Back to Adversarial Search

MAX vs MIN Players: Tic-Tac-Toe

Interactive Game Theory Demonstration

Experience zero-sum games and adversarial decision making
X
MAX Player

Goal: Maximize Utility

Represents: Our AI

Seeks: +1 (Win)

O
MIN Player

Goal: Minimize Utility

Represents: Opponent

Seeks: -1 (Our Loss)

Game Board

MAX (X) always moves first. Click any empty cell!

True utilities only known at game end
Game in Progress

MAX Player (X) turn - Click any empty cell

Live Zero-Sum Utilities

Game in progress - utilities determined at end

MAX (X)
?
Current Utility
MIN (O)
?
Current Utility
Zero-Sum Check
? + ? = ?
Must always equal 0!

Three Perspectives on Game Utilities

True Utilities

When: Only at terminal states
Values: +1 (win), 0 (draw), -1 (loss)
Certainty: 100% known
Use: Final game scoring

Open Lines Evaluation

When: During gameplay
Values: Integer (-8 to +8)
Formula: X_lines - O_lines
Use: Strategic position analysis

Probability Evaluation

When: During gameplay
Values: Decimal (-1.0 to +1.0)
Formula: tanh(lines_diff / 3)
Use: Win likelihood estimation

🎓 Educational Point: Students can see how different evaluation approaches give different numerical ranges but maintain the same relative ordering for decision-making!

Key Adversarial Search Concepts

MAX Player

Represents the AI we're designing. Always moves first and tries to maximize the game's utility value. In tic-tac-toe, MAX wants to win (+1) and avoid losing (-1).

MIN Player

Represents our intelligent opponent. Responds to MAX's moves and tries to minimize the utility value. MIN wants MAX to lose, giving us -1 utility.

Turn-Taking

Players alternate moves until reaching a terminal state (win, loss, or draw). This creates a predictable sequence that algorithms can analyze.

Zero-Sum Nature

Total payoffs always sum to zero. One player's success is exactly balanced by the other's failure. This competitive relationship drives adversarial search.

Why This Matters for AI

AI Game Playing

Understanding MAX vs MIN is fundamental to game-playing AI. Every chess engine, checkers program, or game AI must handle adversarial opponents who actively work against them.

Minimax Algorithm

The minimax algorithm builds on these concepts, recursively exploring game trees where MAX levels maximize utility and MIN levels minimize it, finding optimal play for both sides.