Java • Game Tree Search
Tic-Tac-Toe AI
Tic-Tac-Toe is my compact playground for learning new technologies. This Java edition explores domain modelling, console interaction, and a stubborn intelligence that searches every future.
The Search
It does not guess. It examines the futures left on the board.
Every legal move is classified by recursively exploring the remaining game tree. A move is kept as a victory when one can be forced, as a draw when defeat can be avoided, and as a loss only when no stronger future remains.
Random selection happens only between outcomes of the same quality, so the play stays varied without becoming careless.
Victory first
Winning continuations outrank every alternative and are selected immediately.
Draw secured
When no forced win remains, the search protects the strongest non-losing path.
Varied play
Equivalent moves are chosen randomly, keeping repeated games from feeling scripted.
The Model
A compact domain model keeps the algorithm readable.
Board owns legal placement and win detection. Move carries position and sign. HumanPlayer and AbhishekAI share one Player contract, allowing either side of a match to be human or automated.
Each recursive branch copies the board before evaluating it. The live match remains untouched until the selected move is committed.
Board snapshots
Search branches evaluate isolated copies instead of mutating the active game.
Player strategy
Human and AI players satisfy the same move-selection contract.
Winning strokes
Rows, columns, and diagonals are reported as explicit game outcomes.
The Stage
A terminal application, presented without turning it into a web app.
Rangmanch preserves the program's native console interaction. The browser renders terminal state while the approved Java process remains the application of record.
The production session will be short-lived, rate-limited, isolated from the homelab, and destroyed when the game exits or the visitor leaves.
One fixed process
The runner starts this game directly; visitors never receive a general shell.
Ephemeral session
A backend-authored deadline and cleanup controller bound every execution.
Isolated by default
No public internet, persistent disk, cluster credential, or arbitrary image selection.
The Playground
One familiar game. A different technology lesson each time.
Tic-Tac-Toe gives every experiment the same small, understood problem space. That leaves the interesting differences exposed: language design, state management, rendering, interaction, packaging, and runtime behaviour.
This console edition became a way to explore Java and game-tree search. A later browser edition became my bridge from React into Svelte. The board stays familiar while the technology changes around it.
Java console
Domain modelling, recursive search, and native terminal interaction.
Svelte browser
A practical way to learn Svelte through a UI already familiar from React.
Rangmanch
The same playground now exercises ephemeral application delivery in the browser.
Rangmanch Demo
Play the console edition
Challenge the original Java interface inside a short-lived Rangmanch terminal.
tic-tac-toe-java
Ready
The terminal is ready.
Run the interface rehearsal while the isolated Java runtime is connected.
Built With
Java
Minimax
Console I/O
RangmanchThe Match Ends
Tiny board. Complete search space.
The interesting part is not the size of the game. It is the discipline of modelling every state clearly enough that the strongest move falls out of the system.
Tic-Tac-Toe AI • Built by Abhishek Kashyap
Read the original Java
The source keeps the board, moves, players, win detection, and recursive AI together in a compact console program.
View Repository (opens in a new tab)Runtime, not rewrite
The production demo will execute the console application through Rangmanch rather than replacing it with browser game logic.