Coding and AI classes in the East Midlands

The learner loads thirty complete years of Sutton Bonington rain into a table, one row per year and one column per month. Route one averages each column first, then takes the largest average: June, 63.7 millimetres. Route two takes the largest value in each row first, the wettest month of that year, then averages those thirty maxima: 104.4 millimetres. Both are honest calculations on the same numbers. They answer different questions, and mixing them up leads to badly wrong planning.

Why is the second answer so much bigger? In any one year, some month is unusually wet, but which month that is changes from year to year. Averaging first smooths those wet spells away, spreading them across the column; taking the maximum first catches each year's wettest spell wherever it falls. So the maximum of the averages is always at most the average of the maxima, and here the gap is about 40 millimetres. The tally makes the point another way: June wins eight times, July and August four each, and even January, December and February have their turns.

This matters far beyond rain. A drain sized for the average wettest month would overflow in most years. A shop stocking for its average busiest day would run out on a typical busy day. The learner writes both routes as functions, tests them on a tiny invented table where the answers can be checked by hand, and adds a check that the first route never exceeds the second. The slip to avoid is quietly using one route and describing it in the words of the other.

Knowing that the order of steps can change the answer is a strong reason for an East Midlands teenager to keep coding in 2026. The longer argument is in why teenagers should still learn to code in 2026.