Ultimate Tic-Tac-Toe takes the game everyone finds boring by age eight and makes it genuinely deep, using one extra rule. Nine tic-tac-toe boards sit in a 3×3 arrangement. Win a small board and you claim its cell on the big board; claim three big-board cells in a line and you win the game. The rule that changes everything: the cell you play inside a small board decides which small board your opponent must play in next.
| Level | Search | Character |
|---|---|---|
| Casual | 2 plies, loose move choice | Wins small boards it happens to be in. Rarely builds toward a big-board line, and will send you somewhere useful by accident. |
| Tactician | 4 plies, alpha-beta | Sees where its move sends you and avoids giving away free moves. It will spot a two-board threat. |
| Strategist | up to 7 plies, iterative deepening in a 3-second budget | Plays the big board. It will deliberately lose a small board to set up a meta-board line, and it punishes a careless free move immediately. |
The instinct is to win the small board in front of you. That instinct is wrong often enough to lose games. Three ideas do most of the work:
A common opening is the centre cell of the centre board: it takes the best square of the best board, and it sends your opponent right back into the centre board where you already have a mark.
Ordinary tic-tac-toe has 26,830 reachable positions — a browser can search all of them before you lift your finger, which is why the classic game here has a genuinely unbeatable setting. Ultimate Tic-Tac-Toe has an estimated state space around 1038. No exhaustive search is possible, so an engine must evaluate positions it cannot finish reading.
The evaluation used here is layered: claimed boards weighted by where they sit on the big board, then two-in-a-row threats on the big board, then a heavily discounted term for control inside boards still in play, plus an explicit penalty for handing over a free move. That last term exists because the cost of a free move usually lands just beyond the search horizon — the search cannot see it, so it has to be told.
Nine boards in a 3×3 grid. Win a small board to claim that cell of the big board; three claimed cells in a row wins. The cell you play within a small board dictates which board your opponent plays in next; if that board is finished, they may play anywhere.
No. The state space is roughly 1038, far past exhaustive search. Strong play comes from heuristic search or Monte Carlo tree search, not from a lookup table — which is precisely why the game stays interesting when plain tic-tac-toe does not.
You may play in any board still open. Because this is such a strong position, avoiding giving one away is one of the central skills of the game.
No. A small board that fills with no three-in-a-row is dead — it counts for neither player and can never be part of a big-board line. It still blocks that cell, so it can spoil a line you were building.
Centre cell of the centre board is the standard choice: strongest square, strongest board, and it returns your opponent to a board where you already have a mark. Corner-of-centre is the usual alternative if you would rather avoid the well-known lines.
Sign in with GitHub to share strategies, ask questions, or report a bug.