Chapter 6.1 - 6.4

Games

• Typically, AI games are two-player, turn taking, zero-sum and perfect information.
• A game is a type of search problem that has an initial state, a successor function, a terminal test, and a utility function.
• Game Trees. Figure 6.1.
• An optimal strategy leads to outcomes at least as good as any other strategy when playing an infallible opponent.
• Minimax. Figures 6.2 and Figure 6.3.
• Multiplayer games. Figure 6.4.

Alpha-Beta Pruning

• Figure 6.5 and Figure 6.6.
• α - the best choice so far for MAX.
• β - the best choice so far for MIN.
• Algorithm: Figure 6.7.

Imperfect, Real-Time Decisions

• Replace the terminal test with a cutoff test.
• Replace the utility function with an evaluation functions. Evaluation functions should be accurate and easy to compute.
• Quiescence.
• Horizon Effect.
• Forward Pruning.
• Opening Move Database/Procedures.
• End Game Database/Procedures.

Deep Blue

Exercises

6.1, 6.3, 6.13