Chess AI: How does a chess program "think"?

A brief introduction to the algorithms used to evaluate chess games and how computer programs use that information to play the game.

Frederico Jordan
github.com/fredericojordan
linkedin.com/in/frederico-jordan-68693052/

Transcript

1. Chess AI: How does a chess program “think”? Frederico Jordan
2. Agenda • Introduction • Board Representation • Evaluation • Search
3. Chess Game
4. Chess Automata 1770 - The Turk
5. Chess Engines 1996 - Garry Kasparov vs. Deep Blue
6. Chess Game Tree
7. Computer Chess History Claude Shannon 1950 - “Programming a Digital Computer for Playing Chess” Chess games ~ 10120 Chess positions ~ 1043 Atoms in the universe ~ 1080
8. Computer Chess History Claude Shannon Type A: “Brute Force” Search Type B: Selective Search
9. Board Representation - 8x8 Board
10. Board Representation - Bitboards
11. Chess Game Tree
12. Minimax
13. Chess Piece Relative Value 1 3 3 5 9
14. Chess Piece Relative Value 100 300 300 500 900
15. Mobility
16. Piece-Square Tables
17. Alpha-Beta Pruning
18. Iterative Deepening
19. Horizon Effect
20. Quiescence Search
21. Static Exchange Evaluation
22. Null Move Pruning
23. Transposition Tables
24. Endgame Tablebases Mate in 546
25. Bonus - Heatmap
26. Questions?
27. Thank you!