How to Get Back Into Running After a Break Without Getting Hurt

Start from the volume you ran last month, not the volume you ran before the break, and grow it under three checks: no single run more than about 10% longer than your longest of the past 30 days, no week more than about 30% above your 4-week average, and a monthly growth limit you set yourself. The first two come from cohorts of thousands of runners; the third is a coaching convention, and this guide explains why you need it anyway. All three are drawn as markers on the weekly chart in Trainings on Endurance Analytics. The most common injury pattern in those cohorts is not a wild week. It is a normal week that comes too soon after a gap.

Why your old normal is now a spike

A previous injury is the strongest predictor of the next one. In 706 Dutch recreational runners it was the only factor that predicted a new injury, odds ratio 1.7, and nothing in the training log added to it (Dallinga et al. 2019); reviews across 29 and 73 cohorts agree (van Poppel et al. 2021, Hulme et al. 2017). Since 20 to 40% of runners get hurt within one training block, "previously injured" describes most of us within a few years.

The load does not change; the tissue's tolerance does. In 517 recreational runners the same weekly volume carried a higher hazard in those who had been hurt before (Malisoux et al. 2015). Injury happens when the load on a structure exceeds the capacity it brought into the session, and capacity drops during an injury or a layoff (Bertelsen et al. 2017, Kalkhoven 2026). Your chart shows load. It cannot show capacity.

Fitness fades on its own clock too, and a different one for each system: VO2max drops about 7% in three weeks off (Coyle et al. 1984), while the quads' protection from one downhill run is gone by about nine (Byrnes et al. 1985). Our guide to types of runs lists how fast each one fades and what brings it back.

The gap itself sets the trap. New York City Marathon runners injured in the third month of their build had run less in the second month than those who stayed healthy, 44.6 vs 51.3 km (27.7 vs 31.9 mi) a week (Toresdahl et al. 2023); ultramarathoners with less distance in the 12 weeks before Comrades were more likely to be injured at the race (Burgess et al. 2025); Dutch runners with a low acute-to-chronic ratio had a higher predicted injury probability than those with a high one (Nakaoka et al. 2021). Under-train, then catch up: resume at the old mileage, get the old injury back.

Many also return before the tissue is finished: in a Danish trial the median time to a pain-free area was 56 days, shin injuries took 70, and half of the injured were not recovered after 24 weeks (Mulvad et al. 2018). Kneecap, shin, plantar fascia, iliotibial band and Achilles problems recur rather than happen once (Kakouris et al. 2021).

How much can you increase running mileage per week?

No weekly percentage is proven safe, and the strongest injury signal is not weekly at all. The 10% rule failed its only randomized test, a systematic review found no reproducible threshold, and in marathoners tracked on Strava exceeding the rule was not linked to injury. The one weekly number with support is 30%, and the support is thin: in 874 novices, a week more than 30% above the previous one carried a hazard ratio of 1.59 for distance-type injuries (kneecap, iliotibial band, shins), but the confidence interval touched 1 and overall injuries did not differ. A convention, not a law.

The number that matters after a break is per run. In 5,205 runners over 18 months, the week-to-week ratio showed no relationship with injury; a single run longer than the runner's longest of the previous 30 days did, and the risk rose with the size of the jump (Frandsen et al. 2025). After four weeks off, your "longest run in 30 days" is short, and your first ordinary long run doubles it.

Trained runners confirm it from the other side. In a six-month log of 26 of them, the median biggest weekly jump was 30%, only 2 of 26 stayed within 10%, and 58% put a down week between increases (Johansson 2020, coaching source). A spike is survivable when a rest week follows it.

The rule itself comes from a 1980 book for novices, and no trial has backed it since (Johansson, coaching source). Coaches who write plans mostly do not use it: Jack Daniels adds no more miles a week than the number of days you run, at most 16 km (10 mi), then holds 3 to 4 weeks (Daniels' Running Formula, coaching source); UESCA lets the percentage fall as the base rises and spreads the increase across the week (Prince, Outside, coaching source). The rule survives because it is easy and accidentally protects beginners: at 20 km (12 mi) a week, 10% is 2 km; at 120 km (75 mi) it is 12 km, more than most long runs.

Why 10% a week is not 10% a month

Percentages compound, and most arguments about safe progression are between people quoting the same growth at different horizons. Geometric growth, rounded:

Two things follow. No weekly percentage is a plan: 10% a week for a year multiplies your volume by 142, so everyone who "follows the rule" follows it for a few weeks, then holds or cuts back. And the study numbers are caps on one step, not growth rates: Nielsen's 30% is one week against the one before, Frandsen's 110% is one run against your recent longest. Neither says what your average should do over three months, which is the question after a break, and the one no platform answers.

Three rates that sound alike over one year: +10% a week leaves the chart in 12 weeks, +20% per 4-week block (our default band) triples volume in half a year, +25% a year (an experienced runner's ceiling) is the nearly flat line.

Beginner vs experienced: same rule, different numbers

A beginner going from two runs a week to four in a year doubles their volume, and that is 5.5% per 4-week block, inside any reasonable limit. Coaches set the limit by training age: the Uphill Athlete method lets beginners under about 350 to 400 hours a year add up to 25% a year, caps advanced athletes above 500 hours at about 10%, and expects elites to add nothing and redistribute intensity (Johnston, Training for the Uphill Athlete, coaching source). Those are yearly figures; inside a build the same coaches step volume up 10 to 20% per block, then hold.

What changes with experience is what you add. Beginners add a run day before they lengthen a run, so the long run never becomes the spike. Jason Koop argues that a little every week works for beginners, who adapt fast, and fails for experienced ultrarunners, whose fatigue accumulates over weeks, so he puts the hardest sessions first in a block (Koop, CTS, coaching source). Experience protects nobody, though: in the New York cohort first-timers and veterans were injured at the same rate (Toresdahl, press summary). It changes the numbers you can absorb, not the fact that a jump against your own recent base is what hurts.

Does the acute-to-chronic workload ratio predict injuries?

Not in runners, not as a number. The ratio (last 7 days against the last 28) and its 0.8 to 1.3 "sweet spot" come from team sports, the method manufactures its own sweet spot, and in runners the associations are small, inconsistent, once reversed, and depend on which formula the app uses (Gabbett 2016, Impellizzeri et al. 2020, Toresdahl, de Vos et al. 2026, Dijkhuis et al. 2020, Ding et al. 2026, Cloosterman et al. 2024). Its honest use is the same as any spike marker, ours included: "well above what you have been doing", a prompt to look at the week, never a risk percentage. The one consistent signal in runners is a single run far above your recent longest (Frandsen), which is why our chart marks the day as well as the week.

For overtraining it says even less: one experiment, and the syndrome is diagnosed by performance that stays down after rest, not by a chart (Lehmann et al. 1991, Meeusen et al. 2013). A rising load is a reason to watch recovery, not a diagnosis, and a marker that shouts gets switched off, which is what runners do to watches that call "overreaching" on the plan they prescribed. A band you set yourself, wide enough to let a planned build through, is the alternative; the recovery side, and how the watch platforms compare, is in our guide to the signs of overtraining.

Reading the weekly load chart: three markers, three time scales

This is how the three checks look on the weekly chart of the Trainings page, for whichever volume you track: time, distance, elevation gain, flat-equivalent kilometers, load points or sessions (click the card title to switch). One bar per week, split into its days, and a smoothed 4-week average. The ⋮ menu sets the period, the step, bars or line, the average, and the Overload markers, which are drawn on the one-week step for volume metrics and are on by default. The dashed ceiling line is off by default, because the pink bar already says what it would. Click a week to see its value, average and ceiling ("max") in the header and its sessions below the chart. The chart below is drawn on synthetic data.

The red mark: one day against your last four weeks. A red mark above a bar, with that day's segment in red, means a day that week was more than 10% bigger than your biggest day of the previous 28 days: Frandsen's marker, the one with the strongest evidence. It reads whole days, so a double day counts once. After a gap it fires on your first normal long run, because your recent longest is short. That is the point: split the first long run, or grow it from where you actually are.

The ceiling: this week against your base. The ceiling is 30% above the average of the four weeks before, Nielsen's cutoff against your own recent base rather than against last week. A bar above it turns pink, a spike week; the line itself is hidden by default, and the menu can show it.

The band: this month against last month. The band is drawn around the rolling average as it stood four weeks ago, plus and minus 20% by default. The average line turns red only when it rises above the band; below it stays neutral, because a taper, a recovery week or a cold is not a failure. It exists for the one pattern every weekly rule misses: the climb out of a hole where each week is "only" 25% more than the last. The menu offers 10, 15, 20 and 30%, a horizon of 4 weeks, 12 weeks or a year, and shows the same percentage compounded to the other horizons: ±20% per 4 weeks is +73% per 12 weeks and ×10.7 per year.

A return to running program that reads the chart

Whatever caused the gap, the program has the same shape: find where you are, not where you were, and grow from there.

• Start from your recent baseline. Physiotherapists typically set the first week 10 to 20% below the longest pain-free run you can do now, change one variable at a time, keep a rest day between runs, and progress only when the previous step felt fine (Goom, Running Physio, clinical source). For a bone or a tendon that protocol is the floor, not the plan: tendons settled slowest in the Danish trial (Mulvad), and bone stress injuries reload on a timeline set by energy availability as much as by the scan (Hoenig et al. 2022), so get it from a clinician. The Uphill Athlete rule is stricter still: miss more than about 5% of a month's planned volume and you step back in the plan rather than catch up (Johnston, coaching source).
• Rebuild the long run before the week. Your 30-day longest resets after a gap, so grow the long run in steps of 10% or less from your recent longest, not from memory (Frandsen). Keep it under about 30% of the week by convention; ultra coaches split a big weekend into two back-to-back long days with at least two easy days after (Roche, coaching source).
• Steps and holds, not a ramp. Add a step, hold it 3 to 4 weeks, and insert a lighter week every 3 to 5 weeks with 15 to 25% less volume and the long run cut by at least as much (coaching conventions). A hold is what keeps the average inside the band.
• One axis at a time. Trail runners' own rule: do not add elevation gain and distance in the same week, and do not add intensity in the week you add volume. The iliotibial band is the usual price of too much vert too fast. Rebuild descents separately: see hill training for trail runners and why descents wreck quads.
• Keep hard days hard and easy days easy. Seven similar days score high on Foster's monotony, and load multiplied by monotony predicted illness better than load alone (Foster 1998). A real rest day lowers it arithmetically.

Key takeaways

• The most common injury pattern is a return to the old weekly volume after a gap. After a break or an injury, the same load is a bigger load for the tissue.
• No weekly percentage is proven safe. The 10% rule failed its randomized test; the supported limits are one run no more than ~10% over your 30-day longest and one week no more than ~30% over your recent average.
• Percentages compound: 10% a week is +46% a month and ×142 a year. Study numbers are caps on one step, not growth rates.
• Ratios like ACWR are flags, not forecasts, and different formulas flag different runners.

Sources: van Poppel et al. 2021 · Dallinga et al. 2019 · Hulme et al. 2017 · Malisoux et al. 2015 · Bertelsen et al. 2017 · Kalkhoven 2026 · Toresdahl et al. 2023 · Burgess et al. 2025 · Nakaoka et al. 2021 · Mulvad et al. 2018 · Kakouris et al. 2021 · Buist et al. 2008 · Damsted et al. 2018 · Nielsen et al. 2014 · Frandsen et al. 2025 · Dijkhuis et al. 2020 · Gabbett 2016 · Impellizzeri et al. 2020 · Cloosterman et al. 2024 · de Vos et al. 2026 · Ding et al. 2026 · Lehmann et al. 1991 · Meeusen et al. 2013 · Hoenig et al. 2022 · Foster 1998 · Coyle et al. 1984 · Byrnes et al. 1985