Alpe d'Huez's 6.6% Average Is a Misleading Number

The 6.6 percent average gradient of Alpe d'Huez is one of the most quoted numbers in road cycling, and it is nearly useless as a planning input. Hear us out. From Bourg-d'Oisans at 773 metres to the ski station summit at 1,820 metres, the climb covers 15.97 kilometres and gains 1,048 metres of elevation. Divide one figure by the other and the arithmetic returns 6.6 percent. Climbfinder records a published maximum of 13 percent. No single kilometre of the road behaves like the average. The rest of this piece walks three composite riders through what the profile actually asks of them.

Whether the 6.6 percent number matters to you at all depends on which of three riders you are. It depends on what you brought to Bourg-d'Oisans, what you expect to leave with, and what question you were trying to answer when you first typed the climb's name into a search bar. We will walk through three composite riders below. None of them is a real person we interviewed. Each is a stand-in for a question we get from readers, sharpened by the arithmetic of a climb we have measured.

Scenario 1: The First-Timer Pacing Off a Road Book

Imagine a rider we will call the First-Timer. Picture someone who rides seriously at home — a few hundred kilometres a month, a familiar local climb of eight or nine kilometres — but has never turned the pedals in the Oisans. They have booked four days in Bourg-d'Oisans. They have read three road books, all of which quote the same figure: fifteen kilometres and change, roughly a thousand metres of gain, 6.6 percent average. They have done the maths at home. At their steady tempo pace, six-point-six percent for sixteen kilometres suggests a time they can rehearse in their head. They have a target watts-per-kilo number, they have a target time, and they have a plan to leave Bourg-d'Oisans early, take the first kilometre gently, and hold their average for the rest.

This rider does not survive kilometre three.

The reason is not fitness. The reason is that 6.6 percent is the answer to an arithmetic question — 1,048 metres divided by 15.97 kilometres — and not a description of the road. Somewhere on the way to the ski station, the road hits the 13 percent figure that climbfinder publishes as the maximum. That is not a rounding artefact. It is nearly double the average. If some part of the road is at 13 percent, some other part of the road is well below 6.6 to make the mean work out. The road oscillates around its own average, and the oscillation is what a first-time visitor has never seen at home on a nine-kilometre climb that behaves like itself.

The planning failure is specific and repeatable. The First-Timer sets off from Bourg-d'Oisans at their rehearsed pace. Within the first two or three kilometres, the gradient exceeds their target, sometimes substantially. Their power lifts to hold their speed. It stays lifted, because when the road briefly eases they are still trying to make up the average time they think they are losing. By the time the road settles, the damage is done. The climb has fifteen kilometres left. The rider now rides them all at a power they cannot sustain, because the first three kilometres taught their legs a false lesson about what six-point-six percent feels like.

The correct read of the road book for this rider was never the average. It was the maximum. Thirteen percent tells them the climb is not a metronome. Fifteen-point-nine-seven kilometres times a variable gradient with peaks at nearly twice the mean is not a tempo effort — it is a variable-power effort disguised as one.

Scenario 2: The Repeat Visitor Chasing a Personal Best

Now imagine a rider we will call the Repeat Visitor. Let us say they have climbed Alpe d'Huez three times over five years. They know the road. They know where it steepens and where it releases. They have a personal best they set on their second attempt, and they arrive in Bourg-d'Oisans wanting to take a chunk out of it. They do not need to be told the climb is 15.97 kilometres or that it gains 1,048 metres. They know. They have a Strava file that tells them, kilometre by kilometre, where they lost time last year.

For this rider, the 6.6 percent average is a diagnostic tool, not a planning input. They already know the road oscillates. What they are trying to answer is a different question: given a fixed budget of watts they can hold across roughly an hour of climbing, where on the profile do they spend and where do they save?

The maths for the Repeat Visitor is not about the average. It is about the delta from the average. Every kilometre that the road runs steeper than 6.6 percent is a kilometre where holding target speed costs disproportionate power. Every kilometre that runs shallower is a kilometre where the same power buys extra speed. If the road hits 13 percent somewhere and averages 6.6, the arithmetic of the mean forces the existence of long stretches meaningfully below 6.6 percent to compensate. Those stretches are where a personal best is actually built.

The mistake this rider makes is the opposite of the First-Timer's. They know the steep sections and they attack them, because that is where the previous year's file shows they were slow. But attacking the steepest kilometres of a climb with a 13 percent maximum burns matches at the moment when the road is least generous with them. A twenty-watt overshoot at 13 percent buys almost no extra time. The same twenty watts on a five-percent ramp higher on the mountain buys far more. The 6.6 percent average, read correctly, tells this rider that a Personal Best is a distribution problem, not a courage problem.

The number that would actually help the Repeat Visitor is not on the road book. It is on their own file: the gradient-weighted power curve of their previous attempts. The average gradient is only useful as the axis those curves are measured against.

Alpe d'Huez The print from this article · from €29.95 View the print →

Scenario 3: The Comparison Rider Ranking It Against Other HC Climbs

Picture a third rider — the Comparison Rider — who has never climbed Alpe d'Huez and is not sure they want to. They are planning an Alps trip. They have a list of five or six climbs to choose between. They are comparing them on a spreadsheet: length, gain, average gradient, category. Alpe d'Huez sits in that spreadsheet at 15.97 km, 1,048 m, 6.6 percent. Other climbs on their list sit at similar averages with different lengths, or similar lengths with different averages. This rider is trying to answer the question: which of these will give me the ride I want?

For this rider, the 6.6 percent number is actively misleading in a third distinct way. It creates the false impression that Alpe d'Huez and another climb of the same average gradient are similar experiences. They are not. Two climbs at 6.6 percent can differ by everything that matters — the length of their steepest sections, the position of those sections on the profile, the distribution of hairpins that give the legs a brief geometric reprieve, the altitude at which the effort takes place. Alpe d'Huez tops out at 1,820 metres. A climb of identical length and average gradient that tops out at 2,600 metres is not the same climb. The last three hundred metres of vertical are simply thinner air.

For this rider, the correct use of the 6.6 percent number is as a filter, not a comparison. It tells them Alpe d'Huez belongs in the same broad category as other mid-length HC climbs. It does not tell them how the road behaves. If they want to know how the road behaves, they need to look at the shape of the profile itself, kilometre by kilometre, and specifically at the maximum. A 13 percent maximum on a 15.97 km climb produces a very different suffering than a 9 percent maximum on the same length and average. The mean is the same. The ride is not.

A climb print — the reason our studio exists — is the profile drawn to scale rather than compressed to a single number. The Comparison Rider is exactly the reader we imagine when we draw them. If they want to see Alpe d'Huez next to the others they are considering, the profiles live in the shop, printed at a scale where 13 percent looks like 13 percent and 4 percent looks like 4 percent.

Three riders, three different questions, one number that fails all of them in different ways. The First-Timer needs the maximum, not the average, because the maximum tells them the road is not a metronome. The Repeat Visitor needs the distribution around the average, because a personal best is built by spending unevenly where the road spends unevenly. The Comparison Rider needs the shape, because two climbs with the same average can produce entirely different rides.

What they share is the same underlying arithmetic mistake, applied differently. All three treat 6.6 percent as if it described the road, when in fact it describes only the endpoints. The average gradient of any climb is defined by two data points — start elevation and summit elevation — divided by the horizontal distance between them. It is completely indifferent to what happens in between. A road that climbed in a perfectly straight ramp from Bourg-d'Oisans to the ski station would return the same 6.6 percent. Alpe d'Huez does not climb in a straight ramp. Its published maximum is nearly double its average, which is arithmetic proof that the shape carries almost all of the information about the ride, and the average carries almost none.

This is not an argument against average gradient as a metric. It is an argument for reading it as one of at least three numbers — average, maximum, and length — with the understanding that the maximum is where the climb hurts and the shape is where the strategy lives.

Which Scenario Is You

If you are planning a first attempt at Alpe d'Huez and you have been rehearsing a target time based on the 6.6 percent number, you are the First-Timer. The correction is to plan against the 13 percent maximum and treat the average as a sanity check on total effort, not as a pacing target.

If you have climbed it before and want to go faster, you are the Repeat Visitor. The correction is to stop thinking about the average altogether and start thinking about where on the profile your power is worth the most.

If you are choosing between climbs for a trip you have not yet booked, you are the Comparison Rider. The correction is to compare profiles, not averages. Two climbs at 6.6 percent are not the same climb, and no single number can tell you which one you will enjoy more.

Most riders are one of the three at any given time. Occasionally you shift between them across a season. The number that stays useful across all three is not the average.

Thirteen percent on a road with a 6.6 percent mean. That gap is the number that should change how you plan the ride, how you pace it, and how you compare it to anything else on your list. The average is arithmetic. The maximum is the road.

FAQ

Where does the 6.6 percent average gradient figure come from?

The figure is the elevation gain divided by the horizontal length: 1,048 metres over 15.97 kilometres returns 6.56 percent, which rounds to 6.6. Both underlying numbers in our profile are derived from OpenTopoData's SRTM 30-metre elevation dataset, sampled along the road from Bourg-d'Oisans at 773 metres to the summit at 1,820 metres. It is a mean, not a description — the road oscillates around it, sometimes substantially.

What is the published maximum gradient on Alpe d'Huez?

Climbfinder publishes a maximum gradient of 13 percent for the ascent from Bourg-d'Oisans. That figure is derived from road-book measurement rather than the same SRTM sample that produces our average and total gain. It is the number that matters most for pacing, because it tells you the road briefly runs at nearly twice its own average.

Why do satellite elevation data and road-book maxima sometimes disagree?

Satellite elevation data such as SRTM samples the ground at a resolution of roughly 30 metres per pixel, which is excellent for total gain and average gradient across a long climb but too coarse to catch the peak gradient of a short, steep ramp. Road books tend to record the maximum instantaneous gradient measured on-site. The two methods answer slightly different questions, so a discrepancy of a percentage point or two between them is expected rather than a defect in either source.

How long is the climb from Bourg-d'Oisans to the ski station?

The measured length from Bourg-d'Oisans at 773 metres to the summit at the ski station at 1,820 metres is 15.97 kilometres, covering 1,048 metres of elevation gain. Different sources round the length variously to 13.8, 14, or 15 kilometres depending on where they place the start point — some begin at the roundabout, some at the first ramp, some at the town sign. Our measurement anchors on Bourg-d'Oisans as a settlement.

Is Alpe d'Huez a harder climb than others with a higher average gradient?

Difficulty is not fully captured by average gradient. A climb averaging seven percent over ten kilometres asks for less total work than Alpe d'Huez, which averages 6.6 percent over 15.97 kilometres and gains 1,048 metres of elevation. Length, altitude at the summit, and the position of the steepest sections all matter. Comparing climbs by average alone treats them as arithmetic identities rather than actual roads.

What does the maximum gradient tell a rider that the average does not?

The maximum tells you the road is not a metronome. A 13 percent maximum on a climb averaging 6.6 percent means at least one section runs at nearly twice the mean, which forces at least one other section to run well below the mean for the arithmetic to balance. That variance is what a rider actually experiences, and it is where pacing errors on a first attempt tend to originate.

How reliable is the SRTM 30-metre elevation dataset for computing climb statistics?

SRTM at 30-metre resolution is a solid source for total elevation gain, start and summit altitudes, and average gradient across a full climb. Its known weaknesses are dense vegetation cover, very narrow ridgelines, and short-distance instantaneous gradients that fall inside a single pixel. For a climb of Alpe d'Huez's length, SRTM's aggregate figures are trustworthy; for the instantaneous peak gradient, we defer to road-book sources such as climbfinder.

From the collection