Minimax Algorithm in Game Theory | Set 1 (Introduction)

Minimax is a backtracking-based algorithm used in game theory and AI to determine the optimal move in competitive games by evaluating all possible future outcomes, assuming both players act optimally.

• Works in two-player, turn-based, zero-sum games
• Builds a decision process based on game states and outcomes
• Widely used in strategic games like Tic-Tac-Toe and Chess

Key Concepts

• Maximizer: Player who tries to maximize the final score or outcome
• Minimizer: Player who tries to minimize the final score or opponent’s advantage
• Game State: Representation of the current configuration of the game
• Utility Value: Numerical score assigned to terminal states to represent win, loss, or draw outcomes
• Game Tree: Tree structure representing all possible moves and resulting states
• Heuristic Evaluation: Function used to estimate the value of non-terminal states when full exploration is not possible

Working

• Constructs a game tree where nodes represent game states and edges represent possible moves
• Expands all possible future moves up to terminal states or a depth limit
• Assigns utility values to leaf nodes using game outcomes or heuristic evaluation
• Propagates values upward: maximizer selects maximum value, minimizer selects minimum value
• Final decision is made at the root node based on optimal play assumption by both players

Implementation

Implementing the Minimax algorithm using a simple game tree where leaf nodes represent final scores, and the algorithm finds the optimal value assuming both players play optimally.

Step 1: Importing Required Module

Using the math module to compute the depth of the game tree.

Step 2: Defining Minimax Function

This function recursively evaluates the game tree and returns the optimal value based on maximizing and minimizing turns.

Step 3: Defining game states

We define leaf node values representing final game outcomes.

Step 4: Computing the tree depth

Calculating the depth of the binary game tree using logarithm.

Step 5: Run the Minimax algorithm

We start evaluation from the root node assuming the maximizer plays first.

Output:

The optimal value is: 12

Advantages

• Guarantees optimal move selection under perfect play assumption
• Works well for deterministic, turn-based, zero-sum games
• Serves as the base for advanced techniques like Alpha-Beta Pruning
• Suitable for problems with limited search space

Limitations

• Exponential time complexity makes it inefficient for large game trees
• Explores all possible moves regardless of usefulness
• Not scalable for complex games without optimization
• Requires high computational resources for deeper searches