How to Taper for a Marathon, Half Marathon or Ultra

To taper, cut your training volume by about 40–60% over the last 2 weeks before the race, and keep the intensity and the number of runs per week close to normal. That is the classic taper, and a meta-analysis of 27 studies found it the most effective version (Bosquet et al. 2007). Many elite athletes taper differently: shorter, with a smaller cut and hard sessions until a few days out. This guide covers both, what the evidence says about each, and how long to taper for a 10 km, a marathon or a 100-mile trail race.

Norwegian winter athletes keep turning up at the front of road races. At Hytteplanmila, a 10 km (6.2 mi) race on 17 October 2026, cross-country skier Therese Johaug won the women's race in 30:51 and Olympic speed skater Karoline Haugen was fourth in 31:31. Hans Christer Holund, with 4 world titles in cross-country skiing, an Olympic bronze and a relay silver, ran 29:45, his first 10 km under 30 minutes. His last 14 days before that race look nothing like the taper in most running guides, so we use them as a worked example.

What is tapering in running, and what does it buy you?

A taper is a progressive, non-linear cut in training load before a goal race, meant to clear the fatigue of training while keeping the fitness it built (Mujika and Padilla 2003). Done well, it is worth about 3% of performance, usually between 0.5 and 6%. For a 4-hour marathoner, 3% is about 7 minutes.

It works because fatigue fades faster than fitness. Each session adds a small, slow-building gain and a larger, fast-fading cost; a stretch of lighter training lets the cost drain while most of the gain stays. Our article on overtraining signs explains this fitness and fatigue model.

Here is how the dose looks in our Training Lab, on an illustrative athlete rather than a real one: months of block training, then the Lab's 2-week taper simulation on the right, with volume cut and intensity kept. Hover or tap a day to read fitness, fatigue and form.

Two ways to taper

There are 2 ways to taper that hold up: the classic 2-week taper and the shorter taper many elites use. Both reduce volume and keep intensity; they differ in how long the taper lasts, how deep the cut goes, and how close to the race the last hard session sits.

The classic taper has been tested against alternatives in trials and pooled in a meta-analysis. The short elite taper is mostly described: researchers recorded what champions did before winning. That shows the approach is compatible with peak performance, not that it beats the classic version.

How does the classic taper work?

Cut volume by 41–60% over 2 weeks, fast at first and then leveling off, and keep intensity and the number of sessions the same. That recipe comes from a meta-analysis of 27 studies (Bosquet et al. 2007); a 2023 meta-analysis of 14 studies reached the same picture (Wang et al. 2023). Tapers of 1 week or 3–4 weeks also helped, just less.

The 2-week length rests on the meta-analysis. The individual trials behind each rule used tapers of only 6–7 days:

• Intensity. In 9 trained middle-distance runners, a week of short, hard intervals at much lower volume improved time to exhaustion at 1,500 m race pace by 22%; an easy-running week gave a non-significant 6%, and a week of rest made it 3% worse (Shepley et al. 1992).
• Frequency. Runners who kept training every day through a 6-day taper ran their 800 m about 2% faster; those who rested every third day did not improve, though the groups were 5 and 4 runners and were not compared directly (Mujika et al. 2002).
• Running itself. A taper on the bike with the same interval load kept 5 km performance where it was; the running taper improved it by 3% (Houmard et al. 1994).

The largest real-world data set agrees. Across more than 158,000 recreational marathoners, a disciplined 3-week taper was worth a median 2.6% (about 5.5 minutes), and a disciplined 2-week taper 2.1% for men and 3.1% for women (Smyth and Lawlor 2021). "Disciplined" meant the reduced weeks came in a row, right before the race; the commonest way to lose the benefit was a harder week in the middle of the taper.

Depth is the more forgiving dial. In a very small trial, a 75% cut for 6 days did no worse than a 50% cut (Mujika et al. 2000), while Shepley's rest week was the only option that made runners slower.

What does a short elite taper look like?

Holund's training log for his last 14 days before Hytteplanmila is a clear example: sessions got shorter toward the end, but the hard work stayed, including 2 days with 2 threshold sessions each, and the last full session came 4 days before the race. The table follows a summary of that log by @enduranced_official, translated by the Draft Me Hard channel.

Heart rate is in percent of maximum and lactate presumably in mmol/L; 45/15 and 30/15 are not explained in the original. Treadmill speeds of 13.2–15.7 km/h are 4:33–3:49 per km (7:19–6:09 per mile), at 7% incline.

Two caveats. His weekly volume before these 14 days is unknown, so the size of the cut is unknown; this may be a normal 2 weeks with a lighter last 5 days rather than a "14-day taper". And it is one athlete, racing for 30 minutes on a training base far beyond most readers'.

He is not an outlier among elites. British runners from 3,000 m to 10,000 m tapered for a median of only 6 days, marathoners for 14; they cut easy running to about 70% of normal (53% for marathoners) and ran their fastest taper intervals at 112–114% of race speed (Spilsbury et al. 2015). A review of world-class distance runners puts their taper at 7–10 days (Haugen et al. 2022).

In 11 Olympic and world champions in cross-country skiing and biathlon, volume in the final phase was only about 32% below the preparation phase, and 10 of the 11 did a high-intensity session within 48 hours of winning gold (Tønnessen et al. 2014). They were racing and traveling in those weeks, so their taper is mixed with competition.

The dose itself has been tested once. Trained 1,500 m runners who cut volume by about 30% for 7 days, following the elite pattern, ran a 1,500 m time trial 3.4% faster (Spilsbury et al. 2019). Making the last session faster than race pace did not add anything reliable, in that trial or in a 2021 follow-up.

Three rows need a word. Marius Bakken, who popularized double threshold training, switches to single threshold sessions when racing, close to what recreational runners know as Norwegian singles, and runs a short one 2 days before his own races, so even elites split on the doubles. The heat finding comes from triathletes, where hard heat sessions during a taper caused overreaching and easy ones did not (Schmit et al. 2018); heat training also raised hemoglobin mass in elite skiers without improving performance (Rønnestad et al. 2022). The incline row comes from a treadmill study in which steeper running at the same metabolic effort lowered impact forces (Williams et al. 2020).

So the short elite taper keeps what the trials agree on, intensity and frequency, and spends less on volume reduction. Its signature pieces, late double threshold days and regular heat sessions, are either untested or point the wrong way in the one study that looked.

How do you choose, and how long should you taper?

Use the classic taper as your default, lengthen it for longer races, and move toward the short elite version only if you train high volume for short races and know how you respond to rest. Three things decide it:

• Race length. The longer the race, the more fatigue there is to clear. Recreational trail runners act on this already: in race week they ran about 37% less than in their peak week before long trail races, 54% less before medium ultras and 66% less before the longest ones (Matos et al. 2020). That describes what they did, not what worked best.
• Training volume. The elite data come from athletes who train far more than most readers. Our reading: if you run 40 km (25 mi) a week, a 30% cut over 6 days barely differs from a normal easy week.
• How you respond to rest. If a few easy days leave you flat, keep a short hard session 2–4 days out. If you arrive at races tired, taper longer. Cutting a bit deeper costs little; replacing training with rest is the mistake to avoid.

A marathon taper of 2 weeks is the meta-analysis answer; the recreational data favor 3, and published coaching plans range from about 10 days (Hansons, as summarized) to 3 weeks (Pfitzinger, as summarized). No trial has tested a taper for any race longer than a marathon.

Marathon and half marathon taper week by week

A common marathon taper keeps about 85% of peak weekly volume 3 weeks out, 70% 2 weeks out and 45% in race week, with the last marathon-pace session 6–7 days out (RunnersConnect, coaching practice). A half marathon taper scales the classic taper down to 2 weeks; the split below is our layout of it, not a tested plan.

How do you taper for a 50 km, 100 km or 100-mile trail race?

Taper 10–14 days for a 50 km (31 mi) and 2–3 weeks for 100 km (62 mi) and 100 miles, cut volume in steps that are steepest at the start, keep short bursts of effort each week, and keep running on terrain like the race's. Coaches agree on that much. They disagree on how deep the cut should go.

The deep cut. Jason Koop of CTS starts 2–3 weeks out and cuts fast: for a runner used to 10 hours a week, 4–5 hours in the first taper week, 2–3 in the second, 2 or less in race week. The number of runs stays the same, and quality shrinks rather than disappears: 4 × 10 min of tempo becomes 3 × 5 min at the same effort. Over 3 weeks that is a cut of about 70%, deeper than the meta-analysis optimum.

The shallow cut. Other trail coaches think ultra runners rest too much before big races. David Roche reports more reliable results when athletes keep a shortened long run in the last 2 weeks instead of dropping it, and Uphill Athlete cuts about 25% a week with some intensity kept.

Neither side has a trial. If you are unsure, aim for a total cut of 40–60%, which sits between them. A 2-week plan for a 100 km on a 10-hour training week could look like this:

From the short elite taper, 3 things carry over: incline treadmill or uphill repeats to keep effort high while sparing your legs some impact, an easy bike session in place of a run when your legs are beaten up, and short sharp efforts instead of long ones. Double threshold days do not carry over; no study supports them in a taper at any distance. Heat sessions belong only before a hot race, and only easy ones.

Hard downhill sessions belong before the taper; a short, easy descent 2–4 days out still helps your quads without new damage. Our guide to quad damage on downhill trail races has the timing, and hill training for trail runners covers the climbs.

What should you do in race week?

In your taper week, do the last hard session at race pace 2–4 days out, keep running most days, take at most one full rest day, and run a few short strides the day before. If this is your first trail race, our first trail race guide covers the rest of the checklist.

• Last hard session. Short, at race pace, finished feeling like you could do more. Running it faster than race pace has not proven better.
• The day before. A short run with a few strides is common practice; Uphill Athlete suggests about 5 minutes of 30 s on, 30 s off before a 50K. No study has tested it for endurance races. Our guide to types of runs explains what strides do.
• Fuel. Muscle glycogen rose during both a hard taper and a rest week in the Shepley trial. A single day at 10 g of carbohydrate per kg of body weight on a low-training day nearly doubled it (Bussau et al. 2002), but that is far more than most guts tolerate, so try any loading in training first. Glycogen is stored with water (Kreitzman et al. 1992), so a small rise on the scale is possible and harmless; the one taper study that weighed runners saw no change (Spilsbury 2019).

What is a taper tantrum?

A taper tantrum is the restlessness, phantom aches, doubt and heavy legs many runners feel in the first easy days of a taper. It is normal, and coaches like Koop name it as the main reason runners break their own taper. A pain that gets worse as you run, or that changes your stride, is not a tantrum: treat it as an injury. If your training app warns that you are "too fresh" before a race, our overtraining article explains why that reading is expected after a taper.

Key takeaways

• The classic taper (about 2 weeks, volume cut 41–60%, intensity and frequency kept) is the best-tested approach and the right default.
• Many elites taper shorter, cut about 30% and train hard until 2–4 days out; that pattern has been tested only once, at 1,500 m.
• Don't swap training for rest days (one rest day is fine); illness or pain overrides any plan.
• Taper longer for longer races: 7–10 days for a 10 km, 2–3 weeks for 100 km and 100 miles.
• Late double threshold days and hard heat sessions have no evidence for a taper.

FAQ

Do I need to taper for a half marathon? Yes, a short one. Coaches use 10–14 days for a half, and no evidence supports racing with no reduction at all.

Is 2 weeks enough to taper for a 100-mile race? For most runners, yes. Coaches use 2–3 weeks, with at least 10 days of reduced volume, and suggest the longer end after a very large build.

Can I keep double threshold sessions in my taper? Some elites do, until about 8 days out; others drop to single sessions. No study has tested it. If you train with double threshold, single, shorter threshold sessions in the last 10 days are the safer option.

Why do I feel worse during the taper? The first easy days often feel flat, and with less training stress small aches get noticed. It is common enough to have a name, the taper tantrum. Judge your legs after the race-week strides, not before.

Should I do heat training during my taper? Only if the race will be hot, and only easy sessions (Schmit 2018).

Sources: Mujika and Padilla 2003 · Bosquet et al. 2007 · Wang et al. 2023 · Smyth and Lawlor 2021 · Shepley et al. 1992 · Houmard et al. 1994 · Mujika et al. 2000 · Mujika et al. 2002 · Spilsbury et al. 2015 · Spilsbury et al. 2019 · Spilsbury et al. 2021 · Tønnessen et al. 2014 · Haugen et al. 2022 · Matos et al. 2020 · Schmit et al. 2018 · Rønnestad et al. 2022 · Williams et al. 2020 · Kreitzman et al. 1992 · Bussau et al. 2002 · Hytteplanmila 2026 results