Skip to content

azorfus/Tahl

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

77 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Tahl

An aggressive chess engine written using the Monte Carlo Tree Search with the aim of being trained to play like the legendary grandmaster Mikhail Tal

Input format for the Neural Network

We slice the board into 28 layers of 2D bit arrays to feed the network and expect a 12 layer 2D bit array output.

Legend: 1 = active, 0 = empty
Orientation: rank 8 (top) → rank 1 (bottom), files a → h


SLICES 1 - 4: Piece information

W Pawns W Rooks WKnights WBishops
00000000 10000001 01000010 00100100
11111111 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000

SLICES 5 - 8: Piece information

W Queen W King B Pawns B Rooks
00010000 00001000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 11111111 00000000
00000000 00000000 00000000 10000001

SLICES 9 - 12: Piece information

BKnights BBishops B Queen B King
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
01000010 00100100 00010000 00001000

SLICES 13 - 16: White Castling Rights and Castling availability

White KS White QS White KS White QS
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000

SLICES 17 - 20: Black Castling Rights and Castling availability

Black KS Black QS Black KS Black QS
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000
11111111 11111111 00000000 00000000

SLICES 21 - 24: En Passant

EP - En Passant, EPO - En Passant Origin

EP White EPOWhite EP Black EPOBLACK
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000

SLICES 25 - 28: Empty (Maybe for future use)

t-1 t-2 t-3 t-4
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000

About

An aggressive chess engine built on Monte Carlo Tree Search, designed to be trained with deep neural networks to emulate the playing style of the legendary grandmaster Mikhail Tal

Topics

Resources

License

Stars

Watchers

Forks

Contributors