The north had twenty-five centimetres of ice. The tour was cancelled on eight.

There is a skating race in the north of the Netherlands called the Elfstedentocht — the Eleven Cities Tour. It is almost two hundred kilometres long, it runs on frozen canals, rivers and lakes through the eleven historic towns of Friesland, and it is the largest ice-skating tour in the world: three hundred competitors, and sixteen thousand people skating it as a tour behind them.

It has been held fifteen times since 1909, three times since 1963, and not once since 1997.

Not because anyone lost interest. The organisation is intact. There is an association you must belong to in order to enter, a route, a stamp to collect in each town and at three secret checkpoints, a midnight cutoff, a finish on a canal outside Leeuwarden. What there mostly is not, any more, is ice.

The rule is the interesting part. The tour is held only if the ice is at least fifteen centimetres thick along the entire course, and stays that way. Not on average. Not at the start. Along all two hundred kilometres of it, under the bridges and across the open lakes, for as long as sixteen thousand people take to pass over it. When the ice is judged good the tour is announced and begins within forty-eight hours — a deadline that short because nobody trusts a multi-day forecast of what ice will do.

The fortnight the country got ready

From the end of January 2012 the Netherlands had roughly two weeks of frost: the Dutch meteorological institute, KNMI, dates the cold spell 29 January to 12 February, averaging −5.8 °C at De Bilt. The country began to do the thing it does — there is a word for it, Elfstedenkoorts, eleven-cities fever — and this time it was not only atmosphere. Boats were banned from the canals so they would stop breaking the ice. Hotels in Friesland filled. Estimates of one and a half to two million visitors went around. Mark Rutte, then prime minister, said (as the Associated Press later reported him) that once every fifteen years the country is not governed from The Hague but by twenty-two district heads in Friesland — and that it is in good hands.

Those twenty-two district heads, the rayonhoofden, are the people who go out and drill. On 5 February they met for the first time in fifteen years, and what they reported was not a national verdict but a split one: in the north of Friesland the ice was good. In the south-west it was not. Around the lakes and near Stavoren there were places with two centimetres.

The country mostly heard the first half. On the afternoon of 8 February the association’s chairman, Wiebe Wieling, called it off, and the numbers he gave are the ones worth keeping: on average eleven centimetres of ice, and in some places only eight or nine. Behind that average sat about twenty-five centimetres in the north and around seven in the south. A thaw was forecast within days, so the thin end was not going to be given the time it needed.

Somebody proposed skipping Sloten and Stavoren — the two towns in the worst region. That was rejected, on the grounds that it would leave nine cities.

What the number was measuring

The twenty-five centimetres was real. It simply was not anywhere the decision was about.

That is the whole shape of it. Almost every summary number you meet is an average, and an average is a statement about a population: it tells you what you would find if you sampled at random. A route is not a population. It is a chain, and a chain is scored at its worst link, so the only figure that can decide a chain is the smallest one on it. Nine-tenths of a route at twenty-five centimetres does not average with a stretch at eight into something safe. It produces sixteen thousand people arriving at that stretch, in the dark, at a place where they may already be walking on their blades over matting because the ice is too thin to skate. The 1963 tour, at −18 °C in powder snow and an easterly wind, finished sixty-nine of its roughly nine thousand three hundred tour skaters. That is the distribution the committee holds in its head.

Note that the organisation does not need telling any of this. The whole-route minimum is written into the rule, the forty-eight-hour announcement exists because nobody believes a longer forecast, and the twenty-two people who decide carry drills. The failure mode I am describing is not theirs. It is the one waiting for everybody outside the room, holding a provincial average and a national broadcast and a hotel booking, quite reasonably reading “twenty-five centimetres” as the answer to a question it was not the answer to.

There is a second lesson underneath, and KNMI put it in writing afterwards. The obvious story is that the cold spell was too short. KNMI’s own analysis says the temperatures were essentially sufficient; what beat 2012 was about six centimetres of powder snow lying on the ice, insulating it from the frost, plus an unusually sunny early February. The 1997 tour had a comparable build-up without the snow. So the variable everyone was watching on television — how many days of frost — was not the variable that decided it. The thing that measured the route was twenty-two people with drills, and what they were measuring was not the weather but the ice.

The tour has been called endangered; three times in fifty years, and the 2012 near-miss is itself now fourteen years old. It may be that the last one has already been skated and nobody knew at the time that it was the last one. But that will be the climate’s decision. The committee’s job was only ever to find the thinnest place on the route and answer honestly about it in a week when the entire country wanted a different answer, and in 2012 it did.