• Buğra Fırat 2016, A Visual Look at 2 Million Chess Games, GitHub @ebemunk
  • Visualize the opening tree (very informative!)
  • White wins more often than Black, 39% to 30%, the rest being draws
  • Black was more likely to castle queenside, but White was more likely to not castle at all!
  • Average game length looks like a normal distribution with a mean somewhere around 70-ply (35 moves)
  • Average Material Difference per Ply fluctuates between +/- 0.50
  • Heatmaps for individual pieces are most interesting
    • Square Utilization: the squares that a piece moves to
    • Move Squares: squares that a piece moved from
    • Capture Squares: squares where a capture occurred
    • Checking Squares: squares where the move resulted in a check
  • Source code: parse PGN files and extract statistics, d3.js chess-dataviz

• AlphaZero vs. Stockfish 2017, Self-play Reinforcement Learning masters Chess
  • Superb YouTube commentaries by Jerry, aka ChessNetwork
  • Stockfish open source development: https://github.com/official-stockfish/Stockfish

• Lichess.org REST API use curl to retrieve info by https URL.

• Karsten Müller, Fundamental Chess Endgames, excellent if you find Dvoretsky to be too heavy (Magnus Carlsen uses it in his studies).

Shortcut to his page: https://git.io/chessref | Last update : 2017-12-28