Ask for the shape of the challenge.

Build orders are the curve, and the campaigns teach almost none of what the game actually asks for. The hard AI does not forget its economy while it fights.

A hundred and seventy champions, each with four abilities you are expected to know the cooldowns of, and the tutorial covers none of it.

Every hero is free and every one of them can do something no other game would let a hero do, which is why the learning curve is measured in years rather than hours.

Two games at once — a campaign map that decides the war and real-time battles that can still lose it. Army composition answers most questions before the fighting starts.

Armour has to be stripped before any control lands, surfaces do more damage than swords, and the game explains neither. Act two is where under-levelled parties stop entirely.

The town is gone, so there is no bank of spare heroes to absorb a bad decision — a run is four people and every relationship between them is a mechanic.

Weakness targeting is the whole of combat, and the calendar is the real difficulty: a day spent wrong in November cannot be spent again.

The rules are the difficulty. A fight lost on turn one was lost in the inventory screen, and the game is content to let you walk into an encounter five levels early without a word. Honour mode makes one wipe final; Explorer makes almost nothing final.

A city builder where the scarce resource is time and every law you sign is unrepealable. The storm arrives on schedule whether or not the city is ready.

Classic mode makes every death permanent, and the calendar decides months in advance who will be strong enough to survive. Maddening turns recruitment into its own strategy layer.