presentation.org

Presentation - Positional analysis of chess games

Plot analysis

Heat map grid

Head map single

Histogram grid

Histogram single

Date based plots

Date based plots

Difficulties and issues

Pawn capture on 8th rank

Pawn capture on 8th rank

3.7e) When a pawn reaches the rank furthest from its starting position it must be exchanged as part of the same move on the same square for a new queen, rook, bishop or knight of the same colour. The player’s choice is not restricted to pieces that have been captured previously. This exchange of a pawn for another piece is called ‘promotion’ and the effect of the new piece is immediate.

Source: FIDE Laws of chess

Bug location

def piece_delta(board: chess.Board, count: int, piece_count: Dict[int, int],
                colour: bool) -> Tuple[int, int, int]:
    piece_position = (0, 0, 0)
    for key, value in piece_count.items():
        current_count = bin(board.pieces_mask(key, colour)).count('1')
        if current_count < value: # Detects lost based on previous state
            piece_position = (key, uci_to_1d_array_index(board.peek().uci()), count)
            piece_count[key] = current_count # Modify by object-reference
            break
        elif current_count > value: # Accounts for promotion
            piece_count[key] = current_count # Modify by object-reference
            piece_count[chess.PAWN] = bin(board.pieces_mask(chess.PAWN, colour)).count('1')  # Account for pawn count change
            break
    return piece_position # piece id, position, count

Bug location

board.pieces_mask(chess.PAWN, chess.BLACK)

Data structure

/	<							>
	0	0	0	0	0	0	0	0
	1	1	1	1	1	1	1	1
	0	0	0	0	0	0	0	0
	0	0	0	0	0	0	0	0
	0	0	0	0	0	0	0	0
	0	0	0	0	0	0	0	0
	0	0	0	0	0	0	0	0
	0	0	0	0	0	0	0	0

Board

Culprit

Inital state

Final state

Timezones

Performance

Performance

FISC

$ wc -l *
    3502985 ficsgamesdb_2000.pgn
    ...
   11126076 ficsgamesdb_2020.pgn
  452107755 total

$ du -h ./
17G

$ ls | wc -l
21

Tournements

$ wc -l *
     178 Aachen1868.pgn
     ...
    1196 Zurich2015.pgn
 1786360 total

$ du -h ./
64M

$ ls | wc -l
1059

Profiling

Improvements

Evaluation of positions using engines

Learning to Evaluate Chess Positions with Deep Neural Networks and Limited Look ahead

…We collect around 3,000,000 different chess positions played by highly skilled chess players and label them with the evaluation function of Stockfish, one of the strongest existing chess engines. We create 4 different datasets from scratch that are used for different classification and regression experiments. …

doi:10.5220/0006535502760283

Popular chess engines

• Stockfish
  • Strongest currently
  • Traditional
  • FOSS
• Leela Chess Zero
  • Self-taught through reinforcement learning and repeated self-play
  • FOSS
• AlphaZero
  • First engine to use reinforcement learning and self-play
  • Propriety
• Komodo
  • Propriety
  • Used by Chess.com for it’s celebrity bots

Failed plots

KDE plots

Final words

Final words

Source code: https://github.com/Jake-Moss/chess-analysis

Additions: Deletions:

Broke git once on Mar 7, 2021.

Everything was written in Emacs.