Playground
Tic-Tac-Toe
Choose a square to make the first move.
The search plays random rollouts, then backs up each result through the tree.
SEARCH / LABInteractive algorithm notebook
Runs locally in your browser · no model download
Monte Carlo
Tree Search
Watch a search tree learn which Tic-Tac-Toe move is promising. Change the simulation budget, inspect candidate branches, and play against the search.
Your move
Search trace
Root move candidates
Iterations—
Tree nodes—
Best move—
Each branch is a move from the current board.—
Search tree is emptyRun a search on the computer's turn to compare candidate moves.
Branch inspector
Select a candidate branch to preview the board after that move.
01 Select
Follow promising branches using the UCT exploration score.
02 Expand
Add one unexplored legal move to the search tree.
03 Roll out
Play random legal moves until the game ends.
04 Back up
Update visit counts and values along the path.