TicTacToe-v0/README.md

TicTacToe-v0

Overview

TicTacToe is a classic two-player strategy game played on a 3x3 grid. The goal is to be the first player to align three of your marks, either X or O, horizontally, vertically, or diagonally. This simple yet elegant game tests players’ ability to anticipate, block, and plan moves ahead, making it a suitable environment for evaluating reasoning, prediction, and opponent modeling in large language models (LLMs).

---

Gameplay

• Players: 2

• Symbols: X and O

• Objective: Form a line of three of your symbols before your opponent.

• Board Layout:

   0 | 1 | 2
  ---+---+---
   3 | 4 | 5
  ---+---+---
   6 | 7 | 8
  

Players take turns selecting a cell by its index (0–8). The environment automatically validates moves and announces wins, losses, or draws.

---

Environment Details

• Environment Name: TicTacToe-v0
• Number of Players: 2
• Observation Type: Text-based description of board state and game messages
• Action Type: Integer index (0–8)
• Winning Condition: Three identical symbols in a row, column, or diagonal
• Termination: When a player wins or all cells are filled (draw)

---

LLM Evaluation Purpose

TicTacToe serves as a benchmark for:

• Strategic reasoning: planning moves and anticipating outcomes
• Opponent modeling: predicting and countering adversarial play
• Deterministic decision-making: consistent performance under clear rules

It is also a good starting environment for reinforcement learning or self-play fine-tuning of small or large language models.