Minimax Search

• Search nodes include the following data:
  • Board Configuration
  • Which player is about to move
• Starting node
  • Current board configuration
  • Current player
    • We will call the current player in the starting node the protagonist.
    • The protagonist's opponent we will call the adversary.
• Generating successor nodes
  • Select a legal move for the moving player
  • Generate a new board position resulting from that move
  • Make the other player the moving player
  • If the game is over, no successor nodes should be generated.
• Evaluating a node
  • Terminating nodes:
    • If the protagonist loses, return -1.
    • If the protagonist wins, return 1.
    • If the game is a tie, return 0.
  • Interior nodes:
    • If the current player is the protagonist, return the maximum value retrieved from a successor node.
    • If the current player is the adversary, return the minimum value retrieved from a successor node.
• Evaluation Functions
  • Due to depth cutoffs, terminal nodes are not always reached.
  • A board evaluation function provides an estimate of whether a node will lead to an eventual win or loss.
  • This estimate is a real number between -1 and 1.