Downhill running load: how to count descents in your training
Count downhill running as a number of its own next to your training stress: the distance you ran downhill, weighted by how steep it was, added up week by week and read against your usual weeks and your race. Pace and heart-rate scores give a descent little weight, because running downhill is cheap for heart and lungs (Lu et al., 2025), while your legs brake through all of it. Weighted by the square of the grade, as Pendio's Descent Load Index is, the steep minority carries most of the load: on UTMB's 2026 course file, the stretches steeper than −20 % are a fifth of the downhill distance and three fifths of its descent load.
TL;DR
- Pace and heart-rate scores count a descent by its time and its cost to heart and lungs, which is lowest between about −10 % and −20 %. The braking your legs do isn't in them.
- Pendio's Descent Load Index counts each downhill kilometre as (grade ÷ 10 %)²: 1.0 at −10 %, 4.0 at −20 %. On a steady grade, that's the height lost in kilometres times the grade in per cent.
- D− counts every metre of height alike; by the DLI, 100 m lost at −30 % weighs three times as much as 100 m lost at −10 %.
- On four organisers' course files, the stretches steeper than −20 % were 8–32 % of the downhill distance and 34–78 % of the descent load.
- Read the weekly figure against your own usual weeks. It describes the steep downhill you ran, not what it did to your legs.
Why do pace and heart rate undercount a descent?
Pace and heart-rate scores count a run by its time and by how hard your heart and lungs worked, and a descent is cheap for both. A 2025 review of uphill and downhill running finds that the energy cost of running falls as the ground tilts down, to a minimum typically between −10 % and −20 % (Lu et al., 2025).
Your legs see it the other way round. The same review describes downhill running by “the high volume of eccentric muscle contractions required for braking and shock absorption”: muscles working while they lengthen, to slow each step. At a slope of −9° (about −16 %), it reports braking force peaks 73 % and impact peaks 54 % above level running. Even at −9.1 %, six distance runners on force plates showed more braking (negative) power at the knee after foot contact than on the level (Yokozawa et al., 2005).
So training stress sees a descent through the time it takes. In Pendio’s grade-adjusted pace model, a kilometre at −10 % takes 0.87 times as long as on the flat at the same effort, and at −20 % 1.16 times (guide). At one effort, rTSS grows with that time, and the braking that makes a steep descent slow (why your pace is slower on hills) never enters it. Training load for trail running shows two descents of 1,000 m that the two numbers rank in opposite order.
What can you count from a run’s file?
Any run with elevation gives three downhill numbers: the height lost (D−, the total height a route loses, in metres), the distance or time spent descending, and the grade of each stretch. A 2020 review names a downhill run’s slope, duration and speed as the main factors in how much muscle damage it causes (Bontemps et al., 2020). Each way of counting covers some of the three:
| Count | What it adds up | Slope | Duration | Speed |
|---|---|---|---|---|
| Training stress (rTSS, hrTSS) | Time and intensity, from graded pace or heart rate | Only through time and heart rate | Yes | Yes, as intensity |
| D− | Metres of height lost | No: a metre at −5 % counts as one at −30 % | No | No |
| Downhill distance or time | Kilometres or minutes going down | No | Yes | No |
| Descent Load Index | Downhill km × (grade ÷ 10 %)² | Yes, squared | Yes, as distance | No |
Runalyze’s Impact Load Factor counts from the runner’s side: weight, speed and intensity “combine with running dynamics that include cadence, ground contact time and more”, and uphill and downhill efforts get “special consideration according to the steepness”. Its glossary gives no formula (Runalyze, checked 5 November 2026).
How do you weight a descent by its steepness?
Pendio’s Descent Load Index (DLI) counts each kilometre of descent as (grade ÷ 10 %)², so 1 km at −10 % is 1.0, at −20 % 4.0 and at −30 % 9.0, in units of its own whose weekly trend is what matters (guide). On a steady grade it simplifies to the height lost in kilometres times the grade in per cent:
| Grade | DLI per km of descent | DLI per 100 m of height lost |
|---|---|---|
| −5 % | 0.25 | 0.5 |
| −10 % | 1.0 | 1.0 |
| −15 % | 2.25 | 1.5 |
| −20 % | 4.0 | 2.0 |
| −30 % | 9.0 | 3.0 |
D− counts every 100 m of height as 100 m; the DLI weighs 100 m lost at −30 % three times as much as at −10 %. By hand, a descent of 500 m at −15 % is 0.5 × 15 = 7.5. That slightly undercounts a descent whose grade varies, because squared, a steep stretch adds more than a gentle one of the same length saves.
The square is Pendio’s design choice: the research doesn’t say how to count a descent. The 2020 review calls the literature on the load-response to downhill training “scarce”. Its studies ran downhill at a median −12 % (between −15 % and −10 % for the middle half of them) for a median 40 minutes, and it lists speeds over 15 km/h, slopes steeper than 15 % and runs longer than 60 minutes as still to be investigated (Bontemps et al., 2020). Mountain races spend much of their downhill steeper than that.
Where does a race’s descent load sit?
On mountain courses, the steepest part of the downhill carries far more than its share of the descent load. Worked out with the DLI’s formula from the distance and height lost in each of the 13 grade bands Pendio measured for its race pages, the stretches steeper than −20 % on four organisers’ files were 8–32 % of the downhill distance and 34–78 % of the descent load:
| Course (year of the organiser’s file) | D− | Downhill | Steeper than −20 % | Descent load | Share of it steeper than −20 % | Per 1,000 m of D− |
|---|---|---|---|---|---|---|
| Lavaredo 120K (2026) | ↓5,455 m | 53.3 km | 4.4 km (8 %) | 78 | 34 % | 14.3 |
| OCC (2026) | ↓3,242 m | 28.1 km | 4.7 km (17 %) | 59 | 61 % | 18.2 |
| UTMB (2026) | ↓10,111 m | 80.5 km | 15.6 km (19 %) | 187 | 62 % | 18.5 |
| Val d’Aran EXP (2026) | ↓2,141 m | 15.1 km | 4.9 km (32 %) | 45 | 78 % | 20.9 |
These are Pendio’s measurements of each organiser’s file, as of October 2026; each race page gives the organiser’s own figures, links their file and doesn’t print a descent load. The pages are unofficial: Pendio isn’t affiliated with these races. Counting each band at its average grade slightly undercounts, most in the open-ended band steeper than −20 %, so if anything the steep share is higher.
The last column is the descent’s steepness as the DLI weighs it: on a steady grade, the DLI per 1,000 m of D− is the grade itself. So a metre of the EXP’s descent weighs about 1.5 times one of the Lavaredo 120K’s, though D− counts them alike. At the other end, UTMB’s downhill gentler than −10 % is 47 % of its downhill distance and 6 % of its descent load.
For scale, take an invented training week that loses 1,500 m (4,920 ft) at −12 % on average: 1.5 × 12 = 18. UTMB’s biggest single descent, ↓942 m to Les Chapieux at −17 % on average, comes to about 16 on its own, and the whole course to about ten such weeks. So when you set your training against a race, look at its steep downhill kilometres, not only its D−.
How do you count it week by week?
Add up each run’s descent load over the week (per descent, height lost in km × grade in %) and follow the weekly figure as a trend against your own usual weeks, not a fixed target, next to your training stress rather than instead of it.
Your own weeks are the yardstick because the same descent doesn’t land the same way on everyone. In the 2020 review, the people least used to downhill running generally had the most muscle damage after it, and a first bout of downhill running reduces the markers of muscle damage after a second one, days or weeks later: the repeated bout effect. That protection is “limited in time”: in one study it was gone 9 weeks after a 30-minute downhill run, with no eccentric work in between (Bontemps et al., 2020). The same number can mean more after a winter on the flat than after a summer in the mountains.
The Pendio iPhone app gives every run a DLI and follows it week by week, with an acute:chronic ratio of its own (your last 7 days against your last 28), because “descent can spike while the total stays calm” (guide). A ratio describes a change; it doesn’t forecast what the change will do.
What can’t a descent count see?
A descent load counts slope and distance. It can’t see your speed, the ground underfoot or how tired your legs already were. Speed is one of the three main factors the 2020 review names, and the DLI leaves it out: a −20 % descent jogged and the same descent raced count alike, and a rocky −20 % counts like a smooth one.
It describes, it doesn’t diagnose: a descent load says how much steep downhill you ran, not what it did to your legs or whether you’ll get hurt. If a descent leaves you in pain, talk to a doctor or physio.
Sources
- Lu Z, Suo B, Deng L, Wang J, Fu W, Zhong Y, Jin J. “A review of uphill and downhill running: biomechanics, physiology and modulating factors.” Frontiers in Bioengineering and Biotechnology 13: 1690023, 2025. doi.org/10.3389/fbioe.2025.1690023
- Bontemps B, Vercruyssen F, Gruet M, Louis J. “Downhill running: what are the effects and how can we adapt? A narrative review.” Sports Medicine 50(12): 2083–2110, 2020. doi.org/10.1007/s40279-020-01355-z
- Yokozawa T, Fujii N, Ae M. “Kinetic characteristics of distance running on downhill slope.” International Journal of Sport and Health Science 3: 35–45, 2005. doi.org/10.5432/ijshs.3.35
- Runalyze. “Glossary: Impact Load Factor.” Checked 5 November 2026. runalyze.com
- Pendio. Race course pages: each organiser’s file measured with Pendio’s engine, October 2026. pendio.app/blog/races
- Pendio. GPX course analyzer. pendio.app/tools/gpx-course-analyzer
- Pendio. “How Pendio works”: training on your phone and split times. pendio.app/guide
Questions
Does downhill running count as an easy run?
For your heart and lungs it can be; for your legs it isn't. Running costs the least energy between about −10 % and −20 %, yet downhill running asks for a high volume of eccentric braking work from the leg muscles (Lu et al., 2025). Count it in training stress by its time, and as descent load by its steepness.
Is elevation loss (D−) a good measure of downhill load?
It's half the picture. D− gives each metre of height the same weight, whether it came off at −5 % or −30 %. Weighted by steepness, 100 m lost at −30 % counts three times as much as at −10 %.
How much downhill running should I do before a race?
There's no researched number to give: a 2020 review calls the literature on the load-response to downhill training "scarce" (Bontemps et al., 2020). Compare your recent weeks with the race's descents, the steep ones above all, and decide with your coach. If a descent leaves you in pain, talk to a doctor or physio.
The bottom line
Give downhill running a number of its own: weight each descent by its steepness, add up the week, and read it against your own usual weeks, next to your training stress. Then see where your race's descent load sits: the free GPX course analyzer reads the organiser's GPX for its D−, its biggest descents with their average gradient, and the distance in each grade band.