Matchbox “machine learning”
In 1961, the MENACE computer used 304 matchboxes and beads to learn Tic Tac Toe through reinforcement—one of the earliest public ML demos.
Play vs a friend on the same device or try the computer on Easy, Normal, or Unbeatable. Use Hints to preview the best move, or Undo to rewind. Scores and your preferences are saved on this device only.
Tic Tac Toe—also known in the UK, Ireland, Australia and New Zealand as Noughts and Crosses—is more than a quick pastime. It’s a compact lesson in logic, planning, and pattern recognition. Because the board is only 3×3, new players rapidly see how choices lead to consequences, which makes it perfect for classrooms, home learning, and brain breaks at work.
Grid-based “three-in-a-row” games go back a long way. The ancient Romans played a circle-and-stone game often linked to terni lapilli, where each player had three pieces to move on a small grid. The modern pencil-and-paper form gained popularity in the 19th–20th centuries under names like “Tick-tack-toe” (US/Canada) and “Noughts and Crosses” (UK/Ireland). Today it’s a staple warm-up in maths clubs and coding workshops because the full game tree is small enough to understand—and small enough for a computer to solve perfectly.
Tip: Turn on Hints to see the best move, then try to explain why it’s best. Explaining your reasoning is the fastest way to improve—whether you’re playing in London, Dublin, Toronto, Sydney, or anywhere else in the world.
In 1961, the MENACE computer used 304 matchboxes and beads to learn Tic Tac Toe through reinforcement—one of the earliest public ML demos.
The entire game tree has just 255,168 legal positions. With perfect play it’s always a draw—explorers mapped every outcome decades ago.
A fork creates two winning lines at once; if the opponent can’t block both, you win. Spotting forks is the fastest way past beginner play.
A tied game is often called a “cat’s game” in the U.S.—as in, “no one gets the milk.” It stuck around from mid-20th-century kids’ slang.
Move to a 4×4 or 3D 3×3×3 board and draws vanish—suddenly there are winning strategies again and fresh AI research puzzles.