60 likes | 267 Views
Chess. Widely studied adversarial search problem Most people are familiar with the game, yet it is still complex enough to be interesting to address Existing expertise scale that allows comparison against other human players. Complexities. En Passant Skewers Checks and mates Solution
E N D
Chess • Widely studied adversarial search problem • Most people are familiar with the game, yet it is still complex enough to be interesting to address • Existing expertise scale that allows comparison against other human players
Complexities • En Passant • Skewers • Checks and mates • Solution • Abstract out the moves and game functions • Rely on complex heuristics to evaluate board state
Performance Problems • State-based search requires a lot of memory, recalculation • Solution: • Don’t store full states, only store changes • Mutator object – the Move object • A game state is the initial state plus a set of Moves applied to it • Only have to store one board, search tree of Move nodes • Can apply/undo moves relatively cheaply
Search Solution • State-based greedy search. • Heuristic based on piece values. • Decent AI against absolute beginners (people learning how each piece moves) • Extremely low ELO rating