Coding and AI classes in North West England

The learner adds up twelve months of rain for every year from 1961 to 2020, then splits the years into two thirty-year halves. For each half, Python computes the mean annual total and the standard deviation, which measures how far a typical year sits from that mean. Because a wetter place naturally has bigger swings in millimetres, the fair comparison divides the standard deviation by the mean. That ratio, the coefficient of variation, says how large the typical swing is as a fraction of the typical year.

The spread rose much faster than the mean, so the coefficient of variation rose too: from about 12 per cent of a typical year to about 17. In plain terms, a year at Newton Rigg is now more likely to be far from average, in either direction, even after allowing for the wetter average. The driest year of the later half, 1996, was drier than the driest of the earlier half, and its wettest, 2020, was wetter than any year in the earlier half. Thirty years is not a long sample, so this is evidence of a change at one station, not proof of a regional trend.

There is a technical choice that trips people up. Python's statistics module offers two standard deviations: one divides by the number of years, the other by one fewer. With thirty years they give 111.6 and 113.5 millimetres for the first half, close but not equal, and a report must say which it used. The bigger slip is comparing raw standard deviations between a dry place and a wet one, which almost always makes the wet place look more erratic. The learner computes both versions and writes a check that the coefficient of variation has no units.

Knowing when a bigger spread is only a bigger average in disguise is worth a North West teenager learning to code for in 2026. The longer argument is in why teenagers should still learn to code in 2026.