Coding classes for teens

The hidden bias

The first-empty-square fix nearly doubles the chance of a mine in the top-left corner, to 23.32% against the fair 12.5%. A player who learns that trick wins more often than the game intends, and the code passes every "does it crash?" test.

Two fair fixes

Moving the mine to a random empty square, or dealing a whole new board, both give every square close to 12.5%. The differences left are the normal wobble of random sampling.

The simple one wins

Dealing again sounds wasteful, but it needed only 1.143 attempts on average (theory says 1.141). It is the shortest code, obviously fair and fast enough. The clever fix was the one with the bug.

Boundary cases, as ACARA asks

The same game is full of edges. A corner square has 3 neighbours, an edge square 5 and an inside square 8. Most first versions of the neighbour count crash or miscount on the edges.

Students write a test for each of the three cases before writing the counting code, exactly the boundary testing the curriculum describes.

What the project teaches

Two-dimensional lists, loops, random sampling, recursion for opening empty areas, and the idea that "it works" and "it is fair" are different tests.

It also teaches a habit worth more than any of those: when a result matters, measure it rather than trust it.