The run everyone argues about
Ask a group of marathoners how long their longest run should be and the answers arrive fast: 20 miles, 32 km, never more than three hours, at least one 35 km run, as many 30 km runs as you can absorb. The long run carries more folklore than any other session in a marathon block, partly because it is the one that feels most like the race.
Some perspective before the numbers. For nearly all runners, the biggest improvement comes from running more, consistently, across many months. The exact length and pace of the Sunday long run is a second-order detail sitting on top of that weekly volume. It is still worth looking carefully at what the evidence can and cannot support, because a sensible long run is easier to repeat week after week. It also helps to know who the evidence is about, because a 32 km run is a very different session for a 3-hour marathoner than for a four-and-a-half-hour one.
Key takeaways
- In a study of recreational runners at the 2017 Rotterdam Marathon, who finished in 4 hours 18 minutes on average, a longest training run under 25 km was associated with a finish roughly 13 minutes slower than a longest run of 30 to 35 km. The study is observational, and that figure should not be read as a reason to chase a particular long-run distance.
- In the same model, weekly volume had associations at least as large. The most defensible reading is that more running overall helps, and that the longest-run number may largely be a marker of that volume. These runners typically trained around 40 to 44 km a week.
- An analysis of 92 published sub-elite marathon plans found a peak-week long run of about 30 to 32 km across volume groups. That describes what the plans prescribe, not what works. Only 13 of the 92 stated a target time, ranging from 3 to 5 hours.
- The three-hour cap on long runs is a widely repeated coaching guideline. We found no study that tests it directly. For runners around the Rotterdam group's typical training pace of 5:40 per kilometre, a 32 km run already lasts about three hours, so the cap implies a shorter longest run for slower runners.
- Most plans place the long run at easy, aerobic intensity, with faster race-specific work kept to a smaller share of the week.
What the research says about distance
A 13-minute association, examined carefully
A 2020 study by Fokkema and colleagues, published in the Scandinavian Journal of Medicine & Science in Sports, is the main recent source of numbers on this question. The marathon group was 441 recreational runners entered in the 2017 Rotterdam Marathon. About seven in ten were men, their average age was 41, and they had a median of five years' running experience, with most between three and nine years. In the weeks before the race they typically trained three or four times a week, at around 5:40 per kilometre, averaging 44 km a week (the median was 40 km). Of those 441 runners, 341 finished, in 4 hours 18 minutes on average, with a standard deviation of 37 minutes. The finish-time findings below come from that group of finishers.
They grouped the longest run into four bands: under 25 km, 25 to 30 km, 30 to 35 km (the reference group), and over 35 km. Runners whose longest run was under 25 km finished about 13.4 minutes slower than the 30 to 35 km group, a statistically significant association in a multivariable model that also included weekly volume, training pace, sex, age, body mass index, running experience and recent injury. The 25 to 30 km group finished about 6.4 minutes slower, but the confidence interval just crossed zero, so that gap is uncertain. Runners who went beyond 35 km were not clearly faster than the reference group, though that group was small (38 of the 441). In the same model, compared with runners doing 40 to 65 km a week, those under 40 km finished about 6 minutes slower and those over 65 km about 14 minutes faster. Neither training volume nor longest run was associated with more running injuries in this sample.
Those numbers are associations, and they need to be questioned. The study is observational, so it cannot show that lengthening the long run caused a faster finish. Training was self-reported, on a questionnaire sent two weeks before the race, asking about the previous month's training and the single longest run completed before race day. That longest-run figure is one number. The study did not record how often runners did long runs, only the longest one they reported.
Weekly volume was in the same model, but only as three broad bands: under 40 km, 40 to 65 km, and over 65 km. A runner averaging 41 km a week and one averaging 64 km fall in the same band, so most of the volume difference between them is left unadjusted. At about 44 km a week, a 30 to 35 km long run is already roughly three-quarters of a typical week, so inside those bands long-run length and weekly volume can still travel together. The 13.4-minute gap is what remains after that fairly coarse adjustment, not a clean effect of long-run length.
The authors themselves note that the associations may be confounded by runners' intrinsic speed, or talent: faster runners may simply train longer. In a study like this, the people doing the longer long runs may already have been the fitter ones. The finish-time analysis also included only the 341 runners who finished, out of 441 in the marathon group. The paper does not report how the 100 who did not start or finish had trained, so we cannot see whether shorter long runs were more common among them. All of this is from one recreational field, at one race.
The most defensible reading is that more running overall helps, and that the longest-run number may largely be a marker of that weekly volume. The association with running more than 65 km a week, about 14 minutes faster than 40 to 65 km in the same model, is the more stable signal. None of this is a reason to treat 13 minutes, or a particular long-run distance, as a target to chase. The study also cannot tell us whether the same pattern would appear in a much faster or much slower group.
What published plans prescribe
Knopp and colleagues took a different angle in Sports Medicine – Open in 2024. They coded 92 publicly available sub-elite marathon plans, meaning popular schedules aimed at recreational runners rather than elite programmes, grouped them by peak-week volume (under 65 km, 65 to 90 km, and over 90 km a week) and analysed the last 12 weeks before race day. This describes what the plans prescribe, not what works, and the authors are clear that the plans come with no evidence of effectiveness. They grouped plans by volume because the beginner, intermediate and advanced labels the plans often give themselves were inconsistent. Only 13 of the 92 plans stated a target time. Among those that did, the two high-volume plans aimed at 3:00 to 3:30 (median 3:15), the eight middle-volume plans at 3:00 to 4:30 (median 3:52), and the three low-volume plans at 4:00 to 5:00 (median 4:30).
Across those 12 weeks, the average weekly long run was 27.4 km in high-volume plans, 23.0 km in middle-volume plans and 19.9 km in low-volume plans. The single longest run averaged 35.2 km, 32.5 km and 30.9 km respectively. In peak week the long run converged at about 30 to 32 km in every group, which the authors read as a broad consensus that the longest marathon training run sits around 30 to 32 km, largely independent of weekly volume.
That consensus is a generic prescription, and it shows. A runner on 40 km a week and a runner on 110 km a week end up with roughly the same peak long run, which makes it a very different proportion of each runner's week, and a very different time on feet if one is training for around 3 hours and the other for around 4 and a half. kaizen works from the running you actually do. It is hyper-personalised, using the overall intensity of your real runs, from their pace and elevation, to set your training target for the week, and it recalculates as you run, so the long run is counted inside a total that already reflects your current training.
The broader picture from a meta-analysis
Knopp's paper also summarises a 2020 meta-analysis by Doherty and colleagues, which pooled 127 cohorts of marathon runners. Average weekly distance, the number of runs of 32 km or more, and the longest run completed were each associated with faster finish times, but those measures are strongly linked: runners with long longest-runs also tend to run more often and further each week. Knopp's summary does not describe the performance level of the cohorts, so this should be read as a general pattern, not a finding pinned to one finish-time band.
The three-hour question
Many coaches cap the long run at around three hours, whatever distance that covers. The usual reasoning is that beyond a certain duration the fatigue and recovery cost keeps rising while the extra aerobic benefit flattens out, so a four-and-a-half-hour marathoner covering 32 km may spend far longer on their feet than a 2:45 runner covering the same distance.
The logic is plausible. At the Rotterdam group's typical training pace of about 5:40 per kilometre, a 32 km run already lasts around three hours. For a faster runner, the same distance is done with time to spare; for a slower runner it runs well past three hours, so a three-hour cap means a shorter longest run. We could not find a controlled study that compares the two, so the cap remains a coaching guideline with a sensible rationale. The same Rotterdam study did not find a clear difference for longest runs over 35 km, though only 38 runners were in that group, and it did not look at time on feet.
A fixed cap, like a fixed distance, is a single number applied to every runner. Because kaizen counts each run by the intensity it really carried, a slow three-hour run over hills and a quicker 30 km on the flat both feed the weekly target according to the load they produced.
How hard should the long run be?
Mostly easy
Knopp's analysis placed the long run in its zone 2, "extensive endurance", described in the paper's zone table at about 73 to 80% of maximum heart rate and a light effort of 3 to 4 out of 10. Across all plan volumes, roughly two-thirds of weekly distance sat in that zone, with marathon pace and tempo running grouped into a smaller zone 3 share. In other words, published plans treat the long run as the backbone of easy volume, and they place most race-specific intensity elsewhere or in limited segments.
There is a practical reason for this beyond physiology. A long run done at an easy effort is one you can recover from in a day or two, which protects the rest of the week. A long run that turns into a hard race-effort grind can cost the following sessions, and the weekly total is what sits on the more stable side of this evidence.
Where faster segments fit
Some plans add marathon-pace segments to a portion of long runs later in a block, as rehearsal for race-day pacing and fuelling. We found no study in the sources reviewed here that isolates their effect on recreational marathon times, so their value rests mainly on coaching experience and the logic of specificity. If you include them, the plan-level data suggests keeping them as a minority of the week's running.
Sources and how sure we can be
The distance figures come from two papers read in full: Fokkema et al. (2020, Scand J Med Sci Sports, 441 recreational Rotterdam marathoners, 341 finishers) and Knopp et al. (2024, Sports Med Open, 92 plans). Both are open access. The Doherty et al. (2020) meta-analysis is cited only through Knopp's summary; we did not read it in full. Fokkema's training data was self-reported, on a questionnaire sent two weeks before the race about the previous month, and the finish-time analysis is observational. The 13.4-minute figure comes from a model that included weekly volume only in three broad bands, along with training pace, sex, age, body mass index, running experience and recent injury. The authors flag possible confounding by intrinsic speed. The study recorded a single longest run, not how often long runs were done, and it does not report the training of the 100 runners who did not start or finish. Knopp describes what plans prescribe, not what works, and only 13 of the 92 plans stated a target time. We found no controlled trial on the three-hour cap or on race-pace segments inside long runs. Zone percentages vary between models, so treat every intensity figure here as approximate.
The long run in the bigger picture
For recreational marathoners in the range this evidence actually covers, people finishing around 4 hours 18 minutes, and published plans whose stated goals, where they exist, stretch from 3 to 5 hours, the plans tend to put the longest run somewhere around 30 km, mostly at an easy effort. That is a description of common practice, and the Rotterdam associations are too shaky to treat a particular long-run distance as a target. The more defensible signal is the week itself. For almost every marathoner, building a weekly volume you can sustain, and keeping it going block after block, will do more than fine-tuning the length or pace of Sunday's run.
See how your long runs add up in kaizen
If you have spent a few Saturday evenings deciding between 30 and 32 km, you clearly care about getting this right. kaizen calculates your fitness from your recent running and past races, then sets the weekly training target your marathon goal requires. Your long run counts by its real pace and elevation, alongside every easy run and session, and the target recalculates as you run. You decide how long and how hard Sunday is; kaizen shows you how it fits the week. Open kaizen on iOS or Android and see your fitness.





