Simon Carr: Predicting weather is a mug's game

Three days out and the forecasters are guessing – and it's been proven so

We still fall for it and it's hard to know why we don't learn. The cold, hard truth is that weather forecasting more than three days out is a mixture of salesmanship, superstition and morbid curiosity. So I offer a bet to any professional meteorologists out there. I will bet my farm against yours that my five-day forecasts for the rest of this year will outperform yours.

That's quite a bet, considering you have a multibillion-pound budget at your disposal and a trillion terabytes of computable data. Me, I have £4.50-worth of manuals and a hangover – and I will win without trying.

An initial question: what does a 50 per cent chance of snow mean? That the last 100 times these conditions occurred it rained on 50 of them? This is meaningless – "the conditions" are never the same. Weather never repeats itself.

The weather is the biggest non-linear system we have. Tiny errors in measurements here make larger differences there, and the differences compound on other false measurements. Three days out you guys are guessing. And it's been proven you're guessing (see below).

And what about a 60 per cent chance of next summer being a "barbecue summer"? This really is rubbishy. Do you really want us to believe that you have discerned that in the last 100 years when "current conditions" occurred, 60 of them resulted in long, hot summers?

I say, then, with sublime confidence, that by using seasonal averages, and the naïve forecast that tomorrow will be much like today – my forecasts will outperform any professional's.

The Freakonomics people looked at TV weather forecasts in Kansas and found that the predictions of rain the next day were correct 85 per cent of the time. That sounds impressive until you ask the trick question. How often are you right if you NEVER predict rain? Then you are correct 86.3 per cent of the time. If you exclude settled periods (long, dry summer periods), the success rate drops off – in some cases to zero.

That zero, incidentally, was scored by the National Oceanic and Atmospheric Administration, despite its $4bn budget.

This uncertainty should be taught in schools. It would unsettle all sorts of grandiose pronouncements, schemes, plans and political projects. Global warming, for instance. As we know, sceptics say that our previous weather anxiety in the 1970s was for an ice age. But the reason for that has been forgotten.

It was argued that the climate was warming, the ice caps were melting, the cold water released into the oceans would bring down temperatures with a crash and the ice would return to Europe. And who's to say that global warming won't turn out like that?

I would make one real prediction: if the ice comes, it won't take 2,000 years. When complex systems change from one shape to another they don't do it gradually. They hold their shape until stressed too far and then they flip. If it happened, it would happen fast. In a decade. Maybe in a year. Yes, there's an 80 per cent chance of it.