No opponent around at 3 a.m.? The computer is always awake. Four levels, from a gentle teacher to an opponent that out-counts almost every human — and practice games never touch your rating, so every experiment is free.
▶ Challenge the AIPlays the strong move about a third of the time and a casual move the rest. Wins roughly one game in five against a beginner who knows the rules. Use it for your very first games and for testing a brand-new idea without pressure.
Picks the strong move about half the time. Beats a casual player roughly every second game. This is the level to measure yourself against: when you beat Normal consistently, you are ready for real opponents.
Plays the strong move two times in three and wins about four games in five against average players. Hanging pawns, wasted walls and lazy endgames all get punished. Beat Hard regularly and you will hold your own in public rooms.
No holding back: a full game-tree search that thinks for over a second every move, examining your replies to every line and playing only into positions it can hold. It practically cannot be beaten — play it to find the holes in your game, not to pad your ego.
Honestly, and the same way this guide teaches you to think. Every move it measures both shortest paths — yours and its own — and weighs walls by how much they lengthen your road. Its wall candidates cluster around your shortest path and extend existing structures at choke points, exactly like the wall tactics page recommends. The lower levels simply mix in weaker moves on purpose; Hardcore searches many moves deep and never blunders. If you want to understand why it beat you, replay the game: the endgame math section shows how to count what it counted.