The twistiest roads on Mallorca, measured

Nobody knows how many hairpins the Coll de Sóller has. Duva Pollença says more than 60. Epic Road Rides counted 30 and admits it is difficult to keep count. Love Velo gives 20 on one side and almost 30 on the other. All three describe the same tarmac and none of them is wrong, because a hairpin has never been defined — every count ever published for this island was made by a person on a bicycle deciding, corner by corner, whether that one was tight enough to say out loud.

The road is already a dataset. Every climb in the atlas is measured along the OpenStreetMap centreline of the road, and a centreline knows exactly how far it turns and how much tarmac it takes to do it. So the count can be a measurement: write the rule down, apply it to all of them, and publish what falls out — including the roads where the answer is none.

Hairpins on the island

342

out of 917 turns of 90° or more, on 116 climb-sides

Most on one climb

25

Coll de Sóller from Sóller, all of them inside 8.4 km

Tightest road on the island

3.6/km

Coll des Grau from Puigpunyent — 13 hairpins in 3.6 km

Published counts checked

3 of 4

come back where the guide put them, once it says where on the road it was counting

What counts as a hairpin

A hairpin here is a turn of at least 135° taken inside 150 m of road. Both halves matter. The angle alone would count a long mountain curve that eventually doubles back over two kilometres of valley side; the length alone would count a wiggle. What a rider means by a hairpin is a corner tight enough that you can see the road you were just on, and that is an angle and a distance together.

Every number in that sentence is a judgement, which is why this page publishes three counts for each road rather than one. The strict figure is the hairpins. The second drops the length limit and keeps the angle. The third opens the angle to any turn of 90° or more — the ones you brake for. A guide that counted from the saddle is usually reproducing one of the looser two, and which one tells you what its author was counting.

Where the threshold falls

Here is where the thresholds fall in the pile. Every one of the 917 counted turns is in this histogram, in 15° steps, and the 135° line is drawn on it. The shape is the argument: there is no gap in the distribution for a threshold to sit in, so any line drawn through it moves the count. 241 turns go past a half-circle.

Checking the guides against the road

A published hairpin count cannot be right or wrong until somebody says what a hairpin is — which is why the disagreements below are not errors. But some guides do something better than defining the word: they say where on the road they were counting. That is enough. Given a stretch — or a count and the point it runs out at — this dataset can go to the same tarmac and see what it finds.

Four published counts on this island say where they were taken. 3 of the 4 come back where the guide put them — two of them within a single turn of the printed number, and the third as a run of corners that finishes where its author says it finishes.

3 of 4 is a better hit rate than the state of the published numbers would suggest, and it says something worth saying plainly: the guides are not counting badly. They are counting without a stated rule, on roads where the count depends on the rule — and the moment one of them names the road it counted, the number turns out to be reproducible from the map. The one that does not come back is the one whose page gives the same climb two different lengths, which is the more useful lesson of the four: a count is checkable when its author says where the road starts and then keeps saying the same thing.

For the two climbs the guides argue about most, the whole spread, attributed rather than collapsed.

The guides that publish no number

And the normal case, which is worth recording rather than omitting: a guide describes the hairpins at length, gives the length and the gradient to a decimal place, and never says how many there are.

• Cycle Fiesta — Publishes 7.7 km, 424 m and 5.5% for the climb, and no bend count.
• Café du Cycliste — Sends the reader up “each and every one of the multiple hairpins” towards the noose on Sa Calobra, and never says how many that is.

Every counted climb, ranked

All 116 counted climb-sides, most hairpins first. Sort it however you like; the interesting column is usually the last one. 37 of them have no hairpin at all under the strict rule, which is a fact about those roads and not a gap in the data — and 43 are below the atlas's publication bar, so they are counted here and have no page of their own to link to.

The tightest road, which is not the one with the most

The road with the most hairpins is not the tightest road, only the longest twisty one. Per kilometre — counting only climbs of at least 2.0 km, because two corners in 400 m is a short road rather than a twisty one — the island's tightest stretch of tarmac is Coll des Grau from Puigpunyent: 13 hairpins in 3.6 km, which is 3.6 of them per kilometre. It is not a famous climb and it is not in any list of ten. Second is Coll de Sóller at 3.0 per kilometre.

One road is left out of that by 1 m: mirador de na Burguesa from Gènova packs 3.0 hairpins into a kilometre and is 1,999 m long, so it misses the 2.0 km floor and the table above rounds its length to 2.0 km anyway. Sort the ranking by hairpins per kilometre and it is the row sitting where this paragraph says Coll de Sóller should be. The floor stays where it was published rather than moving to tidy the result — but a threshold that decides a ranking by a metre is worth seeing, and this is what one looks like.

Where the hairpins are

Every counted turn on the island, plotted where it is. The hairpins are not spread over Mallorca; they are stacked into one mountain range. The Serra de Tramuntana holds 255 of them — 75% of every hairpin counted — against 6 comarques with any at all.

One more thing falls out of the plot, and it is a check rather than a fact about the island: a road that climbs a mountainside by switching back has to turn each way the same number of times. Across 917 counted turns there are 458 to the left and 459 to the right. One apart.

Does the counting work?

A measurement nobody can contradict is not a measurement. This one has exactly one independent test, and the island supplies it: there is a single place where a road here passes over itself. OpenStreetMap records that as a bridge on a higher layer — topology, with no angle in any tag — and a road cannot cross its own alignment without having turned through well over a half-circle first. So a counter that is finding real corners has to put a big one there, and nothing about the claim comes from the geometry we measured.

The sharpest single turn in the dataset is 371° on Ermita de Bonany from Petra, at km 3.8 — a full circle and a bit, which is what a road does when it needs to gain height on a spur too narrow for a switchback.

The climbs with no count

10 climb-sides carry no count. The counter re-derives the road from the map rather than reading back the path the gradients were sampled along, so it can check itself: where its own path disagrees with the measured length by more than 2%, it is counting the corners of a different road, and it publishes nothing rather than a number.

How the count was made

One input and one script. The road centreline comes from OpenStreetMap, the same data every gradient on this site is sampled along; no elevation is needed to count a corner, so this measurement never touches the terrain model. The path is walked at a fixed spacing, heading change is accumulated while the road keeps bending the same way, and a turn is closed when the road straightens. What that leaves out, and how the rest of the site is made, is on the methodology page.

Road geometry © OpenStreetMap contributors, ODbL. Counted under the rule above, and every threshold on this page is read from that rule rather than typed beside it.

The tight roads are stacked into one corner of the island, which makes this an unusually easy trip to plan. The town at the foot of the two twistiest climbs on the list is Sóller, and a bed in Sóller puts the Coll de Sóller on your doorstep in both directions. From Palma you will want a hire car to reach the rest without spending the morning on the Ma-11. Both links are marked and pay us if you use them.