MeasuredThe numbers

Ways carrying the route number are stitched end to end, resampled every 30 m, and walked counting every change of turning direction — one lean, one curve. A bend counts once it adds up to 4 degrees, so a camber correction does not. A hairpin is a sustained arc of 120 degrees or more.

Every count here is a count at a stated ruler: the road sampled every 30 m, a bend counted once it adds up to 4 degrees. It is reproducible, not true. Halve the ruler and the number climbs; the same road measured at 1 km gives a twelfth of what it gives at 10 m. What bounds the fine end is the survey, not the asphalt. See `yardstick` in this file.

What the number has to mean. Thai Wikipedia gives the Chiang Mai to Mae Hong Son run by Route 1095 as about 245 kilometres. Spread 1,864 curves across that and you get one every 131 metres — which sounds relentless until you have ridden it, and then it sounds about right.

The same article describes the road as winding กว่า 2,000 โค้ง — more than two thousand. So the two figures in circulation are close to each other, not far apart.

Counting it independently. Every change of turning direction on the OpenStreetMap trace of this route — one lean, one curve — comes to 1,056 on Route 1095 alone, and about 1,310 across the full 245 kilometres. That is one every 187 metres.

Why that count is not a rebuttal — and why no count could be. Ask how many curves this road has and the answer is another question: at what ruler? Sampled every kilometre it has about 105. Every 250 metres, 395. Every 30 metres — what this site publishes — 1,083. Every 10 metres, 1,117. The number does not converge on a true value as the ruler shrinks; it keeps climbing, the way a coastline gets longer the finer you measure it. Richardson found this in the 1950s and Mandelbrot named it.

So 1,864 is not a claim this page can confirm or refute, because it is not the same kind of thing as a measurement with its method attached. It is a count at somebody's ruler, and nobody published the ruler.

What bounds the fine end here is the survey, not the asphalt: OpenStreetMap carries a point roughly every 40 metres, and asking for a reading below that interpolates a straight line between two points that exist, which invents no bend. Volunteer traces also cut corners, smoothing exactly the tight stuff that makes the count.

Inference — which is the useful conclusion: an independent count at a stated ruler lands the same order as the sign. Nobody publishes the survey behind 1,864 and this page cannot confirm it to the digit. What it can say is that the number is the right size, and that anyone who rides it and claims it is not exaggerating.

Count the bends on this road with a coarse rule and the small ones disappear into the straight between two big ones. Count with a finer one and those straights turn out to have bends in them too. At a one-kilometre ruler this run has 105 curves in it. At a ten-metre ruler it has 1,117. Same road, same day, same arithmetic.

Richardson noticed this measuring coastlines in the 1950s and Mandelbrot gave it a name. A coastline has no length until you say how long your ruler is, and this road has no curve count until you say the same. Every number on this site is a count at a stated ruler — thirty metres, four degrees — not a fact about the road. It is reproducible, which is a different and smaller claim than true.

The line bending flat at the fine end is not the road running out of curves. It is OpenStreetMap running out of points: the trace carries one roughly every 40 metres, and asking for a reading below that interpolates a straight line between two of them, which invents no bend. Survey the same asphalt at one metre and the climb would carry on.