olympic Committee now treats RED-S as a clinical diagnosis rather than a single cut-off (Mountjoy et al. 2023); some researchers question whether it is separate from overtraining at all, but the advice for you is the same.

You can't measure your energy availability at home with any accuracy: food logs, body-composition estimates and exercise calorie counts each carry errors, and they compound (Burke et al. 2018). Watch for the signals instead. In elite distance runners, 37% of women had lost their periods and 40% of men had low testosterone, and bone injuries were about 4.5 times more common in the women without periods (Heikura et al. 2018). Lost periods, a stress fracture or unexplained weight loss: see a doctor now, don't wait out the easy fortnight.

Trail and ultra runners are especially exposed. In a survey of 1,899 trail runners, 43% screened at risk of low energy availability, and about half fell short of carbohydrate guidelines in events over 2.5 hours (Henninger et al. 2024).

Carbohydrate is the lever with evidence behind it, though the studies are small. Runners on 8.5 g of carbohydrate per kg per day held their 16 km (10 mi) time through 11 days of hard training, while on 5.4 g it slowed by 8.2% and their mood suffered more (Achten et al. 2004); a later cycling study found no protection at all (Svendsen et al. 2016). Treat carbohydrate as a buffer, not a shield. Fueling during long sessions matters too: elite runners taking 120 g per hour in a mountain marathon had less muscle damage the next day than those taking 60 or 90 g (Viribay et al. 2020). They were gut-trained elites; don't copy 120 g without training your gut first.

Protein is a recovery baseline, not a common cause of overtraining: surveyed ultrarunners met their protein needs while falling short on carbohydrate (Craddock et al. 2025). The reference numbers:

Garmin Training Status and other watch metrics: what can training load tell you?

Training load measures the stress you put in, not how your body is responding to it. Every platform turns your sessions into a load number from heart rate, pace or power, then compares recent load with what you are used to. That comparison can flag a spike or a deep fatigue hole. It cannot see whether you are adapting, which is the only thing that separates overreaching from overtraining (Halson 2014).

Two families of models sit behind the numbers, and their scales don't translate into each other:

The numbers come from each platform's own documentation; the form zones used by TrainingPeaks and intervals.icu come from coach Joe Friel's experience, not from tested thresholds.

None of these consumer metrics has been independently validated against overtraining, injury or performance in published research. The closest evidence is for a heart-rate load weighted by each runner's own lactate profile, which tracked performance in a small group of runners; the generic version your watch computes has no such data (Manzi et al. 2009).

Read the labels for what they are. Garmin's Training Status compares your 7-day load with the trend in its estimate of maximal oxygen uptake (VO2max); "Unproductive" means load went up while the estimate stayed flat. Often the estimate is lagging, not your training failing. On trails the numbers miss something else: pace-based load ignores what descents do to your legs, a blind spot ultra coaches flag openly (Koop); our guide to downhill running muscle damage covers that dose on its own.

Fitness, fatigue and form, explained

Fitness is your long-term average load, fatigue is your short-term average, and form is the gap between them. The classic Banister impulse-response model treats every session as 2 effects: a small, slow one that builds fitness and a larger, fast one that builds fatigue (Morton et al. 1990). Most platforms use 42 days for fitness and 7 days for fatigue by convention. Because fatigue fades faster than fitness, a well-timed easy stretch leaves you fitter and fresh at once. That is the whole logic of a taper.

Fatigue on its own means nothing. A fatigue of 70 is a heavy week for a runner whose fitness is 40 and a normal one at 90. The signal is always relative, which is why form reads best as a percentage of fitness: a form of −30 at fitness 120 is a different situation from −30 at fitness 40.

Read form as a comparison. It tells you how much lighter or heavier your last week was than what you are used to over about 6 weeks: the same idea as a load ratio, only day by day. At −30% your last week ran about 1.3 times your habit; at 0 it was a usual week; +30% is roughly what 3 days of full rest do. Form is not your fitness level, and it does not say how fast you can run today.

That is why a steady block reads as zero, and zero is not a standstill. Repeat the same easy week and fitness and fatigue settle on the same number, so form hovers around 0, while you keep adapting: mitochondrial gains grow with how often you train (Mølmen et al. 2025), and fitness holds even on a third less volume as long as intensity stays (Spiering et al. 2021). Our guide to the types of runs covers what each session adds. In our model a hard session costs more fatigue than it builds fitness, so a steady week with one quality day sits around −10% and one with two around −15%: working load, not a warning.

Here is the Form card from our Training Lab on an illustrative athlete, not a real one: 16 weeks of 3-week builds with a lighter fourth week, one oversized week straight after a recovery week, then the Lab's 2-week taper simulation to race day. Tap or hover a day to read the numbers.

In ordinary build weeks form sits in the working-load band and dips below −30% only on long-run days; the recovery weeks lift it to about −10%, and to zero on the rest day that follows. After the oversized week it falls to about −55% and stays mostly below −30% for 2 more weeks, until the next recovery week: the pattern that deserves a closer look at your sleep, mood and heart rate. The simulated taper ends at about +35% on race morning. Treat that number as a readout, not a score. Studies have tested the dose of a taper, about 2 weeks with volume cut by 41–60% and intensity kept (Bosquet et al. 2007; Wang et al. 2023), and none has found a best form value for race day.

Our Training Lab splits form into 5 bands:

The lower bounds are Friel's. We moved the top of the fresh band from his +25 to +50%: his zones are in points for an athlete at about 100 a day, and as a share of fitness a standard taper lands around +40%. A long rest costs little at first; aerobic fitness drops noticeably only after 2 to 3 weeks, with VO2max down about 7% at 3 weeks (Coyle et al. 1984).

Every model of this kind, ours included, uses population time constants: fitting them to one athlete takes frequent performance tests, and even then the fit is shaky (Hellard et al. 2006; Imbach et al. 2022). The bands are a map of where you are in a training cycle, not a diagnosis.

The second weak spot is how a session becomes one number. Most platforms weight harder minutes more heavily: training impulse (TRIMP), Training Stress Score (TSS) and running-power scores all do. But they ignore how the hard minutes are arranged. A 2026 modeling study showed that TSS rates 20 minutes at threshold as cheaper than 20 × 1 minute at the same power with rest between, while the model says the continuous effort is the harder one (Kontro et al. 2026).

Duration has a similar blind spot. Every mainstream score multiplies intensity by time, so an hour of threshold work counts the same whether you do it in one session or split in two. When trained runners did 6 × 10 minutes in one session, heart rate, lactate and perceived effort drifted up in the second half, and they woke up more tired and sore than after the same work split into 2 sessions 6.5 hours apart (Talsnes et al. 2024). Heart-rate-based load catches part of this on its own, because the drift raises your heart rate at the same pace; pace- and power-based scores miss it.

We built our load around the arrangement problem. Each session gives 2 numbers: a stimulus that feeds fitness and a recovery cost that feeds fatigue. Below your second threshold they are equal. Above it, the recovery cost grows faster than the stimulus as a finite reserve drains, up to 4 times the stimulus, using the reserve-balance model from critical-power research (Skiba et al. 2015). The effect: 20 minutes above threshold in one piece costs more fatigue than 20 × 1 minute with a minute's rest between. We designed this from cycling and lab data; it has not been validated on runners. Heat, altitude and the muscle damage of descents are not in it (descents get their own readiness score in the same Lab). Heart rate is the default input because it is what most runners record, but it lags behind short surges: in a lab test on skiers and cyclists it overestimated the cost of moderate stretches and underestimated 3-minute hard ones (Staunton et al. 2022), so short steep pitches on rolling trails score lower than they cost.

The arrangement problem has a practical side: most runners cross the threshold without meaning to. The coaches of Team Aker Dæhlie, a Norwegian cross-country ski team, name it the most common error they see: sessions planned and logged as threshold but run at interval effort (Team Aker Dæhlie, coaching source). Their checks are simple. Threshold is a pace you could hold for about an hour, every rep goes at the same pace, and heart rate drifts no more than 4 to 5 beats per minute across the session. A session run harder than that costs more recovery than its label says, which is exactly the work our load charges extra for; the types of runs has the details. They also put continuity first: correctly paced training kept up over seasons beats stacking long sessions, extra volume and double threshold days on top of each other.

How fast can you increase training, and does ACWR matter?

There is no proven safe percentage, and the popular 10% rule is not one. In a randomized trial of 532 novice runners, a program that raised volume by 10% a week produced no fewer injuries than a standard program that rose by about 24% a week (Buist et al. 2008). A systematic review found no reproducible threshold at all (Damsted et al. 2018).

The number that did show up is 30% week over week, and it is a weak one. In 874 novice runners, those whose distance in the last 7 days exceeded the previous 7 days by more than 30% had more distance-related injuries, such as kneecap and shin problems (hazard ratio 1.59, but with a confidence interval that touched 1 and p = 0.07), and no more injuries overall (Nielsen et al. 2014). The week ceiling on our weekly load chart borrows the 30% figure from this study as a convention, not as a proven threshold.

The largest study so far points at the single long run instead. Following 5,205 runners through 588,071 sessions, researchers found no link between the week-to-week change in volume and injury. What mattered was any single run longer than your longest of the previous 30 days: the overuse injury rate was 1.64 times higher at +10–30%, 1.52 times at +30–100%, and 2.28 times when the run more than doubled it (Frandsen et al. 2025). For ultra runners this is the long run: grow it in small steps from your recent longest. The weekly chart on our Trainings page flags a day more than 10% above your biggest day of the previous 4 weeks; the comeback guide explains how to read its markers.

And the long term? Ten percent a week for a year would multiply your volume by about 140, so nobody progresses that way, and no trial has tested what works instead. Coaching practice fills the gap: most plans build volume in steps, hold each step for a few weeks, and insert a lighter week every 3–5 weeks with volume cut by roughly 15–25% (coaching conventions).

The most dangerous week is often the one after a break. While you rested, your fitness fell, so your old volume now counts as a jump. Our weekly chart measures each week against the 4 weeks before it, not against the volume you used to run, so a return to old mileage after a pause shows up as a spike. Our guide to getting back into running after a break walks through that comeback week by week.

All three checks sit on one chart: a day far above your recent biggest, a week above 1.3 times the 4 weeks before it, and a 4-week average drifting out of its band. The comeback guide shows that chart and how to read its markers.

Two more things to know about load ratios:

• The famous 0.8–1.3 "sweet spot" comes from team sports. The acute
  workload ratio (ACWR) behind it was built on cricket and football players (Gabbett 2016). Critics showed that the ratio can manufacture its own sweet spot mathematically (Impellizzeri et al. 2020), and in Dutch recreational runners injuries were more common at low ratios, not high ones (Nakaoka et al. 2021). A spike on your chart means "well above what you are used to", not a percentage risk.
• Monotony counts as well as volume. Load multiplied by monotony, where 7 similar days have no real rest, predicted illness better than load alone (Foster 1998). Keep hard days hard and easy days easy.

All of this evidence is about injury, not overtraining. The only runner study we found that deliberately overloaded runners doubled their volume for 4 weeks and produced overreaching in 8 runners (Lehmann et al. 1991); nobody has shown that a spike on a load chart predicts overtraining in everyday training.

Does HRV show overtraining?

Not reliably, and often not in the direction you'd expect. After one hard session, heart rate variability (HRV) drops for a day or 2; that is normal. During weeks of productive overload, HRV often doesn't fall at all and can even rise. A meta-analysis of 27 studies in endurance athletes found a small increase in resting HRV during training that lowered performance (Bellenger et al. 2016). A second meta-analysis found single morning readings useless for telling overreached athletes apart, with their weekly averages running higher, not lower (Manresa-Rocamora et al. 2021).

A steady decline does appear, but later. In an elite triathlete who slid into non-functional overreaching, the 7-day average of HRV fell week after week while a teammate's stayed flat (Plews et al. 2012). So a sustained drop in your 7-day average below your own baseline is a late warning, not an early one. A raised resting heart rate is no more decisive: short overload nudges it up slightly, within normal day-to-day noise (Bosquet et al. 2008; Buchheit 2014).

Use HRV the way the research supports:

• compare it with your own baseline, never with other people or population ranges;
• read the 7-day average, not a single morning;
• put it next to how you feel and how your training is going, which remains the more sensitive signal (Saw et al. 2016).

Half marathon vs ultra: what changes?

The warning signs are the same at every distance; what changes is where the fatigue comes from and how long a race takes to clear. Road training puts more of its load above threshold; ultra training piles up volume, long runs, back-to-back days and hours of descending (see hill training for trail runners for how to dose the descents), and the race is the biggest spike of the season. Ultra runners also carry the highest risk of underfueling, which imitates overtraining.

Recovery after a race grows with its length and how hard you raced, not with kilometers or descent alone. Studies measure different things, so read this table as a rough ladder stitched together from separate groups:

Distance alone does not decide it: after the 330 km (205 mi) Tor des Géants, runners lost less strength than after a 166 km race, most likely because they paced far more conservatively (Saugy et al. 2013). Runners report feeling normal anywhere from a few days to several months after a first 100-mile race, which says more about individual variation than about any rule.

FAQ

How long does it take to recover from overtraining? Days to 2 weeks for functional overreaching, weeks to months beyond that. If 2 easy weeks bring no improvement, see a doctor.

Is overtraining syndrome permanent? Rarely, but recovery can take months to years, and it comes from rest, enough food and a gradual return.

Why does my Garmin say "Unproductive" when I feel fine? Training Status compares your load with the trend in its VO2max estimate. When the estimate lags, you get "Unproductive" without anything being wrong; it is a recurring complaint among runners. Trust your performance and how you feel over the label.

Can HRV tell me if I'm overtraining? Not on its own. During productive overload HRV often stays flat or rises; a sustained fall in your 7-day average below your own baseline is a late sign, and it needs to agree with how you feel and perform.

What is a normal acute

workload ratio? Platforms treat roughly 0.8–1.3 as the productive range. The range comes from team sports, and in runners it does not reliably predict injury. Read a high ratio as "a big jump for you".

Key takeaways

• Overtraining means performance that stays down after rest; overreaching clears in 1–2 easy weeks and is part of normal training.
• Underfueling, especially too little carbohydrate, imitates overtraining. Rule it out before you blame your training.
• Training load shows the stress you put in, not how you adapt. It can flag a spike or a fatigue hole, but it cannot diagnose overtraining.
• Most amateurs dig their hole with intensity, not volume: run threshold at a pace you could hold for an hour and keep easy days easy.
• The 10% rule is unproven. Watch your long run against your longest of the last month, build in steps with lighter weeks, and treat the weeks after a break as a comeback.
• HRV often does not fall during overload; how you feel and how you perform remain the most sensitive signals.

Sources: Meeusen et al. 2013 · Saw et al. 2016 · Seiler-Viken et al. 2026 · Halson 2014 · Stellingwerff et al. 2021 · Mountjoy et al. 2023 · Loucks et al. 2011 · Burke et al. 2018 · Heikura et al. 2018 · Henninger et al. 2024 · Achten et al. 2004 · Svendsen et al. 2016 · Viribay et al. 2020 · Craddock et al. 2025 · Thomas et al. 2016 · Kato et al. 2016 · Manzi et al. 2009 · Morton et al. 1990 · Mølmen et al. 2025 · Spiering et al. 2021 · Bosquet et al. 2007 · Wang et al. 2023 · Coyle et al. 1984 · Hellard et al. 2006 · Imbach et al. 2022 · Kontro et al. 2026 · Talsnes et al. 2024 · Staunton et al. 2022 · Team Aker Dæhlie · Skiba et al. 2015 · Buist et al. 2008 · Damsted et al. 2018 · Nielsen et al. 2014 · Frandsen et al. 2025 · Gabbett 2016 · Impellizzeri et al. 2020 · Nakaoka et al. 2021 · Foster 1998 · Lehmann et al. 1991 · Bellenger et al. 2016 · Manresa-Rocamora et al. 2021 · Plews et al. 2012 · Bosquet et al. 2008 · Buchheit 2014 · Zhou et al. 2026 · Sherman et al. 1984 · Temesi et al. 2015 · Hoffman et al. 2016 · Millet et al. 2011 · Saugy et al. 2013